Are you ready for the next gen GPUs?

[mrgreenthump]

matching ?
they showed 61 fps in Borderlands 3 4K - TPU RTX3080 review shows 70+ in that game
73 fps in Gears 5, RTX 3080 = 84+ fps

From what I see 3080 does a tad bit under 60 dps in BL3 4k Badass settings. So nope.

[kaelleria]

matching ?
they showed 61 fps in Borderlands 3 4K - TPU RTX3080 review shows 70+ in that game
73 fps in Gears 5, RTX 3080 = 84+ fps

[Shakadam]

matching ?
they showed 61 fps in Borderlands 3 4K - TPU RTX3080 review shows 70+ in that game
73 fps in Gears 5, RTX 3080 = 84+ fps

You can't compare benchmarks directly between different test systems and test setups.
TPU for example tested Borderlands 3 in DX11 mode in what I think is ultra quality, whereas the numbers AMD showed was for DX12 and Badass quality mode.

Likewise if we look at Gears 5, TPU has higher numbers for the 3080 than both Guru3d and Sweclockers (just first 2 places I checked) at 84FPS vs 76-77FPS so TPU likely tested at a different quality setting.


What AMD's slide is showing is that's it's in the ballpark of a 3080. You're never gonna get anything more exact than that when trying to compare different tests.

We also don't know that this is the highest end "big Navi". They just said it's an "RX 6000 series". They could have shown a glimpse of their 3080 competitor and have an even higher end model in store for the 3090 competition. It's unlikely but at this point we have no idea.

[Life-Binder]

I take TPU over anyone else

We also don't know that this is the highest end "big Navi". They just said it's an "RX 6000 series". They could have shown a glimpse of their 3080 competitor and have an even higher end model in store for the 3090 competition.

about 0.1% chance of that happening

[kaelleria]

I take TPU over anyone else



about 0.1% chance of that happening

Ok... That doesn't mean their numbers line up with other benches. Running on DX11 vs 12 does not make it a legitimate comparison.

- - - Updated - - -

What we can glean from what they show is they'll have at least a card with similar to the 3080 performance at a much lower power draw.

[Shakadam]

I take TPU over anyone else

That's fine but that also means you can't use TPU's testing and try to extrapolate that versus another test with different test system, test configuration and various other unknown variables. It's not an apples to apples comparison.

[kaelleria]

So...

That wasn't top Navi.... Apparently there's another card with better performance.

https://www.youtube.com/watch?v=SE5eYTIwoT0

We're getting 2 cards on the 28th with wide availability in November and a 6700xt in early 2021.

[Strawberry]

Microsoft and Sony probably poured quite a bit of money into AMD for R&D for their Navi cards. It's the only explanation I can come up with. AMD wasn't even close to Nvidia high end for years. People lost hope they will be able to compete.
But Sony and Microsoft probably didn't want to risk it and those companies are drowning in cash.

[Moozart]

So, just watching right now GN last video and apparently NVidia released a new driver - with the update, owners of the cards that didn't have the CLIC are reporting that crashes disappeared. Still need to investigate what actually the driver does (if it's just a power/frequency curve correction as anyone would do by manually underclock/undervolt the gpu).

The poscap thing was complete BS to begin with. Gigabyte for example is using a ESR 470uF SP-CAP that is actually a higher cost cap than the MLCC's. People kep claiming they were "cheaper caps" . There is a cheaper version of the cap, Zotac used it. The ASUS TUF card that was using 6 MLCCs was also having crashing issues, they all were. My 3080 had some slight issues the day I got it. Once i saw the notice from seisonic to use 2 separate PCIE cables and not the single with 2 6+2 pigtails like everyone usually uses they cleared up. I am willing to be ALOT of people are not running these with 2 separate cables, even the tech tubers if you go back to some of their first 3080 videos were not. I have the Gigabyte Gaming OC. My boost clock hits 2040 at stock settings and I do not have issues.

[Coldkil]

The poscap thing was complete BS to begin with. Gigabyte for example is using a ESR 470uF SP-CAP that is actually a higher cost cap than the MLCC's. People kep claiming they were "cheaper caps" . There is a cheaper version of the cap, Zotac used it. The ASUS TUF card that was using 6 MLCCs was also having crashing issues, they all were. My 3080 had some slight issues the day I got it. Once i saw the notice from seisonic to use 2 separate PCIE cables and not the single with 2 6+2 pigtails like everyone usually uses they cleared up. I am willing to be ALOT of people are not running these with 2 separate cables, even the tech tubers if you go back to some of their first 3080 videos were not. I have the Gigabyte Gaming OC. My boost clock hits 2040 at stock settings and I do not have issues.

Generally, the caps reasoning wasn't a bad one - i mean, you need to have decent stuff on the cards to make them stable, and cheaping out is not a good idea. Objectively the MLCCs are more expensive and do a better job, but you don't need to go ASUS and have 6 of them. Plus, as you said, there are caps even more expensive than MLCCs that do a good job.

The problem isn't even on cables - the cards had a way too stretched frequency table. Partners didn't have hands on proper drivers until like a week before release, that's why FE were more stable (though i've heard of crashes aswell). The point is that different systems have different issues; i am sure there are a plethora of people with underspec PSUs or old ones, or have other issues with the systems.

It's a whole world of things that went the wrong way, first of all not enough testing before release. I'm sure that partners would have changed their firmware with slightly lower boost clocks/improved curves if they found the issue in time.

[mrgreenthump]

Microsoft and Sony probably poured quite a bit of money into AMD for R&D for their Navi cards. It's the only explanation I can come up with. AMD wasn't even close to Nvidia high end for years. People lost hope they will be able to compete.
But Sony and Microsoft probably didn't want to risk it and those companies are drowning in cash.

Afaik, neither Sony or Microsoft has the Infinity Cache that RNDA2 apparently has. It was the Zen 2 team that jumped and helped them with scheduling and caching. But yes the power savings, clock speed and efficiency is probably better due to Sony and Microsoft helping them.

[Thunderball]

Afaik at 4k the 3950x and 10900k were even. So wouldn't think CPU would start mattering now as it's so heavily GPU bottlenecked still.

You know that you cant compare CPUs if there's any form of GPU bottleneck involved right? Same with the other way around - if there's any kind of CPU bottleneck any GPU comparison is irrelevant.

- - - Updated - - -

So...

That wasn't top Navi.... Apparently there's another card with better performance.

https://www.youtube.com/watch?v=SE5eYTIwoT0

We're getting 2 cards on the 28th with wide availability in November and a 6700xt in early 2021.

Yes, there's 2 versions of the 80 CU GPU. That was known for a while. Physically they're the same GPU but the clocks are going to be different. Probably VRAM size aswell. Also why are there so many blantant AMD fanboys on YT these days? Pretty disgusting to watch tbh.

[kaelleria]

You know that you cant compare CPUs if there's any form of GPU bottleneck involved right? Same with the other way around - if there's any kind of CPU bottleneck any GPU comparison is irrelevant.

- - - Updated - - -



Yes, there's 2 versions of the 80 CU GPU. That was known for a while. Physically they're the same GPU but the clocks are going to be different. Probably VRAM size aswell. Also why are there so many blantant AMD fanboys on YT these days? Pretty disgusting to watch tbh.

I mean... watch the video. That guy is the source for most of the Nvidia and RDNA2 leaks from earlier this year.

[mrgreenthump]

You know that you cant compare CPUs if there's any form of GPU bottleneck involved right? Same with the other way around - if there's any kind of CPU bottleneck any GPU comparison is irrelevant.

And how is that different to what I said? At 4k the 3080 is very much the bottleneck. We weren't talking about CPUs.

[Thunderball]

I mean... watch the video. That guy is the source for most of the Nvidia and RDNA2 leaks from earlier this year.

The only guy on YT I've seen that had any kind of truthful info on RDNA2 was the Coreteks guy. Everyone else were just making videos off leakers that tweet, but not make videos.

[kaelleria]

The only guy on YT I've seen that had any kind of truthful info on RDNA2 was the Coreteks guy. Everyone else were just making videos off leakers that tweet, but not make videos.

Coreteks info comes from AIBs... It's not accurate. RDNA info has not been shared with AIBs.

Coreteks claimed Big Navi would only be 15% better than a 2080 TI...

[Thunderball]

And how is that different to what I said? At 4k the 3080 is very much the bottleneck. We weren't talking about CPUs.

You just cant compare the data if the systems have nothing in common, i.e. 3950X with Big Navi and 10900K with 3080. There's no such thing as CPU parity at 4K as you cannot test CPUs at 4K.

[mrgreenthump]

Coreteks claimed Big Navi would only be 15% better than a 2080 TI...

Probably was at some point. At least one of the skus.

[kaelleria]

Probably was at some point. At least one of the skus.

Right... his claim was that was was the top end performance, which isn't the case if their benchmarks are remotely accurate.

[Donald Hellscream]

Nah, I am at a point where I need to either throw money after a new Mobo and by extension a new CPU as my current one doesn't fit anything newer than a 1150 socket, which I can't afford, or buy a new pc, which I also can't afford.

So I am sticking with my rx 570, which is more than suitable for my current needs anyhow.