Does Ryzen bottleneck a 1080ti?

[Ex14]

Hi, It's been awhile since I last posted here, but ran into an interesting situation at work that had me scouring the internet to no avail.

I'm working with an SI right now and Ryzen builds are still relatively hard to come by. So when an order came through for a pretty balling 1800x and 1080ti SLI config I was pretty stoked.

Unfortunately we ran into some pretty puzzling performance scores. With the system performing badly. How badly? With SLI the scores were similar to a SINGLE 1070. With a single card, It performed about 1080 levels only. We tested the cards on another system and it posted normal results in line with what we expected. We also tested the affected system with a 1070 and the scores for the 1070 were perfectly fine

Has anyone else seen any news item or article talking about this? We even swapped CPUs to ensure it wasn't a faulty CPU.

Specs are as follows

1800x

generic thin 120mm AIO cooler.
Asus X370-Pro

16 2133MHZ RAM x 4
850W EVGA G3 PSU
1080ti FE x 2
Using a HB SLI Bridge (non LED)

We tested @ 1080p resolution with heaven @full screen high settings no AA and no tessellation, and just a normal run of 3dmark11

Edited June 22, 2017 by Ex14
added testing methodology

[Nocte]

2133MHz RAM on Ryzen is kinda bad. Ryzen performance scales a lot with RAM speed. 

 

Not sure if that's the only reason though. 

[theminnesotanman]

7 minutes ago, Ex14 said:

Hi, It's been awhile since I last posted here, but ran into an interesting situation at work that had me scouring the internet to no avail.

I'm working with an SI right now and Ryzen builds are still relatively hard to come by. So when an order came through for a pretty balling 1800x and 1080ti SLI config I was pretty stoked.

Unfortunately we ran into some pretty puzzling performance scores. With the system performing badly. How badly? With SLI the scores were similar to a SINGLE 1070. With a single card, It performed about 1080 levels only. We tested the cards on another system and it posted normal results in line with what we expected. We also tested the affected system with a 1070 and the scores for the 1070 were perfectly fine

Has anyone else seen any news item or article talking about this? We even swapped CPUs to ensure it wasn't a faulty CPU.

Specs are as follows

1800x

generic thin 120mm AIO cooler.
Asus X370-Pro

16 2133MHZ RAM x 4
850W EVGA G3 PSU
1080ti FE x 2
Using a HB SLI Bridge (non LED)

The ram speed is pretty low. Generally recommended to use 3000mhz or bigger cause it's been shown to greatly improve performance on Ryzen systems 

[xitywampas]

Ryzen is so new who really knows. I don't know anyone with a 1800x at this point. Still waiting to get hold of one of those myself. Wish I could help but will be following this closely.

[theminnesotanman]

9 minutes ago, Ex14 said:

Hi, It's been awhile since I last posted here, but ran into an interesting situation at work that had me scouring the internet to no avail.

I'm working with an SI right now and Ryzen builds are still relatively hard to come by. So when an order came through for a pretty balling 1800x and 1080ti SLI config I was pretty stoked.

Unfortunately we ran into some pretty puzzling performance scores. With the system performing badly. How badly? With SLI the scores were similar to a SINGLE 1070. With a single card, It performed about 1080 levels only. We tested the cards on another system and it posted normal results in line with what we expected. We also tested the affected system with a 1070 and the scores for the 1070 were perfectly fine

Has anyone else seen any news item or article talking about this? We even swapped CPUs to ensure it wasn't a faulty CPU.

Specs are as follows

1800x

generic thin 120mm AIO cooler.
Asus X370-Pro

16 2133MHZ RAM x 4
850W EVGA G3 PSU
1080ti FE x 2
Using a HB SLI Bridge (non LED)

Also the 850w does seem a little low. Most sli builds want 1000w or higher.

[Ex14]

26 minutes ago, Eibe said:

2133MHz RAM on Ryzen is kinda bad. Ryzen performance scales a lot with RAM speed. 

 

Not sure if that's the only reason though. 

Forgot to mention the additional stuff I did OTL My bad. Switched out the RAM to 2666 Kits (corsair Vengeance LED) and while a slight boost was not the levels of performance we expected it to be.

Also I overclocked the processor to 4GHz to see if the clock speed was the factor but that too netted little results.

[Ex14]

Just now, theminnesotanman said:

Also the 850w does seem a little low. Most sli builds want 1000w or higher.

I was suspecting that too....but that doesn't justify the low single card scores as well

Granted Heaven is not a great benchmarking tool, But even 3Dmark scores were pretty bad.

[Herman Mcpootis]

2 minutes ago, theminnesotanman said:

Also the 850w does seem a little low. Most sli builds want 1000w or higher.

https://www.overclock3d.net/reviews/gpu_displays/asus_gtx1080_ti_strix_oc_sli_overkill3d_review/2c 656w with a 6850k and overclocked 1080ti SLI, the 850w G3 is more than enough.

[Ex14]

added testing methodology on first post 

[Herman Mcpootis]

3 minutes ago, Ex14 said:

added testing methodology on first post 

have you checked temps, gpu/cpu usage, etc. with msi afterburner?

[WereCat]

3 minutes ago, Ex14 said:

added testing methodology on first post 

Something is wrong because besides Physics test the 3DMark doesn't really care about the CPU and likewise the Heaven.

It will affect the total score but not the graphics score.

Besides, 3DMark crushes the physics score with Ryzen so there must be something wrong with that system.

[Ex14]

4 minutes ago, herman mcpootis said:

have you checked temps, gpu/cpu usage, etc. with msi afterburner?

Temps for GPU and CPU are fine.

For usage wise I have not checked on GPU yet (might do it later on when i'm done with my other stuff)

But CPU Wise only one or two core is loaded at a time (100% usage)

[Ex14]

3 minutes ago, WereCat said:

Something is wrong because besides Physics test the 3DMark doesn't really care about the CPU and likewise the Heaven.

It will affect the total score but not the graphics score.

Besides, 3DMark crushes the physics score with Ryzen so there must be something wrong with that system.

We look at only the Graphic score for 3D Mark since this is thje portion we're more concerned with testing at the moment. I'll grab some screen caps later maybe

[WereCat]

1 minute ago, Ex14 said:

We look at only the Graphic score for 3D Mark since this is thje portion we're more concerned with testing at the moment. I'll grab some screen caps later maybe

I would also like to mention that I watch Buildzoid YT videos (he is an extreme overclocker) and he managed to break some records in 3DMark with 1070 on LN2 and Ryzen 7 CPU on water.

[tiggar47]

39 minutes ago, Ex14 said:

Hi, It's been awhile since I last posted here, but ran into an interesting situation at work that had me scouring the internet to no avail.

I'm working with an SI right now and Ryzen builds are still relatively hard to come by. So when an order came through for a pretty balling 1800x and 1080ti SLI config I was pretty stoked.

Unfortunately we ran into some pretty puzzling performance scores. With the system performing badly. How badly? With SLI the scores were similar to a SINGLE 1070. With a single card, It performed about 1080 levels only. We tested the cards on another system and it posted normal results in line with what we expected. We also tested the affected system with a 1070 and the scores for the 1070 were perfectly fine

Has anyone else seen any news item or article talking about this? We even swapped CPUs to ensure it wasn't a faulty CPU.

Specs are as follows

1800x

generic thin 120mm AIO cooler.
Asus X370-Pro

16 2133MHZ RAM x 4
850W EVGA G3 PSU
1080ti FE x 2
Using a HB SLI Bridge (non LED)

We tested @ 1080p resolution with heaven @full screen high settings no AA and no tessellation, and just a normal run of 3dmark11

Overclock the RAM and see the perfomance.

[i_build_nanosuits]

48 minutes ago, Ex14 said:

We tested @ 1080p resolution with heaven @full screen high settings no AA and no tessellation, and just a normal run of 3dmark11

well...1080p for a GTX 1080ti is a very low resolution...i'm not at all suprised by your results.

If you test with a faster CPU like an i7-6700K or 7700K for example the performance will be better...this does not mean ryzen is inadequate, it means you are using a 700$ graphics card on a 100$ monitor.

[WereCat]

10 minutes ago, i_build_nanosuits said:

well...1080p for a GTX 1080ti is a very low resolution...i'm not at all suprised by your results.

If you test with a faster CPU like an i7-6700K or 7700K for example the performance will be better...this does not mean ryzen is inadequate, it means you are using a 700$ graphics card on a 100$ monitor.

That would be true if they didn't tested just 3DMark and Heaven

[i_build_nanosuits]

Just now, WereCat said:

That would be true if they didn't tested just 3DMark and Heaven

1080p is 1080p...whatever you run at this resolution likely won't max out a 1080ti until you have a top-end overclocked gaming CPU.

[WereCat]

5 minutes ago, i_build_nanosuits said:

1080p is 1080p...whatever you run at this resolution likely won't max out a 1080ti until you have a top-end overclocked gaming CPU.

Please, graphics score in 3D Mark scales almost perfectly with GPU performance regardless of CPU.

There is such a little variance that having 1080ti spiting out 1070 graphics score means that something is wrong in the system, not that CPU is a bottleneck.

[Nocte]

Have you downloaded the AMD update that provides the Ryzen Balanced power mode? The "normal" balanced power usage is known to make Ryzen park some cores etc.