7900 XTX Seriously lacking

[LabRat 891]

It was anti-lag+ that was getting VAC for editing files.

Correct. Which is part of the Hypr-RX pre-set package.

 

[faye]

Hyper-RX has gotten people VAC banned. AMD says they fixed that, but I do not trust any of their 'pre-packaged pre-sets' any longer.

That was Anti-Lag+ specifically, it has been removed. I agree it was a mess, but Adrenalin isn't dangerous.

 

[Steevo]

 

[Redwoodz]

I tested many many different combos. Some underpowered just to compare. Best score seemed to be 3000mhz 1035Mv 2700mem and +15% power but that still does not get me close to factory scores consistently. One out of like 15 benchmarks I was close but couldn't get it to do it again.


I don't think it was cpu. This is the exact issue I have had with the gpu since the get go.

Yes, I have ran mem test and it is stable. Xmp enabled

I have had VRAM errors cause some of your symptoms very intermittantly, inc Windows corruption. Could be a bad chip. Maybe try dropping VRAM down even further.
Whatever is going on, it should be consistent if everything is stable.

 

[Zach_01]

why is the power limit -4?, (on the GPU) should be +15…

And it should be reading 468 in the adrenaline overlay. (7900xtx merc310) (Well, my adrenaline overlay says that)

Not sure if you wrote this for my post (but most likely).
I have the -4% power limit. Can’t run power limit to +15% (460+W) on my Nitro+ because of my 750W PSU, so I keep GPU card power under 400W. Plus the extra ~70W won’t give me much more actual gaming performance other than benchmark scores. Also worse thermals too…

 

[Ferrum Master]

I have to chime in too. As a fellow RDNA user. Please excuse for my OpenGL usage, thus the lower scores.

Swapping to newer/less used PCIe power leads, resolved my issue.

I had actually the same problem, I just readjusted the pins using a special rectangular needle. If I would ever mention it, I would be eaten alive. But the real cause ain't the hot wires, but the ground return path. Think about it mate. You have the theory basis well learnt and how it goes hand in hand in with system stability over few GHz realm.

 

[Steevo]

I have to chime in too. As a fellow RDNA user. Please excuse for my OpenGL usage, thus the lower scores.

View attachment 345988



I had actually the same problem, I just readjusted the pins using a special rectangular needle. If I would ever mention it, I would be eaten alive. But the real cause ain't the hot wires, but the ground return path. Think about it mate. You have the theory basis well learnt and how it goes hand in hand in with system stability over few GHz realm.

Grounding is always the issue, voltage is relative to ground and ground at the load is not the same potential as ground at the sink.

 

[mama]

If it helps for comparison. Direct X at 4K. Default settings. Current driver version (driver only install).

 

[DonnieDangerous]

I ran a bunch more tests on individual settings and found that the exact setting that decreases benchmarks by 20% is the "AMD Fluid Motion Frames" enabled setting.

 

[AnotherReader]

I ran a bunch more tests on individual settings and found that the exact setting that decreases benchmarks by 20% is the "AMD Fluid Motion Frames" enabled setting.

In my experience, it's best to enable settings like these on a game by game basis.

 

[Zach_01]

I ran a bunch more tests on individual settings and found that the exact setting that decreases benchmarks by 20% is the "AMD Fluid Motion Frames" enabled setting.

FMF aka FrameGeneration
Runs on driver level so the actual game/benchmark is not aware of it.
Almost doubles frames on screen. What you see as lost performance is the loss of the calculation of generated frames.

Typicaly (by avg) if without FMF you get (for example) lets say 70FPS on a game, when you enable it the output FPS to the monitor would be 70 - 15~20% x 2 = around 110~120FPS

In game/benchmark FPS counters cannot see the generated frames. You need an external counter like AMD overlay, MSI afterburner and such...
So its not meant for benchmarking. Only to make actual gameplay a bit smoother.

 

[Lew Zealand]

FMF aka FrameGeneration
Runs on driver level so the actual game/benchmark is not aware of it.
Almost doubles frames on screen. What you see as lost performance is the loss of the calculation of generated frames.

Typicaly (by avg) if without FMF you get (for example) lets say 70FPS on a game, when you enable it the output FPS to the monitor would be 70 - 15~20% x 2 = around 110~120FPS

In game/benchmark FPS counters cannot see the generated frames. You need an external counter like AMD overlay, MSI afterburner and such...
So its not meant for benchmarking. Only to make actual gameplay a bit smoother.

So tested with FMF/FG on compared to off, benchmarks give a reasonable estimate of the number of real frames lost with FMF as well as its corresponding latency penalties. Though I hear that latency "feels" more than ~20% worse than native.

 

[Zach_01]

So tested with FMF/FG on compared to off, benchmarks give a reasonable estimate of the number of real frames lost with FMF as well as its corresponding latency penalties. Though I hear that latency "feels" more than ~20% worse than native.

I hear you
What I described is at least what the principal of FMF is.

This video talks about FMF on FSR3 specifically

 

[Veseleil]

This thread is the example why forums are so valuable resource to have, as long as community is healthy and has a positive response. The OP came asking for help, and now he helped all of us by making some useful info come to the surface. The importance of getting problems solved, followed by a proper feedback, is immense.

 

[faye]

I ran a bunch more tests on individual settings and found that the exact setting that decreases benchmarks by 20% is the "AMD Fluid Motion Frames" enabled setting.

There's basically no reason to ever turn that on, anyway. It's frame interpolation, we have been turning off the equivalent built in to displays for years.

 

[DonnieDangerous]

I do appreciate everyone's help with this issue. I was able to pick up a 7950x3d for a great price. So, even though I was planning to wait until the 9000 came out I couldn't resist. I ordered a MSI MAG x670E Tomahawk MB and G.Skill Tirdent Z Neo RGB 2x16gb 6000mhz RAM. Will post test results with a fresh OS install just for comparison sake.

 

[Steevo]

I do appreciate everyone's help with this issue. I was able to pick up a 7950x3d for a great price. So, even though I was planning to wait until the 9000 came out I couldn't resist. I ordered a MSI MAG x670E Tomahawk MB and G.Skill Tirdent Z Neo RGB 2x16gb 6000mhz RAM. Will post test results with a fresh OS install just for comparison sake.

With X3D and new BIOS running 64GB of RAM AT 6000+ with good timings is easy FYI.

 

[Toothless]

With X3D and new BIOS running 64GB of RAM AT 6000+ with good timings is easy FYI.

Not with the dual CCD chips.

 

[Makaveli]

Timespy scores are abit low for sure this is my run that is many months old now so older driver.

GPU is reference normal clocks.