• 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.

Is NVIDIA sabotaging their own Ampere launch?

Greebo said:
thought the 3080 and 3090 boards were identical as well?

even if they aren’t it’s pretty easy to use a 3090 reference board to make a 20gb 3080 super vard
Click to expand...
Reference may be the same but I don't think the FEs are. Double-sided (more layers?) and I suspect physcially bigger to accommodate the bigger fan.
 
If you have money burning a hole in your pocket, just get Ryzen 4000 when it arrives this year, I'm waiting because I think RDNA2 will be competitive, RDNA1 wasn't due to tech issues AMD had at the time, presuming they got over that, there's no reason RDNA2 can't be 50% better than what RDNA1 should have been.

Probably still buy Nvidia as their drivers are much better (AMD 290 gave me a lot of issues in the past with dual monitors which apparently they still haven't fixed) but a 3080Ti at a lower price would be ideal.
 
g67575 said:
AMD's last gen (Navi 10 / RDNA v1) was a 9.7 TFlop GPU and was roughly equivalent to a RTX 2070 (7.4 TFlops). Note that Nvidia's RTX 2000 series seems to get more performance per TF.

A RTX 3090 clocked at 2.1Ghz is capable of 44 TFlops. RTX 3080 is 29.7 TFlops, but could do more with overclocking.

I kind of think the RTX 3090 should just come with 2.1Ghz clock frequency as standard, considering the £1400 price tag...

I hate to p*ss on the excitement bonfire, but AMD would have to build a GPU that is 3 or 4 times more powerful than their 5700XT to really compete with Nvidia, probably with IPC improvements too.

AMD will be competing with just the RTX 3070 this year, that's fine. I think the RDNA 2 desktop GPU will have double the TF count of the PS5 GPU (10.2 TFlops)
Click to expand...

You have made an assumption though. Nvidia's performance per TF has dropped this series. The 3090 is only 20% faster than a 3080 despite having 48% more TFlops and a 3080 with 29.7 TFlops should be more than 4 times faster than a RTX 2070 and 2 times faster than a 2080ti (it isnt in either case)

So dont get too hung up over Tflops yet. I say an article somewhere which said Nvidia had dropped 59% in efficiency this generation.

A lot of the extra performance form Nvidia is just coming from the brute force of using 320W and 350W.
 
DarrenM343 said:
October 2021 maybe :).
I really don't see NV launching the Super versions so soon and it's easy to make an assumption there will be one, rightly or wrongly. I think people often have an agenda when releasing rumours
Click to expand...
Aye it'll be a Karen working at reception at nvidia who isn't getting paid enough :p
 
Greebo - You're correct, more than the doubling the TF count (and Shader Units) from the 2080 TI to the 3080 hasn't resulted in the performance increase you might expect. According to this link, the performance improvement in Doom External is ~50%. Link here:
https://www.theverge.com/2020/9/3/2...e-rtx-3080-4k-gameplay-rtx-2080-ti-comparison

Not bad for £650... Compared to the price of the RTX 2080 TI of around £1200.

Comparing the specs of the 2080 TI and the 3080, you can see the shader count has gone up massively, but the TMU count is the same (272) and the ROP count has 'only' increase from 88 to 96. I'm not sure if the specifications on Techpowerup are 100% accurate (full specs are not yet available), but I think they are probably about right.

If they had somehow doubled the texture and pixel rate, perhaps the 3080 would be 100% faster than the RTX 2080 TI. What we got was a 20.7% increase in pixel rate and a 10.6% increase in texture rate.
 
Last edited:
DarrenM343 said:
October 2021 maybe :).
I really don't see NV launching the Super versions so soon and it's easy to make an assumption there will be one, rightly or wrongly. I think people often have an agenda when releasing rumours
Click to expand...
Nothing says that they have to be "Ti" or "Super" versions. In fact, I'd wager they almost certainly won't be if it's just the same card with doubled VRAM. It's hardly the first time we'll have had identically-named cards with two different amounts of VRAM. It happens all the time. Hell, Nvidia even cut down the core of the 3GB 1060 and still kept the name.
 
g67575 said:
Greebo - You're correct, more than doubling the Shader Units from the 2080 TI to the 3080 hasn't resulted in the performance increase you might expect.

Comparing the specs of the 2080 TI and the 3080, you can see the shader count has gone up massively, but the TMU count is the same (272) and the ROP count has 'only' increase from 88 to 96. I'm not sure if the specifications on Techpowerup are 100% accurate (full specs are not yet available), but I think they are probably about right.

If they had somehow doubled the texture and pixel rate, perhaps the 3080 would be 100% faster than the RTX 2080 TI. What we got was a 20.7% increase in pixel rate and a 10.6% increase in texture rate.
Click to expand...

The CUDA cores haven't actually doubled, it's just now they perform 2 instructions as opposed to 1 on turing. The physical number of CUDA cores is equal between the 2080Ti and the 3080.
 
ljt said:
The CUDA cores haven't actually doubled, it's just now they perform 2 instructions as opposed to 1 on turing. The physical number of CUDA cores is equal between the 2080Ti and the 3080.
Click to expand...

Damn, no wonder Nvidia is so far ahead in performance. I guess the best case scenario would be if AMD has managed to accomplish a similar thing with RDNA 2 Compute Units.

EDIT - They haven't cause the console RDNA 2 shader units aren't double what you would expect based on the CU count.
 
g67575 said:
Damm, no wonder they are so far ahead in performance. I guess the best case scenario would be if AMD has managed to accomplish a similar thing with RDNA 2 Compute Units.
Click to expand...

I believe AMD are going a slightly different route (from rumours anyway) where they are going more compute units, but that can do either RT or "normal" rather than splitting RT off onto dedicated cores ala Nvidia
 
I wonder how many fps will be have Quake 2 rtx on rtx 3090. On my currently 2080 Ti WF Aorus ,without dynamic res.scaling i have 48-50fps+- on 1440P. Despite of that fps i see jittery/stutter when moving fast mouse. Guess engine.
So With Rtx 3090 it will be maybe 80-90fps?: )D

This is with all maxed out.
 
I think AMD have done well supporting the Ampere launch.
With nothing to show until November and that will be probably a paper launch anyway. So it's easy Nvidia pickings all the way up to Christmas.
 
You must log in or register to reply here.
Back
Top Bottom