Extremely low DIRECT2D marks for RTX3070 TI

[Mike17]

Hi,

I swapped my GTX 1050 Ti against a RTX 3070 Ti and of course tried the Performance Test.

While the RTX 3070 Ti test results for 3D graphics was at least 10 times higher than those of the GTX 1050 Ti, the results in DIRECT2D were abysmal despite low CPU (on all cores) and GPU load.

I am running W7x64 with the latest NVidia drivers and I am aware that some tests (like PDF and SVG) cannot be run on W7 due to missing OS APIs.

But I fail to understand why the RTX 3070 Ti performs much worse than the GTX 1050 Ti for DIRECT2D in the same PC running the very same OS without any obvious load and perfcut inactive.

Any ideas? Maybe a driver issue?

[Tim (PassMark)]

When you swapped the 3070 in did you do a full clean install of the Nvidia driver (I believe there is a checkbox for this as part of the installer)?

It could also be that no effort was made for the 3070 driver to maximise performance in Win7 since that card was released 10 years since the final update to Win7.

*Update*
We only have 1 result for the 3070 in Win7 (i'm assuming this was your submission) so there aren't any other results to compare to and I'm guessing I found the 1050 result for your system as well.

The 3070 is performing better in most of the tests, including Direct2D (though not by much), so I don't see how you can say it's performing worse. It's mostly that the performance gains can't affect the overall 2D result much due to the missing tests. In a newer version of Windows the 3070 averages over 1000 with a top of 1800 for the 2D score while the 1050 averages around 650 with a top of 1400.

I believe there have been some Direct2D performance improvements between Win7 and Win10/11 as well so the Direct2D results are better in those versions of Windows.

[Robert Childers]

Im a newbie, but Ive got to say something here. I have a above avg system with a Msi 3070ti GamingX. My 2D has been at a embarrasing 560-609 except for a few times installed new hardware or maybe when windows had a major update. At that time the score was an incredible 1400. There was no way to isolate the reason. Ive diagnosed, tweaked, updated, changed drivers, you name it. 99% of your inquiry on that will be answered "Ive tried that 4 times". But Im not quitting. Ive even tried tweaks from other posts, down to disabling prefetch. How could there be another cause of this with good eqt besides the testing. AND i havent looked but It seems many people have this. Are you suspecting anything?

[Robert Childers]

Regarding the previous post 12;44am, the harware utilized include: 10700kf @5.09gz, Msi 3070ti GamingX, Oloy 32gbddr4@3333mz, Evo970Plus 2tb, Corsair RM850, Lg 4k 32UK-T50.

[Robert Childers]

RE: Previous Post "IM a newbie"
There was a time when 2d skyrockets. I have an LG driver which is avalaable on-site. It includes the icc. I tried with a different icc, no help. BUT when the generic driver is used on my 4k monitor 2d goes up. It would be convenient to say "well then thats LG's fault", but how do I know the driver is wrong, it comes off the LG site. When I use the generic my color management will NOT connect to the monitor, so now I have to use it!!
Is 2d supposed to rise when color management is not connected anyway?

[David (PassMark)]

Is 2d supposed to rise when color management is not connected anyway?

It isn't something we have looked at.
I don't know enough about how Windows deals with ICC colour profiles to know what the performance expectation should be.

[Mike17]

Sorry for my late reply. I thought I had set up notification but I didn't get any so I thought nobody had answered.

When you swapped the 3070 in did you do a full clean install of the Nvidia driver (I believe there is a checkbox for this as part of the installer)?

No. I did not notice such an option. I once used some gfx driver removal tool on a different PC but that didn't seem to even work correctly.

It could also be that no effort was made for the 3070 driver to maximise performance in Win7 since that card was released 10 years since the final update to Win7.

That is what I _think_.

*Update*
We only have 1 result for the 3070 in Win7 (i'm assuming this was your submission) so there aren't any other results to compare to and I'm guessing I found the 1050 result for your system as well.

I don't believe that I was able to upload results because that requires some kind of account and I didn't even have this forum account by then.

I believe there have been some Direct2D performance improvements between Win7 and Win10/11 as well so the Direct2D results are better in those versions of Windows

This may well be. But a worse performance on the same system I think must be a driver issue. And I am still wondering about low system load. Just think: if the 3070 driver runs slower somehow for DIRECT2D on W7, how fast would it run in W10 if it was as fast as the older 1050 driver ?

However a friend of mine and I am planning to implant the 1050 and the 3070 into a W10 system and perform the tests there. But I've a lot of work and he has got a cold, so we will not manage to meet earlier than the weekend after the next one.

[David (PassMark)]

I don't believe that I was able to upload results because that requires some kind of account

No account is required.

But there is an option of having an account and seeing all your results in that account (and the performance changes over time).

[Mike17]

Last weekend we found the time for some tests.

My PC is W7x64 running a Core i5 6th generation which despite being old and week was sufficient to run games like Valheim, Minecraft and Roblox without full load on any core.

My friend's PC is W10x64 running a Core i7 6th generation and a GTX 970 which despite its age is performing really fine (of course it looses against a RTX 3070 in 3D).

In my PC the GTX 970 and the RTX 3070 TI perform almost the same in 2D (some 20 points and SVG and PDF not being available).

In my friends PC the GTX 970 and the RTX 3070 TI perform almost the same as well (750 against 780 points) in 2D.

So it looks like 2D performance does not depend on the GFX card but on either CPU or Windows version.

Next test will be swapping the Core i5 6th generation against a Core i7 7th generation in my PC.

I'll keep you informed.

[David (PassMark)]

It is correct that
1) You don't need a high end video card for fast 2D
2) CPU has a big impact

But you might find the results are different in Win11 compared to Win7. As more tests (and newer versions of DirectX) are available.

[Mike17]

Only yesterday I got an i7, 7700K @4.2GHz and I swapped it against the i5 this morning.

After running the performance test I noticed that I seem to have forgotten to save the baseline with the i5, so I cannot compare any more. However the 2D result was 27.

However searching my ptx files I found two about the i5 running along with the GTX 1050 Ti, one created using Performance Test v8 and the other using v10 (on the very same system).

Funnily enough it is interesting to see, that v8 awarded 816 points to 2D and v10 only 17 points, while in all other groups the results are only slightly different.

Chears.