RigPlanner · My stack (Plan) · 171 of 209 components evaluable, 137 priced

No rig to recommend yet — none of the 89 candidate rigs runs every model in "My stack (Plan)" at ≥ 15 tok/s.

NO VIABLE RIG

MEDIUM CONFIDENCE

Every deduction that set this confidence

Start at high (2 points).
−1: 30 daily closes in the window (< 60).
= medium.

Searched 89 candidate rigs built from 171 evaluable components, 137 of them priced, over 8,739 price observations. Every number below expands to the arithmetic that produced it.

Since yesterdayNo comparison yet — this screen states today's position only; it does not claim the recommendation is steady.compared with the same search priced from the append-only history as of Oct 17, 6:00 PM ET (24 h ago)

Why there is nothing to compare

There is no viable rig today, so there is nothing to compare with yesterday's.
Needs: a viable rig in the catalog.

What moved

showing 10 of 171, ranked by membership of the recommendation, then by the size of the 24 h move · the rest are on the Prices screen

What moved: 10 of 171 priced parts, ranked by membership of the current recommendation, then by the size of the 24-hour move on the daily-close series
Part	Close	Δ 24 h	Δ 7 d	Pctile	Close observed	Close: retailer · condition · route
NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GBgpu	$10,795 (observed) OBSERVEDNVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $10,795.47 observed 2026-10-18T06:42:05.271Z observed_at 2026-10-18T06:42:05.271Z series 10:offer293:list age_days 0.6378196759259259 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 102 observations → 30 daily closes, 7 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 6 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 10:offer293:list), 0.6 days old at evaluation time.	down (better for a buyer) −$4,705 (derived) DERIVEDNVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $10,795.47 − $15,499.99 = −$4,704.52 latest_close 1079547 cents — day 2026-10-18, series 10:offer293:list, 4 series closed that day prior_close 1549999 cents — day 2026-10-17, series 7:offer127:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	down (better for a buyer) −$4,105 (derived) DERIVEDNVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $10,795.47 − $14,899.99 = −$4,104.52 latest_close 1079547 cents — day 2026-10-18, series 10:offer293:list, 4 series closed that day prior_close 1489999 cents — day 2026-10-11, series 7:offer127:list, 3 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDNVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 1079547 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 102 observations → 30 daily closes, 7 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 6 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 10:offer293:list), 0.6 days old at evaluation time.	15 h	SabrePCcondition newaccess tier 2, route sabrepc_ngstatealso the cheapest live offerlowest of 4 series that closed that day
AMD EPYC 9755cpu	$5,299 (observed) OBSERVEDAMD EPYC 9755 latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $5,299.00 observed 2026-10-18T20:15:50.819Z observed_at 2026-10-18T20:15:50.819Z series 1:offer35:sale age_days 0.07270916666666667 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 89 observations → 30 daily closes, 9 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 5 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer35:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$1,701 (derived) DERIVEDAMD EPYC 9755 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $5,299.00 − $6,999.99 = −$1,700.99 latest_close 529900 cents — day 2026-10-18, series 1:offer35:sale, 2 series closed that day prior_close 699999 cents — day 2026-10-17, series 7:offer37:list, 2 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD EPYC 9755 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $5,299.00 − $5,299.00 = $0.00 latest_close 529900 cents — day 2026-10-18, series 1:offer35:sale, 2 series closed that day prior_close 529900 cents — day 2026-10-11, series 1:offer35:sale, 3 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDAMD EPYC 9755 percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 529900 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 89 observations → 30 daily closes, 9 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 5 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer35:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
AMD EPYC 8224Pcpu	$2,458 (observed) OBSERVEDAMD EPYC 8224P latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $2,458.35 observed 2026-10-18T20:15:50.949Z observed_at 2026-10-18T20:15:50.949Z series 1:offer284:list age_days 0.07270766203703703 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 56 observations → 28 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 2 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer284:list), 0.1 days old at evaluation time.	up (worse for a buyer) +$1,478 (derived) DERIVEDAMD EPYC 8224P 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $2,458.35 − $980.40 = +$1,477.95 latest_close 245835 cents — day 2026-10-18, series 1:offer284:list, 1 series closed that day prior_close 98040 cents — day 2026-10-16, series 10:offer63:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD EPYC 8224P 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $2,458.35 − $2,458.35 = $0.00 latest_close 245835 cents — day 2026-10-18, series 1:offer284:list, 1 series closed that day prior_close 245835 cents — day 2026-10-11, series 1:offer284:list, 1 series closed that day closes_between 5 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	93% · n 28 (derived) DERIVEDAMD EPYC 8224P percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 25 / (28 − 1) = 0.9259 (92.6%) n 28 daily closes current 245835 cents closes_below_current 25 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 56 observations → 28 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 2 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer284:list), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtime
Cheapest live offer$980shipping not published by this retailer — $1,478 below the closeSabrePCcondition newaccess tier 2, route sabrepc_ngstatestale 2 d — past the 36 hour freshness window
AMD EPYC 9575Fcpu	$4,950 (observed) OBSERVEDAMD EPYC 9575F latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $4,950.00 observed 2026-10-18T20:15:50.800Z observed_at 2026-10-18T20:15:50.800Z series 1:offer29:sale age_days 0.07270938657407408 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 87 observations → 30 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 4 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer29:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$1,450 (derived) DERIVEDAMD EPYC 9575F 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $4,950.00 − $6,399.99 = −$1,449.99 latest_close 495000 cents — day 2026-10-18, series 1:offer29:sale, 3 series closed that day prior_close 639999 cents — day 2026-10-17, series 7:offer322:list, 2 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD EPYC 9575F 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $4,950.00 − $4,950.00 = $0.00 latest_close 495000 cents — day 2026-10-18, series 1:offer29:sale, 3 series closed that day prior_close 495000 cents — day 2026-10-11, series 1:offer29:sale, 3 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDAMD EPYC 9575F percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 495000 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 87 observations → 30 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 4 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer29:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 3 series that closed that day
AMD Ryzen Threadripper PRO 7995WXcpu	$9,900 (observed) OBSERVEDAMD Ryzen Threadripper PRO 7995WX latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $9,899.99 observed 2026-10-18T20:15:50.877Z observed_at 2026-10-18T20:15:50.877Z series 1:offer89:sale age_days 0.07270849537037037 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 59 observations → 30 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer89:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$1,045 (derived) DERIVEDAMD Ryzen Threadripper PRO 7995WX 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $9,899.99 − $10,945.14 = −$1,045.15 latest_close 989999 cents — day 2026-10-18, series 1:offer89:sale, 2 series closed that day prior_close 1094514 cents — day 2026-10-17, series 10:offer90:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD Ryzen Threadripper PRO 7995WX 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $9,899.99 − $9,899.99 = $0.00 latest_close 989999 cents — day 2026-10-18, series 1:offer89:sale, 2 series closed that day prior_close 989999 cents — day 2026-10-11, series 1:offer89:sale, 2 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDAMD Ryzen Threadripper PRO 7995WX percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 989999 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 59 observations → 30 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer89:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
AMD Ryzen Threadripper PRO 9995WXcpu	$12,081 (observed) OBSERVEDAMD Ryzen Threadripper PRO 9995WX latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $12,080.99 observed 2026-10-18T20:15:50.868Z observed_at 2026-10-18T20:15:50.868Z series 1:offer77:sale age_days 0.07270859953703704 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 57 observations → 30 daily closes, 3 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer77:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$978 (derived) DERIVEDAMD Ryzen Threadripper PRO 9995WX 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $12,080.99 − $13,058.67 = −$977.68 latest_close 1208099 cents — day 2026-10-18, series 1:offer77:sale, 2 series closed that day prior_close 1305867 cents — day 2026-10-17, series 10:offer78:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD Ryzen Threadripper PRO 9995WX 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $12,080.99 − $12,080.99 = $0.00 latest_close 1208099 cents — day 2026-10-18, series 1:offer77:sale, 2 series closed that day prior_close 1208099 cents — day 2026-10-11, series 1:offer77:sale, 2 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	28% · n 30 (derived) DERIVEDAMD Ryzen Threadripper PRO 9995WX percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 8 / (30 − 1) = 0.2759 (27.6%) n 30 daily closes current 1208099 cents closes_below_current 8 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 57 observations → 30 daily closes, 3 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer77:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
AMD Ryzen Threadripper PRO 9975WXcpu	$3,900 (observed) OBSERVEDAMD Ryzen Threadripper PRO 9975WX latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $3,899.99 observed 2026-10-18T20:15:50.826Z observed_at 2026-10-18T20:15:50.826Z series 1:offer74:sale age_days 0.07270908564814815 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 55 observations → 30 daily closes, 3 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer74:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$670 (derived) DERIVEDAMD Ryzen Threadripper PRO 9975WX 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $3,899.99 − $4,569.80 = −$669.81 latest_close 389999 cents — day 2026-10-18, series 1:offer74:sale, 2 series closed that day prior_close 456980 cents — day 2026-10-17, series 10:offer75:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD Ryzen Threadripper PRO 9975WX 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $3,899.99 − $3,899.99 = $0.00 latest_close 389999 cents — day 2026-10-18, series 1:offer74:sale, 2 series closed that day prior_close 389999 cents — day 2026-10-11, series 1:offer74:sale, 2 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDAMD Ryzen Threadripper PRO 9975WX percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 389999 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 55 observations → 30 daily closes, 3 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer74:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
Intel Xeon 6787Pcpu	$8,350 (observed) OBSERVEDIntel Xeon 6787P latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $8,349.99 observed 2026-10-18T20:15:50.881Z observed_at 2026-10-18T20:15:50.881Z series 1:offer95:list age_days 0.07270844907407407 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 60 observations → 30 daily closes, 9 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 2 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer95:list), 0.1 days old at evaluation time.	down (better for a buyer) −$542 (derived) DERIVEDIntel Xeon 6787P 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $8,349.99 − $8,892.08 = −$542.09 latest_close 834999 cents — day 2026-10-18, series 1:offer95:list, 2 series closed that day prior_close 889208 cents — day 2026-10-17, series 10:offer290:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDIntel Xeon 6787P 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $8,349.99 − $8,349.99 = $0.00 latest_close 834999 cents — day 2026-10-18, series 1:offer95:list, 2 series closed that day prior_close 834999 cents — day 2026-10-11, series 1:offer95:list, 2 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDIntel Xeon 6787P percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 834999 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 60 observations → 30 daily closes, 9 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 2 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer95:list), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
GIGABYTE GeForce RTX 5080 WINDFORCE OC 16G (GV-N5080WF3OC-16GD)gpu	$1,700 (observed) OBSERVEDGIGABYTE GeForce RTX 5080 WINDFORCE OC 16G (GV-N5080WF3OC-16GD) latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $1,699.99 observed 2026-10-18T20:15:50.893Z observed_at 2026-10-18T20:15:50.893Z series 1:offer138:sale age_days 0.07270831018518518 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 55 observations → 28 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer138:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$494 (derived) DERIVEDGIGABYTE GeForce RTX 5080 WINDFORCE OC 16G (GV-N5080WF3OC-16GD) 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $1,699.99 − $2,193.52 = −$493.53 latest_close 169999 cents — day 2026-10-18, series 1:offer138:sale, 2 series closed that day prior_close 219352 cents — day 2026-10-15, series 10:offer297:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDGIGABYTE GeForce RTX 5080 WINDFORCE OC 16G (GV-N5080WF3OC-16GD) 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $1,699.99 − $1,699.99 = $0.00 latest_close 169999 cents — day 2026-10-18, series 1:offer138:sale, 2 series closed that day prior_close 169999 cents — day 2026-10-11, series 1:offer138:sale, 2 series closed that day closes_between 4 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	56% · n 28 (derived) DERIVEDGIGABYTE GeForce RTX 5080 WINDFORCE OC 16G (GV-N5080WF3OC-16GD) percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 15 / (28 − 1) = 0.5556 (55.6%) n 28 daily closes current 169999 cents closes_below_current 15 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 55 observations → 28 daily closes, 6 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer138:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day
AMD EPYC 9555Pcpu	$5,007 (observed) OBSERVEDAMD EPYC 9555P latest daily closeMeasured or published, passed through unchanged — not computed here. current = latest daily close in the window current = $5,006.99 observed 2026-10-18T20:15:50.658Z observed_at 2026-10-18T20:15:50.658Z series 1:offer25:sale age_days 0.0727110300925926 days Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 57 observations → 30 daily closes, 4 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer25:sale), 0.1 days old at evaluation time.	down (better for a buyer) −$420 (derived) DERIVEDAMD EPYC 9555P 24 h move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ24 h = latest daily close − the daily close as of 24 h ago $5,006.99 − $5,427.00 = −$420.01 latest_close 500699 cents — day 2026-10-18, series 1:offer25:sale, 2 series closed that day prior_close 542700 cents — day 2026-10-17, series 10:offer26:list, 1 series closed that day closes_between 0 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 24 h ago; if the series has a gap, that close may be older than 24 h. A rising price is worse for a buyer; the sign is stated from the buyer's side.	unchanged: $0 (derived) DERIVEDAMD EPYC 9555P 7 d move on the daily-close seriesArithmetic on observed inputs only. No tunable assumption enters it. Δ7 d = latest daily close − the daily close as of 7 d ago $5,006.99 − $5,006.99 = $0.00 latest_close 500699 cents — day 2026-10-18, series 1:offer25:sale, 2 series closed that day prior_close 500699 cents — day 2026-10-11, series 1:offer25:sale, 2 series closed that day closes_between 6 daily closes — the series between the two ends of this window Both ends are daily closes: the last observation per retailer per UTC day, then the lowest of them, so polling frequency does not weight a retailer. The prior end is the last close AT OR BEFORE 7 d ago; if the series has a gap, that close may be older than 7 d. A rising price is worse for a buyer; the sign is stated from the buyer's side.	0% · n 30 (derived) DERIVEDAMD EPYC 9555P percentile in its own 90-day rangeArithmetic on observed inputs only. No tunable assumption enters it. rank = k / (n − 1) where k = #{daily closes < current} (inverse of the type-7 quantile; ties take the lowest rank, as Excel PERCENTRANK.INC) rank = 0 / (30 − 1) = 0 (0%) n 30 daily closes current 500699 cents closes_below_current 0 Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days). Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 57 observations → 30 daily closes, 4 superseded by a later same-day observation. 0 quarantined observations excluded (SPEC §5.6). 3 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close). Current = the latest daily close, 2026-10-18 (series 1:offer25:sale), 0.1 days old at evaluation time.	1 h	Neweggcondition newaccess tier 3, route newegg_realtimealso the cheapest live offerlowest of 2 series that closed that day

Close, both deltas, the 90 d line and the percentile are measured on ONE series: the daily close — the last observation per retailer per UTC day, then the lowest of them, so a retailer polled twelve times a day does not outweigh one polled once. The retailer, condition, route and age in this table belong to the observation that SET that close. The cheapest offer available right now is a different series and appears on its own line beneath the row; the two are never merged into one figure. Where the window is too narrow for every column, the table scrolls sideways inside this panel — no column is dropped and nothing is painted over the panel beside it.

Recommended rig

0/89 viable of 89 candidates

reverse mode — 89 of 89 candidates evaluated, 0 viable

No candidate rig runs every model in this Plan at the target speed. Showing a build here would mean recommending something that does not do the job.

npm run worker -- nightly

1333 platform combinations were collapsed. For each socket and memory size the search carries the CHEAPEST build and, when it is a different part, the one with the most CPU compute (which is what prefill depends on); the rest differ only in price.
8 CPU+GPU hybrid candidates were the same parts as a GPU rig already in the list and are listed once.
No candidate rig meets every Plan model at ≥ 15 tok/s. Most common blockers: DeepSeek-V4-Pro-0813: No quantization chosen or sized. (89 rigs); Qwen3-ASR 1.7B: No quantization chosen or sized. (89 rigs); Qwen3-VL 32B Instruct: No quantization chosen or sized. (89 rigs).

Buy / wait signal

weights 0.50 / 0.30 / 0.20 · buy p10 / wait p75

How the buy/wait signal for NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB was composed: each term with its score, its configured weight, its effective weight after renormalisation, and the data it was computed from
Term of the signal	Score	Weight	Effective	Basis
Trend, 90 d	no push: 0.000	0.50	0.50	0.0 %/30 d, Theil–Sen on ln price over 30 daily closes spanning 29 d
News, 30 d	argues prices rise: 1.000	0.30	0.30	304 classified items, weighted by source tier and half-life · 3 item(s) flagged equity_angle were excluded: a share-price story is not a component-price story.
Product cycle	no push: 0.000	0.20	0.20	no successor on record for this part
Composite	0.300 (estimate) ESTIMATEComposite market signalArithmetic that also depends on a tunable assumption, listed below. market_signal = w_trend·trend_90d + w_news·news_30d + w_cycle·cycle market_signal = 0.5×trend_90d(0) + 0.3×news_30d(+1) + 0.2×cycle(0) = +0.3 trend_90d 0 — weight 0.5 news_30d 1 — weight 0.3 cycle 0 — weight 0.2 Positive = the market points to higher prices (buy sooner); negative = lower prices ahead (waiting pays). Weights from Settings.signal: trend 0.5, news 0.3, cycle 0.2.	1.00		BUY NOW · medium rule: buy_now_price_and_news

Scored on NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 96GB (gpu) — the part at the top of the movers table — there is no recommended rig to take a cost driver from, so the signal is scored on the part that moved most. Buy below

$10,795

DERIVEDBuy thresholdArithmetic on observed inputs only. No tunable assumption enters it.

p10 = type-7 quantile at p = 0.1: h = (n − 1)·p + 1, Q = x(⌊h⌋) + (h − ⌊h⌋)·(x(⌊h⌋+1) − x(⌊h⌋))

h = (30 − 1) × 0.1 + 1 = 3.9 → p10 = x(3) + 0.9 × (x(4) − x(3)) = $10,795.47 + 0.9 × ($10,795.47 − $10,795.47) = $10,795.47

n: 30 daily closes
p: 0.1

Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days).
Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 102 observations → 30 daily closes, 7 superseded by a later same-day observation.
0 quarantined observations excluded (SPEC §5.6).
6 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close).

, wait above

$10,795

DERIVEDWait thresholdArithmetic on observed inputs only. No tunable assumption enters it.

p75 = type-7 quantile at p = 0.75: h = (n − 1)·p + 1, Q = x(⌊h⌋) + (h − ⌊h⌋)·(x(⌊h⌋+1) − x(⌊h⌋))

h = (30 − 1) × 0.75 + 1 = 22.75 → p75 = x(22) + 0.75 × (x(23) − x(22)) = $10,795.47 + 0.75 × ($10,795.47 − $10,795.47) = $10,795.47

n: 30 daily closes
p: 0.75

Window (2026-07-20T22:00Z, 2026-10-18T22:00Z] (90 days).
Daily close = last observation per UTC day per price series (so polling frequency does not weight a retailer); 102 observations → 30 daily closes, 7 superseded by a later same-day observation.
0 quarantined observations excluded (SPEC §5.6).
6 price series; each day's price is the lowest close among the series reporting that day (the best price available at that day's close).

BUY NOW — Current $10,795.47 ≤ 10th percentile $10,795.47 (30 daily closes, last 90 days), and news is neutral-to-rising: 30-day news sentiment +1 > -0.25 (neutral-to-rising means not falling) (SPEC §6.4).
Market signal +0.3: market_signal = 0.5×trend_90d(0) + 0.3×news_30d(+1) + 0.2×cycle(0) = +0.3. Positive = prices expected higher.
Shortage context (SPEC §6.5): flagged shortage_affected, but its own 90-day trend is not rising (0% per 30 days) — the shortage is not visible in this part's price right now.

Verdict vs the offer you can actually buy

The cheapest live offer (SabrePC, new) is $10,795 — $0 ABOVE the $10,795 wait threshold, while the verdict on the close series is BUY NOW. The cheapest thing you can actually buy today is priced in wait territory even though the window is not; the verdict describes the window.

The verdict is computed on the daily-close series: one close per retailer per UTC day, then the lowest of them. A live offer is a different series — what you would pay this minute — and it does not enter the distribution the thresholds are drawn from until its day closes. Where the two disagree the verdict describes the window and the live offer describes today; neither is silently preferred, and both are shown with the series they came from.

That live offer came fromSabrePCcondition newaccess tier 2, route sabrepc_ngstate15 h

Headlines behind the news term

This week on Tom's Hardware Premium: October 17, 2026 tomshardware.com · tier 2 · neutral
Leap seconds being retired next year to avoid wreaking havoc on computer systems tomshardware.com · tier 2 · neutral
Taiwan targets illicit AI server exports to China with five-year prison terms tomshardware.com · tier 2 · neutral
Crafty OpenAI agent willingly stole crypto from an account it didn't own to complete a task during test tomshardware.com · tier 2 · neutral
ASUS Finally Builds An Intel Arc Graphics Card After Two Generations On The Sidelines, And It’s The Full 32 GB Big Battlemage Arc Pro B70 wccftech.com · tier 2 · neutral

Dated catalysts 2026-09-30 Micron FQ4 FY26 results and call (guidance was $50.0B ± $1.0B revenue, about 86% gross margin) (confirmed) · 2026-10-05 Samsung Q3 2026 preliminary earnings guidance (date derived from Samsung's own IR calendar, Oct 5-14 window) (estimated) · 2026-10-20 NVIDIA GTC Berlin 2026 (confirmed) · 2026-10-27 Samsung Q3 2026 earnings conference call (date derived from Samsung's own IR calendar, Oct 27-31 window) (estimated)

Plan fit

5 models · 32,768 ctx · swap on demand

rig outcomes — 0 rows

There is no rig to run the Plan's models on, so none of them has an outcome.

npm run worker -- nightly It re-prices the catalog; the search then runs again on the new observations.

What would change this

in the verdict engine’s own precedence

To becomeWAITany one requirement below is enough
- not yetObserved daily close rises above $10,795 — $0 above today's $10,795.
- not yetA successor lands inside the cycle horizon — Cycle proximity 0 > -0.5 (no successor known)
- not yet30-day news sentiment drops to −0.25 or below — 30-day news sentiment +1 > -0.25 (304 item(s); drivers 2561, 2501, 2754)

Open items

0 gaps · 81 unread · 38 incomplete

Alerts the app raised, newest first
Alert	Kind	Raised
NEW Sign in again to keep account pricesbrowser:cdw, browser:newegg	needs_reauth	14 h
NEW 13 price(s) are staleOlder than 36 h: Central Computers, SabrePC, CDW, Newegg	stale_data	14 h
NEW 6 price(s) quarantinedidentity conflict on mpn: listing shows 100100000725WOF, component has 100000000725 identity conflict on mpn: listing shows 100100000722WOF, component has 100000000722\|100100000722WOZ identity conflict on mpn: listing shows 100000001593WOF, component has 100000001593\|100100001593WOF identity conflict on mpn: listing shows R9700CT32G, component has R9700CT identity conflict on mpn: listing shows 25FC9RTX003P2R, component has VCNRTXPRO6000BQPB\|VCNRTXPRO6000BQBLK\|9005G153220000001 identity conflict on mpn: listing shows TGM100000001136, component has 100000001136\|100000001136	quarantine	14 h
NEW 129 source failure(s) in tonight's run3 blocked, 13 parse, 3 internal, 8 http_status, 49 identifier_unresolved, 12 credentials_missing, 1 schema_mismatch, 29 calculation, 2 rate_limited, 1 cross_source_conflict, 4 robots_disallowed, 2 needs_reauth, 2 sanity_bounds	source_failure	14 h
NEW 75 price(s) are staleOlder than 36 h: Newegg, Central Computers, SabrePC	stale_data	32 h
NEW 129 source failure(s) in tonight's run3 blocked, 23 parse, 8 http_status, 3 internal, 53 identifier_unresolved, 10 credentials_missing, 4 schema_mismatch, 21 calculation, 3 robots_disallowed, 1 needs_reauth	source_failure	32 h
NEW 75 price(s) are staleOlder than 36 h: Newegg, Central Computers	stale_data	2 d
NEW 123 source failure(s) in tonight's run3 blocked, 18 parse, 7 http_status, 1 network, 3 internal, 53 identifier_unresolved, 10 credentials_missing, 3 schema_mismatch, 22 calculation, 1 needs_reauth, 2 robots_disallowed	source_failure	2 d

38 components cannot be evaluated yet — showing the first 6

Components in the catalog whose spec is incomplete, with the fields that are missing and the sources already tried
Component	What is missing, and where it was looked for
AMD EPYC 9016cpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists
AMD EPYC 9116cpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists
AMD EPYC 9176Fcpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists
AMD EPYC 9256cpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists
AMD EPYC 9336cpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists
AMD EPYC 9356cpu	max_memory_gb · 3 sources tried Missing fields: max_memory_gb Sources tried: AMD SKU page for the part (amd.com/.../9006-series/amd-epyc-<sku>.html): System Memory Type, Memory Channels and System Memory Specification only · AMD EPYC 9006 series page (amd.com/en/products/processors/server/epyc/9006-series.html): no capacity figure · AMD server processor specification database (amd.com/en/products/specifications/server-processor.html): no max-memory field exists

Source health

8,739 observations · 345 monitored offers · 209 components (171 complete) · freshness window 36 h

Every price adapter, the access tier and route it last used, when it last succeeded, and what it last failed with
Source	Status	Tier / route	Last success	Fails in a row	Parse confidence	Last error
dataforseo_shopping	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	55413	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	DataForSEO Merchant (Google Shopping) needs credentials that .env does not have: DATAFORSEO_LOGIN, DATAFORSEO_PASSWORD
nvidia_marketplace	FAILING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	14981	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	NVIDIA partner search response does not match the documented shape: searchedProducts.productDetails.0.minShipDays: Invalid input: expected number, received null; searchedProducts.productDetails.0.maxShipDays: Invalid input: expected number, received null; searchedProducts.productDetails.1.minShipDays: Invalid input: expected number, received null
bestbuy_api	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	4987	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	Best Buy Developer API needs a credential that .env does not have: BESTBUY_API_KEY
keepa	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	4972	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	Keepa Product API needs a credential that .env does not have: KEEPA_API_KEY
ebay_browse	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	4971	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	eBay Browse API needs credentials that .env does not have: EBAY_CLIENT_ID, EBAY_CLIENT_SECRET
brightdata_dataset	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	75	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	Bright Data Web Scraper API needs a credential that .env does not have: BRIGHTDATA_API_TOKEN
zyte_extract	CREDENTIALS MISSING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	15	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	Zyte API product extraction needs a credential that .env does not have: ZYTE_API_KEY
zones_html	FAILING	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	14	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	No product title on https://www.zones.com/site/product/index.html?id=301737277NEW
zotac_store_html	FAILING	t2 / zotac_store_html	stale 23 d — past the 36 hour freshness window	11	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	timed out after 30000 ms (policy.timeoutMs) waiting for https://www.zotacstore.com/us/graphics-cards/geforce-rtx-50-series/geforce-rtx-5090
browser:cdw	NEEDS REAUTH	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	1	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	The signed-in state of cdw was not established in this run, so Tier 6 will not read a price that might be a signed-out one.
browser:newegg	NEEDS REAUTH	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	1	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	The signed-in state of newegg was not established in this run, so Tier 6 will not read a price that might be a signed-out one.
walmart_next_data	BLOCKED	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	1	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	https://www.walmart.com/ip/PNY-GeForce-RTX-5090-Overclocked-Triple-Fan-GPU-DLSS4/15046623228 answered HTTP 200 with a bot wall (HUMAN (PerimeterX) verification wall) instead of the page; the __NEXT_DATA__ script is absent.
apple_refurb	HEALTHY	t3 / apple_refurb	15 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
apple_store	HEALTHY	t3 / apple_store	1 h	0	95%	no error recorded
arctic_jsonld	HEALTHY	t1 / arctic_jsonld	1 h	0	90%	no error recorded
browser:best-buy	NEVER RUN	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
browser:bh-photo	NEVER RUN	Unavailable — This adapter has not completed a fetch, so no access tier or route is on record for it. (missing: a successful fetch)	Unavailable — No fetch through this adapter has ever succeeded. (missing: a successful fetch)	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cdw_html	HEALTHY	t2 / cdw_html	1 h	0	90%	no error recorded
cloud:aws	HEALTHY	t3 / json:b0.p.awsstatic.com/ec2-ondemand/us-east-1/linux	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:azure	HEALTHY	t0 / api:prices.azure.com/api/retail/prices	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:coreweave	HEALTHY	t2 / html:www.coreweave.com/pricing	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:crusoe	HEALTHY	t2 / html:www.crusoe.ai/cloud/pricing	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:digitalocean	HEALTHY	t2 / html:www.digitalocean.com/pricing/gpu-droplets	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:gcp	HEALTHY	t2 / html:cloud.google.com/products/compute/pricing	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:hetzner	HEALTHY	t3 / json:www.hetzner.com/live_data_prices.json	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:hyperbolic	HEALTHY	t2 / html:www.hyperbolic.ai/marketplace	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:lambda	HEALTHY	t2 / html:lambda.ai/pricing	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:nebius	HEALTHY	t3 / json:nebius.com/prices#__NEXT_DATA__	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:oci	HEALTHY	t0 / api:apexapps.oracle.com/cetools/api/v1/products	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:openrouter	HEALTHY	t0 / api:openrouter.ai/api/v1/models	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:paperspace	HEALTHY	t2 / html:www.paperspace.com/pricing#h100-footnote	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:runpod	HEALTHY	t0 / api:api.runpod.io/graphql#gpuTypes	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:together_gpu	HEALTHY	t2 / html:www.together.ai/pricing#gpu-clusters	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:together_serverless	HEALTHY	t2 / html:docs.together.ai/docs/serverless/models	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
cloud:vast	HEALTHY	t0 / api:console.vast.ai/api/v0/bundles	1 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
exxact_ngstate	HEALTHY	t1 / exxact_ngstate	15 h	0	50%	no error recorded
framework_html	HEALTHY	t2 / framework_html	1 h	0	80%	no error recorded
itcreations_html	HEALTHY	t2 / itcreations_html	15 h	0	Unavailable — No parse has completed through this adapter, so there is no confidence to average. (missing: a parsed response)	no error recorded
jsonld_product	HEALTHY	t1 / jsonld_product	1 h	0	80%	no error recorded
microdata_product	HEALTHY	t1 / microdata_product	1 h	0	75%	no error recorded
newegg_realtime	HEALTHY	t3 / newegg_realtime	1 h	0	95%	no error recorded
sabrepc_ngstate	HEALTHY	t2 / sabrepc_ngstate	1 h	0	90%	no error recorded
shopify_store	HEALTHY	t3 / shopify_store	1 h	0	95%	no error recorded