-
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.
Reality check, are AMD just as bad as Nvidia.
- Thread starter humbug
- Start date
Soldato
- Joined
- 14 Aug 2009
- Posts
- 2,869
I'm waiting for a 5080 at $1199 for 50% more performance. That will be a good card since 7900xtx at the same $999 is a good card vs 6900xt.
Man of Honour
I wouldn't hold your breath if current rumours are true and the 50 series aren't going to be the performance gains over 40 series that people are expecting, at least not the 4080/4090 anyway. Plus, if it does end up being better by a lot, will the price increase over a 40 series align with the perf gains? It's not a trend we have seen outside of a halo card like the 4090, to which there still is no equal from any other brand, so high prices for high end cards are likely to remain, well, high 
AD102 is already 608mm² on TSMC 4nm (TSMC 5nm and 4nm are almost identical).
en.wikipedia.org
Not that gate pitch is everything.
Later on that Wiki article they mention density. That plus:
en.wikipedia.org
Gets us something like this
I guess +50% could be okay. I also expect they won't go as crazy as they did with AD102 (since nodes are slowing down, going for near the rectile limit right at the beginning is a bit "crazy").
Next gen or the one after, Nvidia too will have to go chiplet.
Now, 600mm² halo cards tell us little what a 5080 and lower could bring.
AD103 is only 380mm², so even if sticking to N4 they could improve it a bit by moving that tier to a 450-500mm² die. All at the cost of margins of course, so they won't be keen to do that.
3 nm process - Wikipedia
| Process | Gate pitch | Metal pitch | Year |
|---|---|---|---|
| 5 nm | 51 nm | 30 nm | 2020 |
| 3 nm | 48 nm | 24 nm | 2022 |
| 2 nm | 45 nm | 20 nm | 2024? |
Not that gate pitch is everything.
Later on that Wiki article they mention density. That plus:
5 nm process - Wikipedia
| TSMC | Mtr/mm² | Mtr for 600mm² die | compared |
| N5 / N5P | 138 | 82,800 | 97% |
| N4 | 143 | 85,800 | 100% |
| N3 | 220 | 132,000 | 154% |
Next gen or the one after, Nvidia too will have to go chiplet.
Now, 600mm² halo cards tell us little what a 5080 and lower could bring.
AD103 is only 380mm², so even if sticking to N4 they could improve it a bit by moving that tier to a 450-500mm² die. All at the cost of margins of course, so they won't be keen to do that.
Tsmc 3nm is yields are still quite bad so not sure when nvidia and amd can make the switch. I guess amd will focus on cost and use 4nm and 5nm instead. Nvidia will want 3nm but yeah it's not great at this stage, who knows what it looks like next year
As for those 3nm yields, tsmc is only getting a yield of 50% while manufacturing Apple's upcoming tiny A17 SoC. The yields are bad enough that Apple has now told Tsmc it no longer wants to pay per wafer and wants to only pay for working chips otherwise the cost is ridiculous. The cost per wafer is about $20k USD
As for those 3nm yields, tsmc is only getting a yield of 50% while manufacturing Apple's upcoming tiny A17 SoC. The yields are bad enough that Apple has now told Tsmc it no longer wants to pay per wafer and wants to only pay for working chips otherwise the cost is ridiculous. The cost per wafer is about $20k USD
Last edited:
Caporegime
I wouldn't hold your breath if current rumours are true and the 50 series aren't going to be the performance gains over 40 series that people are expecting, at least not the 4080/4090 anyway. Plus, if it does end up being better by a lot, will the price increase over a 40 series align with the perf gains? It's not a trend we have seen outside of a halo card like the 4090, to which there still is no equal from any other brand, so high prices for high end cards are likely to remain, well, high
Where are you hearing this from? WTFtech?
Man of Honour
Nvidia RTX 50 GB102 Blackwell early internal performance gains only in range of Pascal to Turing and not 2.6x from AD102, increased SM counts on the anvil
Nvidia's next gen RTX 50 Blackwell with purported 144 SMs isn't reaching ambitious performance gains in internal testing, according to latest rumors. Apparently, Blackwell's performance gains vis-à-vis AD102 are in the range of what we've seen from Pascal to Turing, but Nvidia wants more. This...
www.notebookcheck.net
Caporegime
Nvidia RTX 50 GB102 Blackwell early internal performance gains only in range of Pascal to Turing and not 2.6x from AD102, increased SM counts on the anvil
Nvidia's next gen RTX 50 Blackwell with purported 144 SMs isn't reaching ambitious performance gains in internal testing, according to latest rumors. Apparently, Blackwell's performance gains vis-à-vis AD102 are in the range of what we've seen from Pascal to Turing, but Nvidia wants more. This...
And their source?
Hi, my name is...
Soldato
- Joined
- 14 Aug 2009
- Posts
- 2,869
I was being ironicI wouldn't hold your breath if current rumours are true and the 50 series aren't going to be the performance gains over 40 series that people are expecting, at least not the 4080/4090 anyway. Plus, if it does end up being better by a lot, will the price increase over a 40 series align with the perf gains? It's not a trend we have seen outside of a halo card like the 4090, to which there still is no equal from any other brand, so high prices for high end cards are likely to remain, well, high
To see a bad pricing card being made great just because it looks decent compared to another bad priced card is quite funny
Last edited:
Soldato
Same here.
NVIDIA has crossed a line where I really begrudge giving them any more money. I'm disgusted with Mini-Me-AMD. I am not sure if I can hang on until intel get their act together, but I really hope so, because I sure am going to resent having to pay towards any more leather jackets/kitchen utensils.
Last edited:
Use dlss. I do and its fantastic.
Caporegime
Nvidia crossed that line with me back in the GTX 970 days. But AMD has not made things easy and even they have crossed it too of late with their silly pricing. So what does one do? Many vocal lot seem to just slate Nvidia, recommend AMD and then go buy Nvidia anyways it seems.
Why do you need an upgrade? Why must it have AV1 and why must it have 16gb?
Please provide a good answer
I need something with more muscle.
I'm VRam constrained in some games and that's only going to get worse.
I record a lot of videos, i may edit some soon, i want high image quality without 50GB files.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,070
4090, best value.
Soldato
Fed up with both of them. If Starfield runs OK on my current card,I might just try and hold on with it. We finally are getting some information about the RX7700XT:
AMD Navi 32-based Radeon RX 7800/7700 series reportedly targeting September launch - VideoCardz.com
AMD might already showcase Radeon RX 7800/7700 at Gamescom AMD is reportedly targeting a September release for its Navi 32 desktop GPUs. According to the latest video by Moore’s Law is Dead, the GPU manufacturer has completed the development of its Navi 32 GPU and has already released...
videocardz.com
I really hope it is 54 CUs not 48 CUs,because at least that will be 30% faster than an RX6700XT/RTX3060TI. But with the RTX4060TI being so rubbish,I expect it might be 48 CU!
Last edited:
Associate
Or it will be 54CU's and priced too high.
Soldato
Knowing AMD they will price it a £499.99 and be suprised when the RTX4070 outsells it.
Associate
- Joined
- 22 Nov 2020
- Posts
- 1,459
Last edited:
Fed up with both of them. If Starfield runs OK on my current card,I might just try and hold on with it. We finally are getting some information about the RX7700XT:
AMD Navi 32-based Radeon RX 7800/7700 series reportedly targeting September launch - VideoCardz.com
AMD might already showcase Radeon RX 7800/7700 at Gamescom AMD is reportedly targeting a September release for its Navi 32 desktop GPUs. According to the latest video by Moore’s Law is Dead, the GPU manufacturer has completed the development of its Navi 32 GPU and has already released...
I really hope it is 54 CUs not 48 CUs,because at least that will be 30% faster than an RX6700XT/RTX3060TI. But with the RTX4060TI being so rubbish,I expect it might be 48 CU!
The 48 CU one would put it 30% faster than the 6700XT, that would make it about 90% the performance of the 6800XT or RTX 4070, or a few % faster than the RX 6800.
The 7900XTX was priced in line with the 6900XT, $999
The 7900XT was way over priced at $899 but has now been officially reduced to $749
The 7600 was priced 22% lower than the 6600 while being 27% faster at $269
Whatever comes next must be priced lower than $749.
7800XT: $649?
7800: $579?
7700XT $479?
Those 3 prices are the same as the previous gen they are replacing.
IMO from that slide the 60 CU one is the RX 7800.
The 70 CU one the RX 7800XT.
UK price if current exchange rates hold:
7800XT: £595
7800: £462
7700XT: £439
Last edited: