• 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.

*** The AMD RDNA 4 Rumour Mill ***

What’s the thoughts on us seeing a 9070XTX?

Do AMD have something cooking away? I know they’ve said their concentrating on midrange this time round but there’s been a few rumours circulating
 
233 said:
What’s the thoughts on us seeing a 9070XTX?

Do AMD have something cooking away? I know they’ve said their concentrating on midrange this time round but there’s been a few rumours circulating
Click to expand...

I think they do.... i think they have a 32GB version in the oven for those bargain workstation / AI accelerator people "oh its got 1550 TOPs" why do i as a gamer care????? you didn't bother to print the TOPs on the box of the 7000 series....
 
Last edited:
Tinek said:
Doesn't that only apply to vRAM which actually has ecc enabled? I know some previous generation series had ecc, but then GPU makers started to reserve it again just for pro cards, so it's possible none of the new GPUs have that on and then stability (and visible errors) could be the sign again. I know my 4090 by default has it turned off as it's slowing down vRAM somewhat when it's on and has to be turned on in drivers to work.
Click to expand...

You are correct. It’s not ECC that is used in gaming GPUs, it’s EDR (Error Detection and Replay). Basically if the GPU detects an error, it resends the data… hence the noticeable drop in performance. I can see exactly this behaviour on my 7900 XT and what we are seeing in that video is the exact behaviour we would expect with VRAM being clocked too high.
 
Last edited:
ICDP said:
You are correct. It’s not ECC that is used in gaming GPUs, it’s EDR (Error Detection and Replay). Basically if the GPU detects an error, it resends the data… hence the noticeable drop in performance. I can see exactly this behaviour on my 7900 XT.
Click to expand...

Mine just crashes if i push it over 2600.
 
DerBauer did a video overclocking a 9070XT Red Devil. His conclusion was undervolt, push the power limit and leave the mem clocks alone.
 
Last edited:
ICDP said:
Because it’s not full on ECC.
Click to expand...

Nether is yours and you're not "seeing exactly this behaviour on your 7900 XT" just as i aren't. Because it’s not full on ECC.


ICDP said:
You are correct. It’s not ECC that is used in gaming GPUs, it’s EDR (Error Detection and Replay). Basically if the GPU detects an error, it resends the data… hence the noticeable drop in performance. I can see exactly this behaviour on my 7900 XT and what we are seeing in that video is the exact behaviour we would expect with VRAM being clocked too high.
Click to expand...

ICDP said:
This guy is normally very good but his take on VRAM overclocking is an epic fail. The VRAM has error correction to prevent crashes. So high clocks cause instability and the GPU compensates by reducing the actual performance.

The way to OC VRAM is to keep going until your performance actually regresses, then reduce the OC slightly for extra stability. Keep your VRAM out of the error correction state.
Click to expand...
 
Last edited:
ICDP said:
You are correct. It’s not ECC that is used in gaming GPUs, it’s EDR (Error Detection and Replay). Basically if the GPU detects an error, it resends the data… hence the noticeable drop in performance. I can see exactly this behaviour on my 7900 XT and what we are seeing in that video is the exact behaviour we would expect with VRAM being clocked too high.
Click to expand...
EDR makes sense, that's usually not even shown in specs, though we had for a while ECC too. At least they have EDR now, in the past there was nothing, just crashes.
 
humbug said:
Nether is yours and you're not "seeing exactly this behaviour on your 7900 XT" just as i aren't. Because it’s not full on ECC.
Click to expand...

I’m not getting into a pointless debate on semantics with you. My post was a quick outline to why overclocking VRAM beyond limits results in the behaviour seen in the video. And indeed with my experience. The fact yours crashes does not remotely make that point wrong.
 
233 said:
What’s the thoughts on us seeing a 9070XTX?

Do AMD have something cooking away? I know they’ve said their concentrating on midrange this time round but there’s been a few rumours circulating
Click to expand...

The 9070XT is using the full Navi 48 and AMD aren't doing a bigger GPU until RDNA5/UDNA.

We could see a 9070XT+ with higher power limits, Clocks and more memory but this so far is just rumour.
 
Tythus said:
It's a midpoint—it can correct errors, but it's highly inefficient at doing so. And once you go beyond a certain threshold, ECC or not, it won't be of any help. :P
Click to expand...

Yep, this is exactly what it does. But ultimately my point was to explain the “why” without getting too deep in to the technicalities. I obviously failed because some people want to argue semantics :D
 
ICDP said:
Yep, this is exactly what it does. But ultimately my point was to explain the “why” without getting too deep in to the technicalities. I obviously failed because some people want to argue semantics :D
Click to expand...
Truly shocking revelations here, folks! Some people… arguing about semantics?! We never saw this coming. Experts say it was only a matter of time before a discussion turned into a meticulous dissection of word choice. More at 10—see you in the studio.

In other news, haters gonna hate!
 
You must log in or register to reply here.
Back
Top Bottom