• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Intel Arc series unveiled with the Alchemist dGPU to arrive in Q1 2022

Until I see the frametime analysis of a whole benchmark suite, I remain to be convinced.

Manufacturer slides are mostly worthless especially with the issues already uncovered for their other cards.
 
isn't the 750 basically the best gpu they have? only competing with a 3060 doesnt sound that great.
Hm, state of flux. TPU and W1zzard can usually be relied on to update their GPU database, but it still list only A380, A770, and A780.
HIBwMNe.png
No A750 there. It might be another cut-down DG2-515. Maybe similar to the A730M and it's a 3/4 of the full die?
Or they've come to their sense and have renamed them all lower. If top end DG2-512 isn't quick enough, call it A760 instead?
 
isn't the 750 basically the best gpu they have? only competing with a 3060 doesnt sound that great.
It doesn't sound great until you see what the majority are gaming on. Priced right it would be a nice upgrade for those on old slow cards. This place is unusual, most have a nice system, but even then only a minority buy the really high end.
 
A780 might come out October to Jan 2023, it was suggested that it would be out in June this year. I suppose it gives Intel more time to optimise the drivers. As mentioned already its not a bad start from Intel.
 
Beats the 3060, price is key.

PGP7SCX.jpg


I thought it looked stuttery and thought I could see tearing in that video. :confused:
 
Seems PCGH somehow got a table from somewhere:
ZnrybeK.png
So ACM-G10 is the same as TPU called DG2-512. So that implies the the A750 limited edition is 3/4 of the full die, while A580 is cut straight in two.

Since TSMC 6nm wafers aren't cheap but should have similar yields to 7nm, I can't figure out the half-die cut. Especially since TPU also think it's around 400mm².
GA106 is 276mm² but of course Samsung 8nm. 12 billion transistors, whereas TPU has DG2-512 as 21 billion. 3/4 of which (I know some thing don't scale like that), is around 16 billion. So that implies they are matching RTX3060 performance (in the limited amount of modern games they are willing to pick for marketing benchmarks) with only 33% more transistors.

Do we have any idea about TPU? Guess RTX3060 isn't that great at around 170W (6600 XT is about 160W).

EDIT: seems the A750 board has 6-pin and 8-pin PCIe (good for 300W?), but it might be over-engineered as Shroud likes to shout.
 
Right at 47 Minutes.

Shrout pushing using the iGPU for streaming with the ARC GPU, as if you can't do that with Nvidia and AMD GPU's, and Linus, the Nvenc encoder makes almost no difference to the frame rates no matter how much load the GPU is under, stop trying to crawl up Shrout's rear end.
 
Performance results shown here are from a small subset of the games, that work very well with Intel® Arc™ and the Alchemist architecture. I’m not asserting that ALL GAMES will show these results, but it’s a view of what Intel Arc A-series cards are capable of with the right software and engineering enablement.

Given how badly the 380 fell off if the game was the wrong type I think any graph from intel showing performance will involve cherry picking at astronomical highs.

You're not gonna show your top card card beating a 3060 in F1 then losing to a 580 in GTAV are you. GTAV won't exist to you.

Wait for unrestricted 3rd party reviews :eek:
 
They were quite open about concentrating on DX12. And really, I think that those who were dreaming of 3080 / 3090 / 69xx performance were hopelessly optimistic. This is Intel's first try in many, many years.

But when AMD faced such critique on titles where nvidia done better on DX11 there was no empathy, so it is only right they come under the same scrutiny! ;)
 
Back
Top Bottom