RTX2080 . . . (MSI Ventus) Do you own one?

[motherboard1]

I managed 7159 @ those settings, no overclock. I doubt i'll get much more as my old 2600K is probably bottlenecking a bit.

Z

I wonder what the hell is going on on my end then? That's a big difference.

 

[Zuzzz]

I wonder what the hell is going on on my end then? That's a big difference.

Try the latest nVidia driver. I noticed slight improvements. I'm using 430.53. Being a rather new card Im guessing there will be bug fixes and as well some optimization of the drivers

Z

 

[yoadknux]

Thanks for your results. The 2080 is a really good card. When it launched it was slower than 1080ti according to some benchmarks, but now when things have stabilized, I see that it's about 10% faster (my 1080ti does 6450 with a stable overclock). Although I'm still on 399 drivers so maybe it'll improve a bit if I update.

Also, I just want to add that 77c for superposition is kinda hot (at least for Pascal), so maybe that's why you two have a difference in temps. Anything below 65c is definetly slowing down.

 

[motherboard1]

I just ran it on the new Driver No OC and got 6524... wtf ? 7031 with an OC.

Not impressed.

Edit: And now 6396. I watched the GPU Utilization of the program sit at around 85% for most of the test, which was previously @ 99%

Gonna go test on Windows 7

 

[Zuzzz]

I just ran it on the new Driver No OC and got 6524... wtf ? 7031 with an OC.

Not impressed.

Edit: And now 6396. I watched the GPU Utilization of the program sit at around 85% for most of the test, which was previously @ 99%

Gonna go test on Windows 7

This will be interesting. You are on stock clocks? 1515/1710? And yes my GPU was 99% as well.

Z

 

[motherboard1]

This will be interesting. You are on stock clocks? 1515/1710? And yes my GPU was 99% as well.

Z

I ran the test 4 or 5 times and with stock clocks and OC. Results are trash on the 430.39 Driver.

My CPU is always @ 4.4 I don't touch it.

-----

This new Screenshot is one run Windows 7 OCed 100Mhz/500Mhz Driver Version " 419.67 "

 

[motherboard1]

And now, with NO OC still on Windows 7 MSI afterburner closed I scored 6666. I then proceeded to OC again and scored 7165

When I OC I'm running a custom fan curve which is automatically present when MSI Afterburner is running. So maybe this is a Hot run Cold run situation plus the Fan curve altering results wildly I don't know.

Only thing I can do is run again with the new drivers on Windows 10 tomorrow morning when it's completely cold with no OC and see what happens then. But for sure that GPU Utilization 85% +/- 2 % looks way out of wack and doesn't occur at all on my runs with other driver/OS

 

[EarthDog]

Makes sense. YOu drop boost bins over ~65C so...

 

[motherboard1]

Makes sense. YOu drop boost bins over ~65C so...

Ya but that GPU utilization is at a solid 99% in the good runs and mostly 85 ish on the bad ones.

 

[Zuzzz]

Yes I also run a custom fan curve set to ramp up VERY fast @60C. These cards seem to 'down-clock' rather than 'throttle' so keeping the beast as cool as possible will yield higher and more stable GPU speeds. OC'ed or not, its all about the temps on these things.

Z

 

[motherboard1]

Yes I also run a custom fan curve set to ramp up VERY fast @60C. These cards seem to 'down-clock' rather than 'throttle' so keeping the beast as cool as possible will yield higher and more stable GPU speeds. OC'ed or not, its all about the temps on these things.

Z

What is the difference? I thought downclock was throttling ??

This morning I ran it stock on the newest drivers Win 10 and scored 7065. I just got off work and OCed it now and ran for a score of 6894. This test is all over the place it makes no sense and after about scene 10 this run GPU utilization tanked to the mid 80's Temps maxed out at 67'c which I'm certain didn't happen on this mornings run.

Edit : As always, there is something wonky going on here. My GPU utilization was in the low 80's for the first half of the test this time, though I set the fan curve to keep the Temp under 90'c. Still poor results in the 6000's and this time my Boost clock stayed +/- 20mhz of 2000. Whatever is choking my utilization is killing it.

Edit: I meant to say Fan curve to keep the temps under 65'c I don't know where 90 came from.

 

[yoadknux]

What is the difference? I thought downclock was throttling ??

This morning I ran it stock on the newest drivers Win 10 and scored 7065. I just got off work and OCed it now and ran for a score of 6894. This test is all over the place it makes no sense and after about scene 10 this run GPU utilization tanked to the mid 80's Temps maxed out at 67'c which I'm certain didn't happen on this mornings run.

Edit : As always, there is something wonky going on here. My GPU utilization was in the low 80's for the first half of the test this time, though I set the fan curve to keep the Temp under 90'c. Still poor results in the 6000's and this time my Boost clock stayed +/- 20mhz of 2000. Whatever is choking my utilization is killing it.

This is actually kinda weird. Out of all the benchmarks and stress tests I've tried, Superposition has always been the most optimized one. Relatively low power consumption, and can run high clocks consistently, and the results for me have been something like 6400+-50. So we're talking about at worst 1.5% deviation, usually less. Do you have a clue what's going on? Are other benchmarks, let's say Heaven, give you the same issues? Does the low usage occur after a while, or does it sometimes start with low usage and sometimes it doesn't?

Maybe you need to tweak some configurations with the nVidia driver (there are high performance settings) + set high performance power plan in Windows, but frankly it shouldn't be affecting this.

 

[motherboard1]

This is actually kinda weird. Out of all the benchmarks and stress tests I've tried, Superposition has always been the most optimized one. Relatively low power consumption, and can run high clocks consistently, and the results for me have been something like 6400+-50. So we're talking about at worst 1.5% deviation, usually less. Do you have a clue what's going on? Are other benchmarks, let's say Heaven, give you the same issues? Does the low usage occur after a while, or does it sometimes start with low usage and sometimes it doesn't?

Maybe you need to tweak some configurations with the nVidia driver (there are high performance settings) + set high performance power plan in Windows, but frankly it shouldn't be affecting this.

I've tried it with the Nvidia High performance setting as well as the Power Plan High performance. Doesn't seem to affect anything. BTW in a previous post I said Fan curve was set to keep temps under 90'c but that was some kind of typo, I meant to say under 65'c after a couple posters mentioned throttling starts @ 65'c Temps have since shown to change nothing. It's that GPU utilization thing I'm sure of that much. But why I don't know.

Yes it sometimes starts with low utilization and sometimes it kicks in towards the end, and sometimes it's off and on throughout the middle ish. I've ran the test so many times now and I've been keeping an eye on the utilization and there doesn't seem to be a predictable pattern at all. But if I get a run where it stays at 99% the whole way through I get a high score. OC or not it's really down to whether or not the utilization ****s off on me.

Can't be a CPU bottleneck because the graphs are showing the CPU utilization is pretty low for all cores. Well under 50%

The inconsistent results even followed me to another Driver version on another OS. I'm going back to Windows 7 to run a few more and watch the utilization again.

 

[yoadknux]

I've tried it with the Nvidia High performance setting as well as the Power Plan High performance. Doesn't seem to affect anything. BTW in a previous post I said Fan curve was set to keep temps under 90'c but that was some kind of typo, I meant to say under 65'c after a couple posters mentioned throttling starts @ 65'c Temps have since shown to change nothing. It's that GPU utilization thing I'm sure of that much. But why I don't know.

Yes it sometimes starts with low utilization and sometimes it kicks in towards the end, and sometimes it's off and on throughout the middle ish. I've ran the test so many times now and I've been keeping an eye on the utilization and there doesn't seem to be a predictable pattern at all. But if I get a run where it stays at 99% the whole way through I get a high score. OC or not it's really down to whether or not the utilization ****s off on me.

Can't be a CPU bottleneck because the graphs are showing the CPU utilization is pretty low for all cores. Well under 50%

The inconsistent results even followed me to another Driver version on another OS. I'm going back to Windows 7 to run a few more and watch the utilization again.

Maybe a PSU thing? Do you use a daisy chain connector or two separate PCI-E cables?

 

[motherboard1]

Maybe a PSU thing? Do you use a daisy chain connector or two separate PCI-E cables?

I use a single PCI-E cable W/daisy config, should I be using 2 separate cables? I didn't think that could effect power draw.

 

[Zerileous]

Yes you should. Well at least if the manual indicates it. Mine does advise against using the daisy chain for GPUs over a certain power draw. Technically each connector I think can draw 150W, but I'm just going by memory so you might want to check the spec.

When your usage is 80%, it's likely a CPU bottleneck. Is the 80% GPU utilization happening in both CPU overclocked and non-overclocked scenarios, or are you running 4.4GHz for both? Try it at 4K and see if it consistently uses the GPU. Not sure why you get the inconsistency from the CPU end, but it is similar to the inconsistency I encountered running Heaven which is very CPU constrained on my system.

 

[motherboard1]

Yes you should. Well at least if the manual indicates it. Mine does advise against using the daisy chain for GPUs over a certain power draw. Technically each connector I think can draw 150W, but I'm just going by memory so you might want to check the spec.

When your usage is 80%, it's likely a CPU bottleneck. Is the 80% GPU utilization happening in both CPU overclocked and non-overclocked scenarios, or are you running 4.4GHz for both? Try it at 4K and see if it consistently uses the GPU. Not sure why you get the inconsistency from the CPU end, but it is similar to the inconsistency I encountered running Heaven which is very CPU constrained on my system.

CPU is always overclocked. MSI Afterburners graphs don't indicate much action on any CPU core during the benchmarks AND if it was a CPU bottleneck I believe the GPU utilization drop off would occur in a consistent manner at specific points in the test.

 

[Zuzzz]

This seems to be the key. The 80% or less usage of the GPU during the test. I have not seen the GPU useage drop ever during the test. My results are also on an old CPU(i7--2600K). So I doubt that 6700k is bottlenecking. Find out why the GPU is dropping to 80% and I think we will find the answer. I am actually trying to duplicate the results but so far I cannot.

Z

PS: imho throttling lowers the work load while downclocking lowers the speed and maintains the full load.

 

[yoadknux]

Superposition is extremely GPU limited, and very well optimized at that. This is a driver issue or power issue.
Again, I would start by doing Heaven, Valley, Timespy, Firestrike, Sky Diver... And check usage.

 

[motherboard1]

Superposition is extremely GPU limited, and very well optimized at that. This is a driver issue or power issue.
Again, I would start by doing Heaven, Valley, Timespy, Firestrike, Sky Diver... And check usage.

The only one I have installed right now is Heaven. Are any of the others free ?

Here's is a graph from the Heaven Benchmark run on Extreme preset. I let it loop for 2 or 3 rounds, before I realize it wasn't in benchmark mode. Still the graph shows plenty of dips especially early on. You can see where the session ends and begins as well sorry about that. The line is obviously where I closed the utility

Definitely something going on here as well, notice how the dips in GPU usage seem to correspond with dips in Power Percentage.

Edit: Btw the dips are shorter here. It's not exactly what I'm seeing in SuperPosition in which the GPU utilization definitely stays low for extended periods, I'll throw up a graph for that too a little later on.