Then compare it to RTX 2070.
TU-106: (Full die)
36 SMs, 64 ROPs, 144TMU, 2304 Shading Units, 13.8B Transistors @ 220watts @ 445mm^2 die size.
AMD Radeon RDNA2 Navi (RX 6500, 6600, 6700, 6800, 6900 XT)
Then compare it to RTX 2070.
TU-106: (Full die)
36 SMs, 64 ROPs, 144TMU, 2304 Shading Units, 13.8B Transistors @ 220watts @ 445mm^2 die size.
That can be true in general, but not particulary when talking about Ampere. GDDR6X requires more power and offers higher bandwidth. But if your product isn't able utilize all the bandwidth, it just consumes more power and becomes less power-efficient. Look at RTX 3070 with GDDR6 and RTX 3080 with GDDR6X. The later has 70 %(!!!) higher theoretical bandwidth than the former, but offers only 26-32 % (1440p, 4k, ComputerBase) higher performance. RTX 3080 has the same number of ROPs as RTX 3070, slightly lower boost clock (so fillrate is probably lower for RTX 3080) and the bandwidth cannot be efficiently utilized. But the GDDR6Xs run at full speed and consumes more power.
Wait a sec... TU106 contains 10.8B transistors
Yeah, both Navi10 & TU-106 have 10B+ transistors.
That was the entire point of my posting TU-106 numbers, but somehow I used 2070 SUPER's transistor number. You are correct.
thnx.
Who knows. Same reason they didn't showcase their fastest card on the 8th?
Subtlesnake
Regular
Being at best a bit slower than the 3080 isn't a great marketing point. And based off the other two ray tracing benchmarks we have, I wonder if performance falls off a lot more in heavy ray tracing applications like Control.
Hey if 6800 is just 'a bit slower' in RT than 3080 I'll take it. I think it's more likely the settings used were not the same.
Yah, 6800XT has 20% more RT cores. They both have the same bandwidth to VRAM though. Cache will be a bit faster on the XT because of the clock increase. Will be interesting to see if RT performance scales to the higher end SKUs like 6800XT and 6900XT, or if RT is just too cache unfriendly and you don't get linear scaling with RT cores. Really depends where the bottleneck is.
Subtlesnake
Regular
On another forum I saw someone score 56 FPS with a 3080 on high, so that would be 22% faster. On the other hand, Paul from RedGamingTech is reporting he gets 51, so I guess we will have to see.
Again, the guy who leaked this scores is definitely using custom settings not the default presets. You can use your phone and google translate. I think he disabled or reduced some of the settings.
EDIT: yeah so apparently at the same custom settings 3080 scores 20% higher.
Last edited:
20% higher than the RX 6800 score or 20% higher than some listed 3080 score?
D
Deleted member 2197
Guest
I guess we'll have to wait for the reviews. I was going to compare AMD's Battlefield V 4K score against a recent Guru3D's review but noticed AMD chose DX11 for their testing. Guru3D in all their reviews uses DX12.
TechPowerUp uses DX11 on BFV, here's 3080 for example: https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-founders-edition/9.html
that's super interesting! It would be perfect to know the performance of the RTX 3070 at that game with RT enabled. Also wonder whether the game supports DLSS 2.0 or not. Things are starting to look promising and I am not entirely sure my next GPU is going to be from nVidia. I still have some months to think about it, and decide in the meanwhile.
Frenetic Pony
Veteran
DLSS 2.0 doesn't really matter here. What does is the apparently solid performance of RT, at least in this case. Which is to say here better than a 3070 for a card that costs more, which is fine. With AMD's own "Super Resolution" coming up probably next year and next gen TAA that delivers potentially quite big improvements (UE4's new one already equals DLSS 2.0 more or less) DLSS 2.0 isn't even going to be a totally relevant comparison point in the near future. More important for arch considerations, and anyone looking to buy, it seems RT on RDNA2 can be fine.
With different acceleration techniques and different cache structures I'd imagine performance on it probably differs a solid bit title to title, but showing up runnable performance for the price is a good indicator.
True, it doesn't seem like the extra bandwidth of the 3080 is being utilized and you have to wonder why they didn't just stick to G6. And even the 3070Ti continues to use G6X which seems unnecessary. Framebuffer size aside, the 3070 is decent value as a 1440p card considering that the 3080 is only 26% faster for 40% more money.
It will also be interesting to see how N22 and N23 fare with their respective configurations. N23 on paper looks like it's bandwidth deficient.
I honestly have no idea at this point so I'll just keep my mouth shut and wait for real data.
