Low 2D Performance scores

[Globespy]

If you have RTX 4090, then your 2D and 3D scores are both low. The average 3D Mark for RTX 4090 is around 39,000. For the 2D tests, all the tests score lower, but the Image Filters and Image Rendering are magnitudes lower than what is typical seen for RTX 4090.

What OS are you using? Is cooling adequate? Power Supply providing enough power? Is the video card in correct PCIe slot?

Win 11 22H2 (Fresh Install), Fantastic cooling, 1200W Platinum PSU, yeah it's in the correct PCIe slot.
Gaming performance is stellar and in line with how the GPU should be performing, so that just leaves the elephant in the room which is something must be going wrong with your software.
Sorry to point the finger back, but this is the only test software that is showing laughable results.

[David (PassMark)]

We have had the software run on 900 different systems with a GTX 4090.
The average 3D result was 39,261 (median result 40,275).
Your result was 11,157.

Won't even mention the other million machines the same software ran fine on......

So while the software might not be perfect, you need to explain how it can run fine on 900 other systems, but not yours. The odds are that greatly in favour of there being something 'special' about your machine. Blaming the messenger is a common reaction however.

[Globespy]

We have had the software run on 900 different systems with a GTX 4090.
The average 3D result was 39,261 (median result 40,275).
Your result was 11,157.

Won't even mention the other million machines the same software ran fine on......

So while the software might not be perfect, you need to explain how it can run fine on 900 other systems, but not yours. The odds are that greatly in favour of there being something 'special' about your machine. Blaming the messenger is a common reaction however.

Touche!
It's all ok, perhaps there's something odd going on that only seems to show itself in this benchmark test. I am willing to fully accept that as a possibility.
That said, I don't have any performance issues in day to day use, have never seen a BSOD or other unexplained oddity with this current build, temperatures are stellar, all the expected performance is there, and for all intents and purposes it's the best PC I've built in over 20 years of building PC's as an enthusiast.
I appreciate the opportunity to try out your software for 30 days, but I trust you can also acknowledge that there's literally no red flags that would warrant me going down rabbit holes to identify something that likely just isn't there, and this result was perhaps a very odd fluke. PC's can be odd sometimes.
If there's something that you perhaps think of that I could check, feel free to reply - maybe I missed some setting I was supposed to change in the NVCP or something, because it does seem truly odd that the strongest component in my build was the only one that seemed to have a 'bad' moment, then get on with business as usual?
Thanks.

[David (PassMark)]

If the performance meets you expectations, then no need to spend time 'fixing' it.
I would fully expect it to be some config issue (probably not a hardware fault). But there are dozens of factors that influence the 2D /3D results.

If you look at the distribution of 4090 results there is a huge range. Many many people are running these cards in a sub-optimal configuration (and some are pushing the envelope on the high end as well).

[Globespy]

David (PassMark) - thanks again for the reply.
It's indeed very strange, I ran all the other usual benchmarks (I purchased all the 3D mark ones, and tested Firestrike Extreme, Port Royal, Speed Runway etc.) and scores are all above average.
I read through your troubleshooting page, checked that 3D acceleration was on in DxDiag, I don't use any overlays or skins (forgot about Window Blinds - blast from the past, which I never used either) and I'm pretty much stumped.
Like I said, it's a fresh installation of Win 11 22H2 with nothing on my main M.2 NVME that I keep exclusively for the OS and the usual motherboard chipset drivers and the like.
I checked the drives and everything is fine.
The only thing I can think of is that I don't use Windows's pagefile (64GB DDR4 on a gaming PC), but I can't see why that would be the issue.

Very odd.

Current Nvidia driver is 526.56 - I may try the one that came out last week to see if that makes a difference, but unlikely

[David (PassMark)]

It is covered elsewhere, but the most common problems are,
- High temperatures (i.e. poor cooling) leading to thermal throttling of CPU / GPU.
- Wrong PCIe slot (or a slot that looks OK, but actually only functions as a x1 rather than x16. e.g. bad link training, or BIOS setup, dirty connector)
- Right PCIe slot, but card is, for example, PCIe 4.0 in a PCIe 3.0 bus.
- Wrong device drivers or driver bugs
- All manner of dubious 3rd party software. Including streaming apps, overlays, frame rate counters, screen sharing, remote desktops, virtualization, windows skins / themes, etc...
- Under powered CPU (compared to the GPU). i.e. bottlenecks.
- Bad RAM setup. e.g. not in dual channel.
- Not enough video RAM, so slower system RAM might get used.
- Frame rate limits due to monitor refresh rate and GSync like features.
- Resource hogging background apps
- Missing OS functions in old versions of windows, like the ability to render PDF files (in Win7).
- Windows power saving setting (mainly in laptops)
- Rendering running on the wrong video card (where both the integrated and discrete card are available). Mostly a laptop issue
- 'Bad' or unexpected device driver settings, forcing features like morphological filtering / full screen anti-aliasing on even when the app doesn't ask for it to be on. There are dozens of possible settings like this people turn on then forget about.
- Power supply not being adequate. (but the most common result from this is a crash or failed boot).
- Unexpected overclocking impacts (clock speed is faster, but performance is lower)
- Messed up Windows install.

But there are also other rare issues. Some people reporting results vary depending on the Windows desktop resolution. Some saying ICC colour profiles effect things. Some saying running multiple desktops effect things. Some people (rightly) saying that recent CPU/OS security patches effected things, sometimes there is also a real hardware failure as well, etc, etc..... list is huge unfortunately.

[Globespy]

David (PassMark)- I was reading through this and something stuck out to me.
"some people reporting results vary depending on Windows desktop resolution" - I'm running an 3440x1440 ultrawide (21:9) with DLDSR at 2.25 supersampling that results in a resolution of 5160x2160 which is what I use to play all games as the 4090 was literally at 40-45% utilization otherwise.
Maybe that's it.
Thanks, I am going through the rest of the list, some are things I know are ok, some I have to dig into a little more.

[AntonG]

I think it would be a good idea to allow future Performance Test version to be configured to completely skip the 2D benchmark section. Too many variables contribute to low 2D scores on high-end machines. Frankly, one could make the case to simply remove that section of the benchmark all together.

[David (PassMark)]

Too many variables contribute to low 2D scores on high-end machines

Frankly, one could make the case to simply remove that section of the benchmark all together.

But the purpose of benchmarking is to measure performance and then improve it (or allow comparison between different machines or settings).
So the more complex the area the more useful benchmarking is.

If it turns out that very expensive video cards (and some are extremely expensive nowadays) don't produce much performance benefit, then this is very useful information. Not something we should hide. I suspect some people are shooting the messenger here as they don't like the message.

But what we have done in the PerformanceTest V11 (coming soon) is rebalance the overall PassMark rating value, to reduce the weight of the 2D test in the overall system benchmark result.

[AntonG]

But the purpose of benchmarking is to measure performance and then improve it (or allow comparison between different machines or settings).
So the more complex the area the more useful benchmarking is.

If it turns out that very expensive video cards (and some are extremely expensive nowadays) don't produce much performance benefit, then this is very useful information. Not something we should hide. I suspect some people are shooting the messenger here as they don't like the message.

But what we have done in the PerformanceTest V11 (coming soon) is rebalance the overall PassMark rating value, to reduce the weight of the 2D test in the overall system benchmark result.

If you've re-evaluated the 2D weighting, to "rebalance" as you've put it ... doesn't that sort of hint to something along the lines of being Subjective? I'm not being contrarian on purpose here ... just trying to understand. Logic dictates that if the weighting factors are rejiggered, than something was determined to be amiss.

[AntonG]

Just ran the V11 Beta. With the weighting 'rebalanced', my overall Passmark Rating jumped from a V10's (Build 1016) score of ~8500 to 13891.

[David (PassMark)]

doesn't that sort of hint to something along the lines of being Subjective? I'm not being contrarian on purpose here ... just trying to understand. Logic dictates that if the weighting factors are rejiggered, than something was determined to be amiss.

Yes, it is subjective. How important 2D,3D, Disk speed and everything else varies between different people and different computing tasks. So there is no one value / weighting that makes sense for everyone. 2D isn't important at all for a file server. But it is very important for Photoshop and Microsoft Word.

But over the last 10 years 2D performance hasn't improved much. While disk performance has got like 10x better. So 2D was looking like the bottleneck.

my overall Passmark Rating jumped from a V10's (Build 1016) score of ~8500 to 13891

What was your CPU model, GPU model and 2D score?

[AntonG]

5950x
6750 xt

[David (PassMark)]

AntonG,

Your 2D result is on the low side, relative to your other components. So overall score looks a lot better if 2D isn't given as much weight. Your improvement in DiskMark also helps.

Here is a graph of the 2DMark for systems with your 5950X CPU with a selection of different video cards. These were just randomly selected baseline submissions. Your system is below the slowest machine for 2D. 925 isn't really a bad result. But is weak relative to the other high end hardware you have.

[AntonG]

Thanks for posting that up David.

Yes, I'm not that concerned about the real-world, gross effect of a score that is low. Eveything I do on this machine performs sollidly. I do quite a bit of photo editing, video editing and audio editing on this machine and everything runs quick & smooth. I've tried disabling Freesync Premium Pro on the monitor and in the GPU control software, along with reducing Windows 'fluff' in the Advance Visual Effects section to try to bump the 2D score... but nothing makes a significant impact. It piques my curiosity to know what has the biggest impact... but I'm not compelled to chase it.

The 6750 XT performs like a beast in 1440p gaming, so I'm content.

I am a little shocked to see all those RTX 3090's near the bottom tho...