AMD RX 7900XTX and RX 7900XT Reviews

[Digidi]

Is there any news on the frontend discussion?

 

[eastmen]

Yea thats one thing. Hard to tell how much and what a gpu has run.

yup which is why when buying used I just recommend putting aside a budget for new thermal pads and paste. hopefully buy one that people have taken apart and redone before so you know the pad thickeness and everything.

 

[Clukos]

That seems pretty serious. Basically avoid the reference model and you should be fine but if he's right then AMD can only do a recall in this case.

 

[Jawed]

Is there any news on the frontend discussion?

I'm not sure whether Gravity Mark performance falls into this topic:

Radeon Gaming/Graphics Performance: Windows 11 vs. Linux GPU Benchmarks - Phoronix

www.phoronix.com

but there's some strange deltas.

 

[Man from Atlantis]

RDNA3 and too high hotspot temperatures on some AMD Radeon RX 7900 XT(X) - Cause research | igor'sLAB

Of course, you can discuss BIOS problems with the changed telemetry of the new RDNA3 cards, faulty thermal sensors in an alleged amok mode or even tangible mechanical problems, but in the end only AMD…

www.igorslab.de

 

[Digidi]

I'm not sure whether Gravity Mark performance falls into this topic:

Radeon Gaming/Graphics Performance: Windows 11 vs. Linux GPU Benchmarks - Phoronix

www.phoronix.com

but there's some strange deltas.

Thank you the questions is which renderer could they use? In this benchmark you can use Compute Shaderr oder Mesh Shader oder Default (old Pipline). In 3dCenter Froum we tested MS and it was not running on 7900xtx

7900xtx with new driver:

3DCenter Forum - Einzelnen Beitrag anzeigen - AMD/ATI - Radeon RX 7900 XT & XTX Review-Thread

Diskussions- und Hardwarehilfe-Forum von 3DCenter.org

www.forum-3dcenter.org

4090:

3DCenter Forum - Einzelnen Beitrag anzeigen - nVidia - GeForce RTX 4090 Review-Thread

Diskussions- und Hardwarehilfe-Forum von 3DCenter.org

www.forum-3dcenter.org

 

[Jawed]

Thank you the questions is which renderer could they use? In this benchmark you can use Compute Shaderr oder Mesh Shader oder Default (old Pipline). In 3dCenter Froum we tested MS and it was not running on 7900xtx

I'm afraid I don't know. I think Phoronix has an article somewhere that states how it configures each test it runs.

Part of what's interesting here is that the open source driver is sometimes much faster than the AMD driver (on Windows 11). It's always faster with Vulkan on RDNA 3, despite often being slower on RDNA 2.

Curious stuff... It's as if the AMD code is as "experimental" as the open source code

 

[Scott_Arm]

That seems pretty serious. Basically avoid the reference model and you should be fine but if he's right then AMD can only do a recall in this case.

Yikes. That is a huge screw up. Good thing they barely built any of them if the stock reports are correct.

 

[digitalwanderer]

That seems pretty serious. Basically avoid the reference model and you should be fine but if he's right then AMD can only do a recall in this case.

I missed this earlier figuring it was FUD, but it's Der Bauer who I hugely respect so I gotta say damn this is seriously not good for AMD.

 

[Wesker]

That seems pretty serious. Basically avoid the reference model and you should be fine but if he's right then AMD can only do a recall in this case.

Geez, talk about dropping the ball at the worst possible time.

 

[homerdog]

That seems pretty serious. Basically avoid the reference model and you should be fine but if he's right then AMD can only do a recall in this case.

So the vapor chamber goes full vapor and the gas never liquifies again when it reaches the cold end of the loop? Never go full vapor.

 

[trinibwoy]

Geez, talk about dropping the ball at the worst possible time.

It might actually be the best possible time to screw up as the 7900XTX wasn’t going to move major volume anyway. Save the perfect execution for when you can make some money from it.

 

[Frenetic Pony]

It might actually be the best possible time to screw up as the 7900XTX wasn’t going to move major volume anyway. Save the perfect execution for when you can make some money from it.

I don't know if it's sampling bias on my part or an actual effect, but rushing products out to meet arbitrary deadlines seems to have gotten more common over time (say the past decade).

 

[BRiT]

Performance Comparisions @ https://forum.beyond3d.com/threads/gpu-ray-tracing-performance-comparisons-2021-2022.62346

 

[GhostofWar]

If what debauer is saying turns out to be true, the Gamers Nexus interview with the nvidia engineer talking about the vapor/condensation/hotspot considerations takes on a strangely fascinating new light.

 

[Kaotik]

It could be QC issue, since if it was just bad design we wouldn't have perfectly fine working cards, would we?

 

[Bondrewd]

It could be QC issue

https://twitter.com/x/status/1609848148058406912

Probably more of a subcontractor that makes said vapor chambers issue.

 

[Lightman]

https://twitter.com/x/status/1609848148058406912

Probably more of a subcontractor that makes said vapor chambers issue.

Yes, Der8auer himself in the video tested his good card and it was perfectly fine no matter how it was mounted. He theoretises either it's something to do with internal structure of the vapochamber or the amount of phase change liquid varies between samples.

 

[CarstenS]

Hopefully he'll be doing the easiest cross-check to prove his theory: swap the supposedly faulty cooler with one on one of the other cards.

 

[Deleted member 2197]

AMD partner suspects a faulty batch of reference Radeon 7900 GPUs is affected by thermal issues - VideoCardz.com

There might be a fault batch of reference AMD Radeon 7900 GPUs Igor Wallossek has provided a brief update to Der8auer findings. As a reminder, Der8auer has recently posted a video focusing on AMD Radeon 7900 reference card thermal throttling related issues. Although he has not yet provided a...

videocardz.com