Question What could be causing my (seemingly) very below average score ?

[124124]

Just ran Port Royal a couple times, keep getting 400~ points under the average.
Temps seem fine, everything seems alright. Not sure why my GPU is apparently underperforming.
https://www.3dmark.com/3dm/137126354
Rig is visible in 3DMark, but here it is anyway

PC Specs
GPU: 4070 VENTUS 3X OC 12GB
CPU: i5-12600KF
MOBO: B760M GAMING X AX DDR4
RAM: Kingston FURY Beast DDR4 3200 MHz 16GB 2x8GB CL16
SSD: WD Black SN850X
PSU: ATX 750W THERMALTAKE SMART BM2
HDD: Seagate barracuda st1000dm010 1tb 3.5" sata3
MOUSE: Corsair Harpoon
KEYBOARD: Tactical Ozone Mini.

 

[Lutfij]

Rig is visible in 3DMark, but here it is anyway
What BIOS version are you on for your motherboard? How are you cooling the processor? Make and model of your case?

PSU: ATX 750W THERMALTAKE SMART BM2
The Smart series of PSU's from Thermaltake aren't exactly a smart investment as they aren't a reliably built unit.

 

[VizzieTheViz]

Isn’t 400 points more of a variance in scoring that’s to be expected with different configurations, ambient temperature, stuff running in the background, having luck with cpu/gpu binning etc?

Is there actually a problem using the system or would you just expect/ like to see a higher number in a benchmark?

 

[124124]

Isn’t 400 points more of a variance in scoring that’s to be expected with different configurations, ambient temperature, stuff running in the background, having luck with cpu/gpu binning etc?

Is there actually a problem using the system or would you just expect/ like to see a higher number in a benchmark?

Normal variance is no more than 3% (according to devs). Just worried something might be wrong with my PC performance wise as I have been seeing some visual glitches in the past and have been getting my event viewer bombarded with livekernel errors.

 

[COLGeek]

Is your motherboard BIOS up to date? Same for all of your major drivers and Windows.

 

[124124]

Rig is visible in 3DMark, but here it is anyway
What BIOS version are you on for your motherboard? How are you cooling the processor? Make and model of your case?

PSU: ATX 750W THERMALTAKE SMART BM2
The Smart series of PSU's from Thermaltake aren't exactly a smart investment as they aren't a reliably built unit.

My BIOS version is the F14 for the Gigabyte B760 GAMING X AX. It's from 2023, you reckon I should update it? I know my PSU is bad. Been thinking about swapping it out. I'm cooling the processor with a Deepcool AK400 Zero Dark. My case is a Deepcool Matrexx 4 3F.

Is your motherboard BIOS up to date? Same for all of your major drivers and Windows.

My windows and major drivers, are, in theory, all up to date. My BIOS isnt. Should I update?

 

[COLGeek]

Yes, update your BIOS and check performance afterward.

 

[124124]

Yes, update your BIOS and check performance afterward.

Just updated my BIOS - performance is still the same.

 

[COLGeek]

Is your GPU installed in the x16 slot nearest the CPU?

 

[Eximo]

2x8 GB of DDR4 3200 is not what most enthusiasts who submit to 3D Mark are likely to be using with 12th gen, so there is that. No mention of CPU cooling? If the CPU is running hot during the benchmark, that would affect the score. Though the clock speed looks reasonable, no temperatures shown.

And of course enthusiasts are likely to have overclocked/undervolted their GPUs and CPUs. Since you can't overclock the CPU, that would put you behind most users of similar hardware.

I run 2x16GB DDR4 4000 with my 12700KF for example. I don't really overclock my GPU since it meets my needs currently, but it is water cooled which makes the clock speeds extremely stable during benchmarks. CPU is water cooled as well, with mild overclock.

 

[124124]

Is your GPU installed in the x16 slot nearest the CPU?

Yes, its in the upper slot of the two that I have. Has a tiny bit of sag if that matters.

2x8 GB of DDR4 3200 is not what most enthusiasts who submit to 3D Mark are likely to be using with 12th gen, so there is that. No mention of CPU cooling? If the CPU is running hot during the benchmark, that would affect the score. Though the clock speed looks reasonable, no temperatures shown.

And of course enthusiasts are likely to have overclocked/undervolted their GPUs and CPUs. Since you can't overclock the CPU, that would put you behind most users of similar hardware.

I run 2x16GB DDR4 4000 with my 12700KF for example. I don't really overclock my GPU since it meets my needs currently, but it is water cooled which makes the clock speeds extremely stable during benchmarks. CPU is water cooled as well, with mild overclock.

I'm cooling my CPU with a Deepcool AK400 Zero Dark. I definitely do need to look into faster RAMs. My CPU temps dont go over 75 during this test.

 

[Eximo]

I'm cooling my CPU with a Deepcool AK400 Zero Dark. I definitely do need to look into faster RAMs. My CPU temps dont go over 75 during this test.

That should be just fine then.

Yeah, 2x16GB DDR4 4000 CL18 is about $80 right now. You can go for more, 4400 tends to work, but the price starts rapidly going up.

 

[124124]

That should be just fine then.

Yeah, 2x16GB DDR4 4000 CL18 is about $80 right now. You can go for more, 4400 tends to work, but the price starts rapidly going up.

You think swapping my RAM up could make the scores better? It's just quite confusing because my GPU didn't underperform about a year or so ago and as far as I know GPUs shouldn't really deteriorate in the sense of performance unless you put them under constant heavy load (which I havent done). Do have in mind Port Royal is a very GPU heavy test

 

[Eximo]

VRAM overages do end up in system memory. Why you see all the hate against 8GB cards today, their performance plummets when the system memory starts getting used.

A few things that can go wrong on a GPU that don't mean it breaks.

The temperature sensor you see in the monitoring isn't the whole picture. It is either an average of several sensors or a particular sensor. Internally the silicon has many more local temp sensors to make sure it doesn't cook itself. Thermal paste applications can degrade. Long term sag can cause the thermal paste to break off due to mechanical stress. All about the heat cycles, not necessarily running the card hard. So, if only a small part of the GPU is overheating, that could result in seeing normal temperatures, but less performance.

GDDR5X and up, if I recall correctly, have partial ECC, so memory errors can be recovered from live. This means interrupting the pipeline to re-do a render. One way to test this would be to lower the memory clocks a bit to see if the GPU performs any better.

As mentioned before, your PSU is not the best. Running something like HW Monitor might reveal a power limit perfcap reason, and if the GPU power limit isn't reached, it means something is wrong.

Keep an eye on the 12V voltage during your test, if it sags a lot, that could be the PSU failing to give the GPU what it wants.

Nvidia has also been putting out 50 series drivers that have had adverse effects on older cards. Might just be that.

 

[124124]

VRAM overages do end up in system memory. Why you see all the hate against 8GB cards today, their performance plummets when the system memory starts getting used.

A few things that can go wrong on a GPU that don't mean it breaks.

The temperature sensor you see in the monitoring isn't the whole picture. It is either an average of several sensors or a particular sensor. Internally the silicon has many more local temp sensors to make sure it doesn't cook itself. Thermal paste applications can degrade. Long term sag can cause the thermal paste to break off due to mechanical stress. All about the heat cycles, not necessarily running the card hard. So, if only a small part of the GPU is overheating, that could result in seeing normal temperatures, but less performance.

GDDR5X and up, if I recall correctly, have partial ECC, so memory errors can be recovered from live. This means interrupting the pipeline to re-do a render. One way to test this would be to lower the memory clocks a bit to see if the GPU performs any better.

As mentioned before, your PSU is not the best. Running something like HW Monitor might reveal a power limit perfcap reason, and if the GPU power limit isn't reached, it means something is wrong.

Keep an eye on the 12V voltage during your test, if it sags a lot, that could be the PSU failing to give the GPU what it wants.

Nvidia has also been putting out 50 series drivers that have had adverse effects on older cards. Might just be that.

Just got done with Port Royal while also checking my sensors with HWiNFO. Everything seemed fine on the voltage and wattage. GPU wattage was 195-200W constantly and voltage was 1-1.10. Junction temperature was about 75-80~ and hotspot was 80-86~. All seems fine here. Will try to lower the memory clocks and retest.

 

[Roland Of Gilead]

Just updated my BIOS - performance is still the same.

Did you clear CMOS after the bios update? If not, you must do this to be sure it's not a bios issue.

 

[124124]

Did you clear CMOS after the bios update? If not, you must do this to be sure it's not a bios issue.

No, I haven't. However, I have upped my core clock by just 200mhz and that was enough to elevate my score enough. Not sure if I should keep it like this or not - because these aren't the stock settings but this is what the GPU should actually be showing in terms of performance.

 

[Roland Of Gilead]

No, I haven't. However, I have upped my core clock by just 200mhz and that was enough to elevate my score enough. Not sure if I should keep it like this or not - because these aren't the stock settings but this is what the GPU should actually be showing in terms of performance.

It would be a good idea anyway. Afterwards you can dial in your OC again, and test.