[SOLVED] GPU memory clock is lower than expected

[TomFiddle]

Recently got a brand new computer with a non-k 11700 and a ASUS TUF Gaming 3070 Ti.

However, I noticed that in full load, my GPU's memory clock is lower than it advertised. It ran at 1156.5 MHz instead of 1188 MHz (default memory clock of 3070 Ti)

In addition, I understand that GDDR6X memory will run hotter than regular GDDR6 memory, will the memory temperature capped at 86 °C when it is in full-load be too hot?

Also, my CPU is running at stable 4.39 GHz when gaming, is it a way to increase the clock speed or should I increase it?

I am running the PC with a 750W gold PSU. Additionally I have one gen4 M.2 SSD and one SATA SSD, 2x 8GB 3200 RAM on an ASUS TUF GAMING Z590 motherboard

 

[dorsai]

How is your case cooling ? The GPU will definitely start to throttle if it gets too hot. DDR6X runs hot anyway so case airflow will be all that much more important when under full load. If you have the option you can try setting a more aggressive fan curve on both your case fans and your GPU fans.

The 11th gen Intel chips are known to be very hot and very power hungry when overclocked...and since you don't have the K version overclocking becomes very dependent on motherboard support. It may be better to focus on keeping the CPU cool since it appears to be turboing up to full speed already.

 

[TomFiddle]

How is your case cooling ? The GPU will definitely start to throttle if it gets too hot. DDR6X runs hot anyway so case airflow will be all that much more important when under full load. If you have the option you can try setting a more aggressive fan curve on both your case fans and your GPU fans.

The 11th gen Intel chips are known to be very hot and very power hungry when overclocked...and since you don't have the K version overclocking becomes very dependent on motherboard support. It may be better to focus on keeping the CPU cool since it appears to be turboing up to full speed already.

Thanks for the reply!

I am using antec df600 flux as the case and FUMA2 as the CPU cooler, In normal circumstances (gaming, casual rendering) with ambient temp at 30 °C, my GPU runs at 60-75 °C in full load, similar to my CPU Temp.

The only thing that concerns me is that the memory clock, I suspect that it is because the auto OC turn down the memory clock to get a better temp and higher core clock, because I look up videos showcasing other 3070 Ti with higher memory clock, they got 90 °C something, so I guess the GDDR6X is just too hot to cool down.

For the 11700, after some researches, I think if it is not in heavy load, it will just keep at 4.4 Ghz, I have seen the same thing happens in the benchmark videos on YouTube.

 

[Cubrix]

Quick question, how is your GPU connected to your Power supply? Are you using a 1 PCI cable with a splitter to your GPU or are you using 2 PCI cables? The reason I am asking is because the PCI slot gives about 75watt and 1 PCI cable from the PSU can only supply 150 watt. That totals to 225 watt. Your GPU requires around 290 watt. So what you can maybe test is to use 2 x 8 pin PCI cables from your PSU instead of 1 cable with 2 x 8 pin splitters.

 

[TomFiddle]

Quick question, how is your GPU connected to your Power supply? Are you using a 1 PCI cable with a splitter to your GPU or are you using 2 PCI cables? The reason I am asking is because the PCI slot gives about 75watt and 1 PCI cable from the PSU can only supply 150 watt. That totals to 225 watt. Your GPU requires around 290 watt. So what you can maybe test is to use 2 x 8 pin PCI cables from your PSU instead of 1 cable with 2 x 8 pin splitters.

I think it was plugged with 2 x 8 pin, my friend built that for me

 

[dotas1]

Quick question, how is your GPU connected to your Power supply? Are you using a 1 PCI cable with a splitter to your GPU or are you using 2 PCI cables? The reason I am asking is because the PCI slot gives about 75watt and 1 PCI cable from the PSU can only supply 150 watt. That totals to 225 watt. Your GPU requires around 290 watt. So what you can maybe test is to use 2 x 8 pin PCI cables from your PSU instead of 1 cable with 2 x 8 pin splitters.

What you are describing is the daisy chain which has two 6+2 pins. All modern PSUs have daisy chain cables and they are working as intended. For a two 6+2 pin daisy chain, it can provide 150w each for a total of 300w.

The only issue with daisy chain cables is heat. That is why some people use two separate cables and not daisy chained. There is no power issue when using a daisy chain.

A GPU must have all available power connectors plugged in, otherwise it won't work or in some OLDER GPUs, will not boost to higher clocks.

 

[Cubrix]

What you are describing is the daisy chain which has two 6+2 pins. All modern PSUs have daisy chain cables and they are working as intended. For a two 6+2 pin daisy chain, it can provide 150w each for a total of 300w.

The only issue with daisy chain cables is heat. That is why some people use two separate cables and not daisy chained. There is no power issue when using a daisy chain.

A GPU must have all available power connectors plugged in, otherwise it won't work or in some OLDER GPUs, will not boost to higher clocks.

Mmm I guess I can be wrong. I just followed what JayzTwoCents said on this video:

View: https://www.youtube.com/watch?v=3puInnaG5ok&ab_channel=JayzTwoCents

@8:50

At least I learned something today, thanks!

 

[spentshells]

1156.5 MHz instead of 1188 MHz
making it run at 1188 won't change anything no additional fps
the memory is fairly hot, you could try a spot fan or another case fan that blows on the card directly

If I were you I wouldn't even bother, odds are some minor change was made when oc'd
you could fight for another lil bit and nothing will change.

 

[dotas1]

Mmm I guess I can be wrong. I just followed what JayzTwoCents said on this video:

View: https://www.youtube.com/watch?v=3puInnaG5ok&ab_channel=JayzTwoCents

@8:50

At least I learned something today, thanks!

It's the safe way to go, yes. It's also what I have in my GPU so I do agree that it's better not to use daisy chain. It's NOT the problem here though. If you listened when Jay changed the cables he said it's an old PSU although platinum efficiency. Not to mention that in the end, the problem was the GPU and not the PSU.