News Intel May Be Readying up to 32 Arc Alchemist GPU Models

InvalidError

Titan
Moderator
It might also be just planning ahead for the next 2-3 generations.
To me, it seems improbable that Intel would lump next-gen hardware with a block of old-gen PCI IDs.

Between different memory sizes (4/8GB, 6/12GB, 8/16GB, etc.) for a given GPU trim, possibly different VRAM speed grades and other feature trims, it wouldn't be too hard to burn through 32 IDs.
 
Reactions: thisisaname
Aug 31, 2021
18
15
15
0
It would be helpful to mention how many Ampere and RDNA2 PCI IDs there are. A lot of people may take the wrong thing away from this story.
 

Chung Leong

Respectable
Dec 6, 2019
425
145
1,860
0
It would be helpful to mention how many Ampere and RDNA2 PCI IDs there are. A lot of people may take the wrong thing away from this story.
There're 61 Ampere devices, not counting cards with GA100. Quite a few are engineering samples.

2200
GA102​
2204
GA102 [GeForce RTX 3090]​
2205
GA102 [GeForce RTX 3080 Ti 20GB]​
2206
GA102 [GeForce RTX 3080]​
2208
GA102 [GeForce RTX 3080 Ti]​
220d
GA102 [CMP 90HX]​
2216
GA102 [GeForce RTX 3080 Lite Hash Rate]​
222b
GA102 [GeForce RTX 3090 Engineering Sample]​
222f
GA102 [GeForce RTX 3080 11GB / 12GB Engineering Sample]​
2230
GA102GL [RTX A6000]​
2231
GA102GL [RTX A5000]​
2235
GA102GL [A40]​
2236
GA102GL [A10]​
2237
GA102GL [A10G]​
223f
GA102GL​
228b
GA104 High Definition Audio Controller​
2302
GA103​
2321
GA103​
2420
GA103M [GeForce RTX 3080 Ti Mobile]​
2482
GA104 [GeForce RTX 3070 Ti]​
2483
GA104​
2484
GA104 [GeForce RTX 3070]​
2486
GA104 [GeForce RTX 3060 Ti]​
2487
GA104 [GeForce RTX 3060]​
2488
GA104 [GeForce RTX 3070 Lite Hash Rate]​
2489
GA104 [GeForce RTX 3060 Ti Lite Hash Rate]​
248a
GA104 [CMP 70HX]​
249c
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
249d
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
249f
GA104M​
24a0
GA104 [Geforce RTX 3070 Ti Laptop GPU]​
24ac
GA104 [GeForce RTX 30x0 Engineering Sample]​
24ad
GA104 [GeForce RTX 3060 Engineering Sample]​
24af
GA104 [GeForce RTX 3070 Engineering Sample]​
24b0
GA104GL [RTX A4000]​
24b6
GA104GLM [RTX A5000 Mobile]​
24b7
GA104GLM [RTX A4000 Mobile]​
24b8
GA104GLM [RTX A3000 Mobile]​
24bf
GA104 [GeForce RTX 3070 Engineering Sample]​
24dc
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
24dd
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
2501
GA106 [GeForce RTX 3060]​
2503
GA106 [GeForce RTX 3060]​
2504
GA106 [GeForce RTX 3060 Lite Hash Rate]​
2505
GA106​
2520
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2523
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
252f
GA106 [GeForce RTX 3060 Engineering Sample]​
2560
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2563
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
2583
GA107 [GeForce RTX 3050]​
25a0
GA107M [GeForce RTX 3050 Ti Mobile]​
25a2
GA107M [GeForce RTX 3050 Mobile]​
25a4
GA107​
25a5
GA107M [GeForce RTX 3050 Mobile]​
25af
GA107 [GeForce RTX 3050 Engineering Sample]​
25b5
GA107GLM [RTX A4 Mobile]​
25b8
GA107GLM [RTX A2000 Mobile]​
25e0
GA107BM [GeForce RTX 3050 Ti Mobile]​
25e2
GA107BM [GeForce RTX 3050 Mobile]​
25e5
GA107BM [GeForce RTX 3050 Mobile]​
 
Intel's new GPU drivers imply on the number of DG2 SKUs.

Intel May Be Readying up to 32 Arc Alchemist GPU Models : Read more
If I were to wager, they will only release a few. Probably very 1 perfect yield, and one chopped version to cover defective units of the same die model. The excess models is to confuse competition as to the exact targets.

It's a complex np complete problem based on yields.

You could sell only 100% perfect chips at a high margin. But if only 50% of the chips are perfect that's 50% waste.
You could sell everything that has 1 execution unit more than the next lower chip bin and get 95% yield. But then you lose valuable margin in a more crowded market.
This is where algebra comes in where you draw a graph of lines that represent total yield vs binning x = perfect EU * margin + y = reduced EU * margin where x + y = total usable GPUs./average margin You draw several of these lines to figure out where your maximum margin is based on area under the graph.

Who said algebra wasn't useful?

You could also do a Calculus of the equation as it's a curve based on total usable dies, take the derivative to find the maximum inflection point
 
Last edited:

lazyabum

Prominent
Jan 16, 2021
99
21
535
0
There're 61 Ampere devices, not counting cards with GA100. Quite a few are engineering samples.

2200
GA102​
2204
GA102 [GeForce RTX 3090]​
2205
GA102 [GeForce RTX 3080 Ti 20GB]​
2206
GA102 [GeForce RTX 3080]​
2208
GA102 [GeForce RTX 3080 Ti]​
220d
GA102 [CMP 90HX]​
2216
GA102 [GeForce RTX 3080 Lite Hash Rate]​
222b
GA102 [GeForce RTX 3090 Engineering Sample]​
222f
GA102 [GeForce RTX 3080 11GB / 12GB Engineering Sample]​
2230
GA102GL [RTX A6000]​
2231
GA102GL [RTX A5000]​
2235
GA102GL [A40]​
2236
GA102GL [A10]​
2237
GA102GL [A10G]​
223f
GA102GL​
228b
GA104 High Definition Audio Controller​
2302
GA103​
2321
GA103​
2420
GA103M [GeForce RTX 3080 Ti Mobile]​
2482
GA104 [GeForce RTX 3070 Ti]​
2483
GA104​
2484
GA104 [GeForce RTX 3070]​
2486
GA104 [GeForce RTX 3060 Ti]​
2487
GA104 [GeForce RTX 3060]​
2488
GA104 [GeForce RTX 3070 Lite Hash Rate]​
2489
GA104 [GeForce RTX 3060 Ti Lite Hash Rate]​
248a
GA104 [CMP 70HX]​
249c
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
249d
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
249f
GA104M​
24a0
GA104 [Geforce RTX 3070 Ti Laptop GPU]​
24ac
GA104 [GeForce RTX 30x0 Engineering Sample]​
24ad
GA104 [GeForce RTX 3060 Engineering Sample]​
24af
GA104 [GeForce RTX 3070 Engineering Sample]​
24b0
GA104GL [RTX A4000]​
24b6
GA104GLM [RTX A5000 Mobile]​
24b7
GA104GLM [RTX A4000 Mobile]​
24b8
GA104GLM [RTX A3000 Mobile]​
24bf
GA104 [GeForce RTX 3070 Engineering Sample]​
24dc
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
24dd
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
2501
GA106 [GeForce RTX 3060]​
2503
GA106 [GeForce RTX 3060]​
2504
GA106 [GeForce RTX 3060 Lite Hash Rate]​
2505
GA106​
2520
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2523
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
252f
GA106 [GeForce RTX 3060 Engineering Sample]​
2560
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2563
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
2583
GA107 [GeForce RTX 3050]​
25a0
GA107M [GeForce RTX 3050 Ti Mobile]​
25a2
GA107M [GeForce RTX 3050 Mobile]​
25a4
GA107​
25a5
GA107M [GeForce RTX 3050 Mobile]​
25af
GA107 [GeForce RTX 3050 Engineering Sample]​
25b5
GA107GLM [RTX A4 Mobile]​
25b8
GA107GLM [RTX A2000 Mobile]​
25e0
GA107BM [GeForce RTX 3050 Ti Mobile]​
25e2
GA107BM [GeForce RTX 3050 Mobile]​
25e5
GA107BM [GeForce RTX 3050 Mobile]​
[/QUOTAn
There're 61 Ampere devices, not counting cards with GA100. Quite a few are engineering samples.

2200
GA102​
2204
GA102 [GeForce RTX 3090]​
2205
GA102 [GeForce RTX 3080 Ti 20GB]​
2206
GA102 [GeForce RTX 3080]​
2208
GA102 [GeForce RTX 3080 Ti]​
220d
GA102 [CMP 90HX]​
2216
GA102 [GeForce RTX 3080 Lite Hash Rate]​
222b
GA102 [GeForce RTX 3090 Engineering Sample]​
222f
GA102 [GeForce RTX 3080 11GB / 12GB Engineering Sample]​
2230
GA102GL [RTX A6000]​
2231
GA102GL [RTX A5000]​
2235
GA102GL [A40]​
2236
GA102GL [A10]​
2237
GA102GL [A10G]​
223f
GA102GL​
228b
GA104 High Definition Audio Controller​
2302
GA103​
2321
GA103​
2420
GA103M [GeForce RTX 3080 Ti Mobile]​
2482
GA104 [GeForce RTX 3070 Ti]​
2483
GA104​
2484
GA104 [GeForce RTX 3070]​
2486
GA104 [GeForce RTX 3060 Ti]​
2487
GA104 [GeForce RTX 3060]​
2488
GA104 [GeForce RTX 3070 Lite Hash Rate]​
2489
GA104 [GeForce RTX 3060 Ti Lite Hash Rate]​
248a
GA104 [CMP 70HX]​
249c
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
249d
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
249f
GA104M​
24a0
GA104 [Geforce RTX 3070 Ti Laptop GPU]​
24ac
GA104 [GeForce RTX 30x0 Engineering Sample]​
24ad
GA104 [GeForce RTX 3060 Engineering Sample]​
24af
GA104 [GeForce RTX 3070 Engineering Sample]​
24b0
GA104GL [RTX A4000]​
24b6
GA104GLM [RTX A5000 Mobile]​
24b7
GA104GLM [RTX A4000 Mobile]​
24b8
GA104GLM [RTX A3000 Mobile]​
24bf
GA104 [GeForce RTX 3070 Engineering Sample]​
24dc
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
24dd
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
2501
GA106 [GeForce RTX 3060]​
2503
GA106 [GeForce RTX 3060]​
2504
GA106 [GeForce RTX 3060 Lite Hash Rate]​
2505
GA106​
2520
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2523
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
252f
GA106 [GeForce RTX 3060 Engineering Sample]​
2560
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2563
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
2583
GA107 [GeForce RTX 3050]​
25a0
GA107M [GeForce RTX 3050 Ti Mobile]​
25a2
GA107M [GeForce RTX 3050 Mobile]​
25a4
GA107​
25a5
GA107M [GeForce RTX 3050 Mobile]​
25af
GA107 [GeForce RTX 3050 Engineering Sample]​
25b5
GA107GLM [RTX A4 Mobile]​
25b8
GA107GLM [RTX A2000 Mobile]​
25e0
GA107BM [GeForce RTX 3050 Ti Mobile]​
25e2
GA107BM [GeForce RTX 3050 Mobile]​
25e5
GA107BM [GeForce RTX 3050 Mobile]​
Not to mention that
There're 61 Ampere devices, not counting cards with GA100. Quite a few are engineering samples.

2200
GA102​
2204
GA102 [GeForce RTX 3090]​
2205
GA102 [GeForce RTX 3080 Ti 20GB]​
2206
GA102 [GeForce RTX 3080]​
2208
GA102 [GeForce RTX 3080 Ti]​
220d
GA102 [CMP 90HX]​
2216
GA102 [GeForce RTX 3080 Lite Hash Rate]​
222b
GA102 [GeForce RTX 3090 Engineering Sample]​
222f
GA102 [GeForce RTX 3080 11GB / 12GB Engineering Sample]​
2230
GA102GL [RTX A6000]​
2231
GA102GL [RTX A5000]​
2235
GA102GL [A40]​
2236
GA102GL [A10]​
2237
GA102GL [A10G]​
223f
GA102GL​
228b
GA104 High Definition Audio Controller​
2302
GA103​
2321
GA103​
2420
GA103M [GeForce RTX 3080 Ti Mobile]​
2482
GA104 [GeForce RTX 3070 Ti]​
2483
GA104​
2484
GA104 [GeForce RTX 3070]​
2486
GA104 [GeForce RTX 3060 Ti]​
2487
GA104 [GeForce RTX 3060]​
2488
GA104 [GeForce RTX 3070 Lite Hash Rate]​
2489
GA104 [GeForce RTX 3060 Ti Lite Hash Rate]​
248a
GA104 [CMP 70HX]​
249c
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
249d
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
249f
GA104M​
24a0
GA104 [Geforce RTX 3070 Ti Laptop GPU]​
24ac
GA104 [GeForce RTX 30x0 Engineering Sample]​
24ad
GA104 [GeForce RTX 3060 Engineering Sample]​
24af
GA104 [GeForce RTX 3070 Engineering Sample]​
24b0
GA104GL [RTX A4000]​
24b6
GA104GLM [RTX A5000 Mobile]​
24b7
GA104GLM [RTX A4000 Mobile]​
24b8
GA104GLM [RTX A3000 Mobile]​
24bf
GA104 [GeForce RTX 3070 Engineering Sample]​
24dc
GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]​
24dd
GA104M [GeForce RTX 3070 Mobile / Max-Q]​
2501
GA106 [GeForce RTX 3060]​
2503
GA106 [GeForce RTX 3060]​
2504
GA106 [GeForce RTX 3060 Lite Hash Rate]​
2505
GA106​
2520
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2523
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
252f
GA106 [GeForce RTX 3060 Engineering Sample]​
2560
GA106M [GeForce RTX 3060 Mobile / Max-Q]​
2563
GA106M [GeForce RTX 3050 Ti Mobile / Max-Q]​
2583
GA107 [GeForce RTX 3050]​
25a0
GA107M [GeForce RTX 3050 Ti Mobile]​
25a2
GA107M [GeForce RTX 3050 Mobile]​
25a4
GA107​
25a5
GA107M [GeForce RTX 3050 Mobile]​
25af
GA107 [GeForce RTX 3050 Engineering Sample]​
25b5
GA107GLM [RTX A4 Mobile]​
25b8
GA107GLM [RTX A2000 Mobile]​
25e0
GA107BM [GeForce RTX 3050 Ti Mobile]​
25e2
GA107BM [GeForce RTX 3050 Mobile]​
25e5
GA107BM [GeForce RTX 3050 Mobile]​
Not to mention there's as many SKUs for the GTX 1000, RTX 2000 series each including the GTX 1600. Several hundred models for five years down the drain.
 

kal326

Distinguished
Dec 31, 2007
1,178
39
19,320
1
Between mobile, desktop, and potentially professional series workstation class cards 32 isn't that much in the grand scheme of things. That's like a new Xeon line release number of skus.
 

ASK THE COMMUNITY

TRENDING THREADS