HELP! R9 390X unable to sustain full/stable utilisation.

[MMSIND]

This is my first ever build and it's making me really nervous. Please help me out.

 

While gaming, the GPU usage is all over the place. There was significant stuttering occuring occasionally.
Photo below.

 

Even in stress test with and without CPU stressed (AIDA64) GPU is unable to sustain 100% utilisation.
Photo below.

 

My temps after stressing both GPU and CPU at the same time are, GPU-76C , CPU- 75C
Neither GPU nor CPU is overclocked. Precision core boost on CPU is disabled so it's running at 3500 Mhz without XFR.
CPU is able to sustain 100% utilisation during stress test with and without GPU stressed. There are no signs of overheating of either component (as far as I know).

 

Full specs
Processor- Ryzen 1500X
GPU-Saphire R9 390X
RAM-Kingston Fury 8 GB DDR4 (2400 Mhz)
SSD-Sandisk 240 GB (sorry can't remember exact model)
HDD-Seagate Barracuda 1 TB
Motherboard-Asus Prime B350M-E
PSU-Antec VP600P (600 watts)

Any logical advice is appreciated. Thanks in advance.
PS- I've used several monitoring softwares to measure temps and usage so all of my readings are correct and I have freshly installed Windows 10 yesterday and my graphic drivers are updated.

[ScratchCat]

Non 100% utilization does not mean thermal throtteling, it means that your GPU was getting bored at those points e.g. if the test was paused for a second. That is completely normal. It would simply mean that there is a bottleneck elsewhere or that point in time was not particularly intensive.

 

In gaming that is completely normal, these could be points where the CPU is limiting the framerate (less likely) or you have v-sync on and therefore do not need to full utilize the GPU.

 

[MMSIND]

7 minutes ago, ScratchCat said:

Non 100% utilization does not mean thermal throtteling, it means that your GPU was getting bored at those points e.g. if the test was paused for a second. That is completely normal. It would simply mean that there is a bottleneck elsewhere or that point in time was not particularly intensive.

 

In gaming that is completely normal, these could be points where the CPU is limiting the framerate (less likely) or you have v-sync on and therefore do not need to full utilize the GPU.

 

I had the Vsync off and there was significant stuttering while gaming (GTA V), also I think during stress test GPU utilisation should be 100% like that of CPU. Test was not paused at any moment of low/ erratic usage.

[ScratchCat]

Just now, MMSIND said:

I had the Vsync off and there was significant stuttering while gaming (GTA V), also I think during stress test GPU utilisation should be 100% like that of CPU.

If the CPU was at 100% utilization at that point then the GPU would be limited by the speed of the CPU. Check what background processes you are running, they may put additional strain on the CPU.

 

The GPU must wait for the CPU to process that frame before it can begin rendering. If the CPU cannot keep up the GPU must wait around ,therefore lowering the utilization, for the CPU to pass the details it requires in order to render the frame.

 

In the stress test the CPU must also provide the GPU with data, however that is almost never the limiting factor. Most likely another item needed to be rendered such as a OS UI element or the GPU stress test was temporarily paused for some reason.

 

Once again, it is perfectly normal. Anything which isn't a specific GPU benchmark may have this problem. Try the superposition benchmark and view utilization.

[MMSIND]

Just now, ScratchCat said:

If the CPU was at 100% utilization at that point then the GPU would be limited by the speed of the CPU. Check what background processes you are running, they may put additional strain on the CPU.

 

The GPU must wait for the CPU to process that frame before it can begin rendering. If the CPU cannot keep up the GPU must wait around ,therefore lowering the utilization, for the CPU to pass the details it requires in order to render the frame.

 

In the stress test the CPU must also provide the GPU with data, however that is almost never the limiting factor. Most likely another item needed to be rendered such as a OS UI element or the GPU stress test was temporarily paused for some reason.

 

Once again, it is perfectly normal. Anything which isn't a specific GPU benchmark may have this problem. Try the superposition benchmark and view utilization.

The CPU never went over 60% utilisation during gaming (and it's 1500x) rules out CPU bottleneck. I freshly installed Windows 10 yesterday so that rules out background processes (also Internet was not connected). Please suggest some good GPU stress test. Also the stutters while gaming suggests there ia something wrong with GPU utilisation.

[ScratchCat]

7 minutes ago, MMSIND said:

The CPU never went over 60% utilisation during gaming (and it's 1500x) rules out CPU bottleneck. I freshly installed Windows 10 yesterday so that rules out background processes (also Internet was not connected). Please suggest some good GPU stress test. Also the stutters while gaming suggests there ia something wrong with GPU utilisation.

The CPU need not reach 100% utilisation in order to bottleneck the CPU, if the game can only utilise 2 cores on a 4 core CPU then the maximum CPU load would be 50%. Try reduce your resolution to 720p and test again, if you still have stuttering then you are CPU bottlenecked.

 

Superposition is a good GPU benchmark to test if you really have a problem, it is GPU limited in 720p already.

I would also recommend making sure your drivers are up to date and perform a malware scan.

[MMSIND]

My drivers are updated and there is no malware. I have 1 GB data per day so I can't download Superposition today. Will Furmark be good enough? And I really don't think that 1500X will bottleneck R9 390X in GTA V. Though I'll change resolution and let you know.

[ScratchCat]

2 minutes ago, MMSIND said:

My drivers are updated and there is no malware. I have 1 GB data per day so I can't download Superposition today. Will Furmark be good enough? And I really don't think that 1500X will bottleneck R9 390X in GTA V. Though I'll change resolution and let you know.

Furmark would be fine, however you may run into power limitations. A r9 390X has a rated TDP of around 350W , if power consumption exceeds this the card will throttle either through lowering clock speeds or inducing clock modulation where empty cycles are taken to reduce power consumption. Watch your power consumption with a program like Hwinfo or MSI afterburner and see if the utilisation drops occur at the same time as the power spikes.

[MMSIND]

2 minutes ago, ScratchCat said:

Furmark would be fine, however you may run into power limitations. A r9 390X has a rated TDP of around 350W , if power consumption exceeds this the card will throttle either through lowering clock speeds or inducing clock modulation where empty cycles are taken to reduce power consumption. Watch your power consumption with a program like Hwinfo or MSI afterburner and see if the utilisation drops occur at the same time as the power spikes.

MSI Afterburner is showing that my card is running at 0 volts, I'll use HWinfo for this. And Antec really promises 600 Watts with this PSU so I don't think I'll run into Power limiting.

[ScratchCat]

2 minutes ago, MMSIND said:

MSI Afterburner is showing that my card is running at 0 volts, I'll use HWinfo for this. And Antec really promises 600 Watts with this PSU so I don't think I'll run into Power limiting.

I meant that the GPU will limit it's power consumption to stay within spec. That is what the power limit in radeon settings is for

[Techo238]

1. I'd suggest reseating the card as it may have a stable enough contact to boot your pc but not enough to give good results in games/stress tests.

2. try a different PCI slot. You are blaming the GPU at the moment but it could be an issue with the slot on your Mobo 

[ScratchCat]

For reference, this is my utilisation playing Enderal - a Skyrim mod. The game still stutters now and again ,CPU usage is around 50%.

 

Additionally observe your VRAM usage, if you are running very memory intensive settings your graphics card may run out of memory and have to swap to main memory over PCI-E. This will be noticeable as large stutters as the GPU has to wait for the information to arrive/ be sent and therefore could drop in utilisation.

[MMSIND]

@ScratchCat Please tell me that Furmark is also not a very reliable software. Even in Furmark stress test my GPU usage was occasionally dropping below 50%.  R9 390X have 8 GB DDR5 VRAM and I'm also not using any memory intensive settings. I still believe that in GPU benchmarks the utilisation should be 100% all the time which I can't seem to achieve.

[MMSIND]

2 hours ago, Techo238 said:

1. I'd suggest reseating the card as it may have a stable enough contact to boot your pc but not enough to give good results in games/stress tests.

2. try a different PCI slot. You are blaming the GPU at the moment but it could be an issue with the slot on your Mobo 

I only have one PCI-E slot for graphic card and it's pretty snug in there (I think)

[Techo238]

33 minutes ago, MMSIND said:

I only have one PCI-E slot for graphic card and it's pretty snug in there (I think)

well it was worth a shot, it was just cause my gpu when i first installed it in my rig didnt work cause it didnt seat right. it was down to the little latch that locks the card in place not being fully open which meant the card didnt seat.

 

maybe also try an older gpu in the slot or some different PCI device to see if it works normally

[ScratchCat]

2 hours ago, MMSIND said:

@ScratchCat Please tell me that Furmark is also not a very reliable software. Even in Furmark stress test my GPU usage was occasionally dropping below 50%.  R9 390X have 8 GB DDR5 VRAM and I'm also not using any memory intensive settings. I still believe that in GPU benchmarks the utilisation should be 100% all the time which I can't seem to achieve.

That is indeed strange. If you are worried contact your manufacturer and RMA the card, that will be less of a hassle.

What sort of framerate are you achieving in furmark and what settings are you using?