• 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.

Blackwell gpus

spudbynight said:
The last I heard the B770 was unlikely to happen
Click to expand...
Off topic for this thread, but there's still hope

www.guru3d.com

Intel Working on Higher-end Graphics card, BMG-G31 GPU Export Data Appears

Intel is progressing with the development of its Battlemage-generation graphics card, known internally as the BMG-G31.On December 3, the company announced two new models in the Battlemage series: the Intel Arc B580 with 12GB of memory and the Intel Arc B570 with 10GB of memory, both built on the...
www.guru3d.com www.guru3d.com
 
flat,800x800,075,f.jpg
 
Armageus said:
Off topic for this thread, but there's still hope

www.guru3d.com

Intel Working on Higher-end Graphics card, BMG-G31 GPU Export Data Appears

Intel is progressing with the development of its Battlemage-generation graphics card, known internally as the BMG-G31.On December 3, the company announced two new models in the Battlemage series: the Intel Arc B580 with 12GB of memory and the Intel Arc B570 with 10GB of memory, both built on the...
www.guru3d.com www.guru3d.com
Click to expand...

The problem the b770 will be scaling, A770 had nearly double the amount of cores compared to A580 but it was only 20% faster, so it cost much more to build but gave little benefit.

Whether b770 launches depends if Intel has fixed the scaling problem - AMD and Nvidia also have scaling problems of their own, Nvidia for example if you double the number of cores you only get 50% more performance but Intel's scaling is much worse, you only get 20-25% when you double cores.

And given Intel's financials I can't see them willing to sell at a loss. A770 was sold at a loss, B580 is not and that's much better considering Intel are bleeding money so I don't see any way with their current position that they can afford to sell B770 at a loss, and therefore it needs to be priced high enough to make a profit and that's only possible if they fix the scaling issue with the architecture
 
Last edited:
Perfect_Chaos said:
But nope, got to milk for everything..
Click to expand...
It's NVIDIA, who nicely grown into a monopoly - expect things to get worse, not better. At least not till Intel and AMD start taking bigger chunks of low-mid market away from them. Problem isn't just pricing and performance but also mindset, though - neither Intel nor AMD have a big GPU following currently, people just choose NVIDIA even if it's demonstrably worse in a given price range, just because they know it and trust it.
 
Ascendancy said:
Zotac have seemingly let slip their launch lineup:
videocardz.com

ZOTAC confirms GeForce RTX 5090 with 32GB GDDR7 memory, 5080 and 5070 series listed as well - VideoCardz.com

ZOTAC confirms NVIDIA RTX 50 launch lineup Two years ago, just before NVIDIA announced the GeForce RTX 40 series, a ZOTAC design based on the Ada Lovelace architecture was leaked ahead of the official showcase. Last month, there was a leak from ZOTAC’s GPU factory claiming that the company was...
videocardz.com videocardz.com
Click to expand...
So, no 5080D, which strongly suggest 5080 < 4090 perf-wise.
 
Tinek said:
So, no 5080D, which strongly suggest 5080 < 4090 perf-wise.
Click to expand...

To be fair it means the AI and Compute performance of the 5080 is lower than the 4090, but it doesn't have to mean the gaming performance is lower
 
Last edited:
Grim5 said:
To be fair it means the AI and Compute performance of the 5080 is lower than the 4090, but it doesn't have to mean the gaming performance is lower
Click to expand...
US export restrictions are based on "total processing performance". Which means mostly CUDA cores, which are directly responsible also for performance in modern games (excluding RT/PT games for the time being). They might speed up tensor and RT cores, but I wouldn't count on some magical speed-ups in gaming generally that's not related to processing performance increases. And that would be export-limited if it crosses 4090D's speed (not even 4090). To me it's a clear indicator 5080<4090D in general performance. Edge cases can be just that - edge cases, ergo largely meaningless. There is a reason they stopped production of 4090 a while ago, to sell them all out before these "great" news, I reckon. ;)
 
Last edited:
Ascendancy said:
Pretty wild if that's the case. Has that ever happened before? Where the 80 class has been worse than the previous flagship?
Click to expand...
I like to remind people that NVIDIA CEO repeats all the time for a while now that GPUs don't get much faster anymore, AI is the future instead. We're almost by the very wall of what's possible in physics without running into too many wild quantum effects. Maybe different materials will let us increase clocks, heat transfer, power etc. but transistors' size hasn't changed for years now - only their construction did, but the actual size of each transistor is pretty much the same since "10nm" processes (and that already didn't mean 10nm per transistor either!). The real size planned in few generations of production (so not even current top ones!) will be about 13nm in size per transistor, last I've seen online recently. To get down to 3nm is just not possible currently, as that would be right into quantum realm of weirdness - it's all just marketing names that means "we've improved things a bit here and there." without touching the transistor sizes.

Anyway, all the low-hanging fruits of performance have been already reached, now it's getting harder and harder each generation to get any speed-ups. Same with CPUs (vide 9000 Ryzen vs 7000 - faster clocks, bit less power used, but for gaming it's near the same CPU).
 
Last edited:
5070ti is the only one I see having a chance of catching my interest, as 16GB is absolute minimum and the gap from the rumoured 5080 spec in CUDA cores isn't huge... all depends on pricing of course, there's a good chance both will land at more than I'm willing to consider.
I'd like more than I'm getting from my 7900XT, but only up to about the same I paid for that (£700) - somehow I don't really fancy my chances, but would be glad of a pleasant surprise :)
 
Tinek said:
I like to remind people that NVIDIA CEO repeats all the time for a while now that GPUs don't get much faster anymore, AI is the future instead. We're almost by the very wall of what's possible in physics without running into too many wild quantum effects. Maybe different materials will let us increase clocks, heat transfer, power etc. but transistors' size hasn't changed for years now - only their construction did, but the actual size of each transistor is pretty much the same since "10nm" processes (and that already didn't mean 10nm per transistor either!). The real size planned in few generations of production (so not even current top ones!) will be about 13nm in size per transistor, last I've seen online recently. To get down to 3nm is just not possible currently, as that would be right into quantum realm of weirdness - it's all just marketing names that means "we've improved things a bit here and there." without touching the transistor sizes.

Anyway, all the low-hanging fruits of performance have been already reached, now it's getting harder and harder each generation to get any speed-ups. Same with CPUs (vide 9000 Ryzen vs 7000 - faster clocks, bit less power used, but for gaming it's near the same CPU).
Click to expand...

What? You realize we actually USE quantum tunnelling in component design, right? Getting smaller doesn't preclude performance improvement because of quantum effects in the slightest, it just requires accounting for it and in some cases utilizing them. 'We can't use the same design on a smaller process' =/= 'We can't design on a smaller process'
 
My prediction on 50 series prices

Worst case they are closeer to the top end price, but they need to be much cheaper ideally

5090/D - £1800-2800
5080 - £1300-1500
5070 Ti - £950-1050
5070 - £700-850
5060 Ti - upto £550
5060 - £400 minimum
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom