REALLY Weird DX12 results.

[Avro Arrow]

Ok, so... I ran PerformanceTest and noticed that my 3D scorea are really low. I realised that it was because I didn't have Windows set to 2160p, so I did that to avoid the score being penalised and ran it again, only to be baffled again.

My DX9 Score: 243 (85th percentile)
My DX10 Score: 132 (81st percentile)
My DX11 Score: 319 (95th percentile)
My DX12 Score: 44 (58th percentile)

Here are my system specs:

CPU: Ryzen 7 5800X3D
Motherboard: ASRock X570 Pro4
RAM: 32GB DDR4-3200 (running at XMP Profile 3200MHz)
Video Card: ASRock Radeon RX 7900 XTX Phantom Gaming OC 24GB <-Not exactly a joke of a video card, eh?
OS SSD: WD Black SN770 500GB PCIe4 NVMe

There has to be a problem with the testing methodology because my GPU Util percentage in the Adrenalin Overlay never goes above 45% during the Passmark DX12 test. It is difficult for me to believe that what is currently the second-most-powerful video card in the world is only hitting the 58th percentile in DX12. Hell, I run Spider-Man Remastered at 2160p FULLY MAXXED-OUT and my FPS never dips below 100.

I've attached my Unigine Superposition Score (DX12, 4K Optimised): 22191

Anyone have any ideas as to what's wrong?

[David (PassMark)]

See this page for general performance issues
https://www.passmark.com/support/per...erformance.php

Most common issues are,
- Thermal throttling / cooling
- G-Sync / Free Sync / Refresh rate frame rate limiting.
- Under powered or throttled CPU, bottlenecking the GPU
- Video card device driver issues
- Card in wrong PCIe slot, or PCIe slot at wrong speed
- Wrong monitor resolution for the test (should be OK in this case however)
- Slow RAM or single channel. This is more a problem with embedded video however.
- You are running older versions of PerformanceTest, which didn't load the video card as much.

Note that the quantity of video RAM doesn't help in the test. The test uses only around 1.5GB. So your 24GB doesn't help here.

[David (PassMark)]

Just for fun, I had a look at the result distribution for this video card in real world. It is super wide.

So lots of people who have this card seem to have it setup in some sub-optimal fashion, and there are lots of different setup options that give a lot of different performance levels. Average is 116fps.

[Avro Arrow]

[QUOTE=David (PassMark);n55794]See this page for general performance issues
https://www.passmark.com/support/per...erformance.php

Most common issues are,
- Thermal throttling / cooling

I just ran the test, the card never even reached 40°C so that can't be it.

- G-Sync / Free Sync / Refresh rate frame rate limiting.

I don't have an adaptive sync display. My display is 2160p60Hz so that can't be it.

- Under powered or throttled CPU, bottlenecking the GPU

It's a AMD Ryzen 7 5800X3D with an average temperature of 43°C during the test. It's a prime gaming CPU so that can't be it.

- Video card device driver issues

That's possible but unlikely since I was playing Spider-Man Remastered maxxed out at 4K while never dropping below 100FPS (not something that a driver problem would allow). Still, installing a new driver set wouldn't hurt so I'll give that a shot. The drivers that I have are slightly out-of-date (1 month max) and probably aren't the problem but I'll swap in the newer package anyway.

EDIT: As expected, the score remained EXACTLY the same with the new Adrenalin Package. Your benchmark therefore has a bug.

- Card in wrong PCIe slot, or PCIe slot at wrong speed

No, that's not it either. The card is in the primary (armoured) slot on my X570 Pro4 and GPU-Z has verified that it is running at PCI-Express v4.0 x16.

- Wrong monitor resolution for the test (should be OK in this case however)

I don't normally keep my display at 2160p (because everything gets REALLY small) and I did initally get penalised for that but then I set it to 4K and ran it again. The scores I showed were at 4K. It does change resolution correctly if I'm only running a single test. A full suite requires me to set Windows to 4K.

- Slow RAM or single channel. This is more a problem with embedded video however.

32GB of DDR4-3200 running stable at their XMP Profile speed of 3200MHz. Slow RAM would also be mitigated by the 3D V-Cache of my R7-5800X3D so that can't be it either.

- You are running older versions of PerformanceTest, which didn't load the video card as much.

Version 11.0, downloaded today. The version I had before was almost two years old and I knew better than to use it.

Note that the quantity of video RAM doesn't help in the test. The test uses only around 1.5GB. So your 24GB doesn't help here.

No, but the ungodly amount of GPU horsepower should. The DX11 and GPU Compute tests both confirmed that.

I really appreciate you trying to help me. The PCIe speed was something I hadn't considered (but ended up eliminatiing with GPU-Z) and I'll install a new driver package. Unfortunately, none of the other possibilities held any water. I really wish that they had.

Keep in mind though that this isn't an emergency for me. My gaming performance is nothing short of incredible. I reported this more as a bug to you than as a problem for me. I get great scores on every benchmark except Passmark. If my setup were the issue, I would have RX 6600-level scores in ALL benchmarks, not just Passmark (that's just how tech logic works).

I've been building PCs since 1988 (Harris-branded 80286-16MHz) and am not prone to the usual noob errors like (gasp!) single-channel RAM or trying to use an RX 7900 XTX with a Devil's Canyon i5-4670K (and wondering why my FPS is so much lower than Steve Burke's....LOL ). I assure you that whenever I encounter something that I can't solve myself, it's never something simple (although I often wish it were).

Ok, I looked at the established baselines and I'm scoring between the RX 470 and the RX 580. There's not a chance in hell that I am running my card in "such an un-optimised state" that I could get such a low score.

There's a problem with the benchmark and the fact that you've seen such a wide range of scores for such a high-end card while ALL other benchmarks have no issue with it whatsoever is proof of that. This is now a bug report.

Oh, one other thing.." Problem: Slow 3D Graphics


Solution: Make sure you have the latest copy of the device driver for your video card and the most recent version of DirectX. Video device drivers are complicated pieces of software and as a result, it is very common that they are buggy or unreliable. If you have the latest device driver and are still encountering problems, consider rolling back to a previous version as it may be more stable.

Also check you video hardware acceleration is turned on from the Windows, "display properties" dialog and in DirectX (use Dxdiag.exe for this)."

You might want to remove this from the "Common issues" page as nobody uses Windows XP anymore and not even the W7 Control Panel has "Display Properties" as a choice. Also, everyone with W7 or newer has DX12 so that's also pretty anachronistic.

[David (PassMark)]

Same DX12 benchmark has been run on millions of machines now. So a systematic bug effecting majority of machines is now pretty unlikely. So much more likely it is something special about your machine.

Other slightly less common issues
- 3rd party overlay software. Most are OK nowadays. Some aren't.
- Video capture software running
- Screen sharing / streaming software / remote desktop software. Sometimes people even forget they have this stuff running
- General crapware running in the background
- Running the tests on the wrong video card. Sometimes the integrated card might be used if not disabled, or sometimes users have more than 1 discrete card installed.
- Windows power settings (more commonly a problem on laptops)
- Device driver settings overriding application settings to force higher quality than PerformanceTest requests
- From time to time we see driver bugs that only appear in certain setups or certain driver versions. For example, dual monitors only, HDMI only (but not display port), HDR turned on, DPI scaling, etc..

Post a debug log of you like, maybe we'll notice something amiss
https://www.passmark.com/support/per.../debug-log.php

Can you also save a baseline file with all the benchmarks in it (from the File menu)

[BRENNANX]

Oh sh*t you serious Dave,does the rx 7900 get 112 fps in dx12? and I was so proud of my 89fps with my rx6600xt lol...suppose it stands to reason as the cost of a rx7900 down here is the same as a good 2 year old car....keep well...B