Question 7950X single core doesn't hit 5.7/5.8GHz

[Steelbom]

Hi,

Recently picked myself up a 7950X and Aorus Xtreme X670E w/ 2x32GB Corsair Vengeance 5600MHz RAM.
I'm using a Seasonic 1000W Titanium PSU and a Noctua NH D15 CPU cooler w/ a Fractal Torrent case (2x200mm front fans, 3x120/140mm bottom fans).

I ran a Cinebench single core benchmark and could only hit 5.5GHz and the multi-core I could only hit 5GHz all-core.

The latter maxed out CPU temp at 95 (as expected) but the former was only hitting 65-70c.

Anything I can do without overclocking to help the single core reach 5.6GHz+?

Cheers,
SB

 

[IEC]

These chips (goes for both major vendors) are basically factory overclocked out of the box. Both also implement some form of AVX offset - in AMD's case, it is dynamic based on workload. The more intense the load, the more it scales back to stay within parameters.

5.5GHz ST sounds about right for Cinebench which is an AVX workload
5 GHz sounds about right for MT

Techspot got similar results in their 7950X review

As long as your benchmark scores are similar to what you read in reviews, I wouldn't worry about it.

 

[eek2121]

These chips (goes for both major vendors) are basically factory overclocked out of the box. Both also implement some form of AVX offset - in AMD's case, it is dynamic based on workload. The more intense the load, the more it scales back to stay within parameters.

5.5GHz ST sounds about right for Cinebench which is an AVX workload
5 GHz sounds about right for MT

Techspot got similar results in their 7950X review

As long as your benchmark scores are similar to what you read in reviews, I wouldn't worry about it.

While you are correct, another issue is that the latest AGESA limits boost clocks for 4+ cores to 5.5ghz. This means some workloads and benchmarks may not perform as well. There is a BIOS option on ASUS boards to disable the limit, but I haven’t been able to get it to work.

This matters, because (at least on my machine) if any combination of 4 cores is active at all, you won’t see peak single core clocks.

 

[A///]

asus saves the day, hurrah.

 

[Timmah!]

Perhaps bigger AIO, some 360 or even 420 one, could help to get those additional 100, 200 MHz. Noctua might allow only 5,5 before it reaches 95C.

 

[Steelbom]

These chips (goes for both major vendors) are basically factory overclocked out of the box. Both also implement some form of AVX offset - in AMD's case, it is dynamic based on workload. The more intense the load, the more it scales back to stay within parameters.



5.5GHz ST sounds about right for Cinebench which is an AVX workload

5 GHz sounds about right for MT



Techspot got similar results in their 7950X review



As long as your benchmark scores are similar to what you read in reviews, I wouldn't worry about it.

Ahhh I see. I did get slightly lower scores on the single threaded side of things (1950) and a fair bit lower on the multithreaded side (32000). I've recently changed the power plan from "Power saving" to "Ultimate performance". Not sure how it got onto that.

While you are correct, another issue is that the latest AGESA limits boost clocks for 4+ cores to 5.5ghz. This means some workloads and benchmarks may not perform as well. There is a BIOS option on ASUS boards to disable the limit, but I haven’t been able to get it to work.



This matters, because (at least on my machine) if any combination of 4 cores is active at all, you won’t see peak single core clocks.

Interesting, that would definitely be the case for me since I had a fair few programs open at the time.

I wonder why they made that change?

Perhaps bigger AIO, some 360 or even 420 one, could help to get those additional 100, 200 MHz. Noctua might allow only 5,5 before it reaches 95C.

Yeah could be the case. Although, it's not actually hitting 95c on the single core test (unless maybe one chiplet is?)

 

[alexruiz]

What UEFI version do you have?
Did you install AMD chipset drivers?

 

[Steelbom]

What UEFI version do you have?
Did you install AMD chipset drivers?

Not sure about the first question but I did install the latest AMD chipset drivers.

 

[lopri]

de8auer's 7950X does 5.4 GHz with a custom water setup. 5.45 GHz causes a crash. (around 7:20)

 

[lopri]

At around 10:50 into the video there is a temperature scaling chart. You need a peak temperature of 70C for 5.5 GHz all-core boost. Not sure if it's doable with air or water.

 

[Markfw]

At around 10:50 into the video there is a temperature scaling chart. You need a peak temperature of 70C for 5.5 GHz all-core boost. Not sure if it's doable with air or water.

What I want to know is why is it a big deal to hit over 5.1 even all-core boost ? Mine run at 4.8 ghz with all 32 threads@100% all day, and they just kick butt.

Oh, and that with CO at -25 and 142 watt usage. And still kicks the 5950x right where it hurts, but like 30-50% based on the exact load.

 

[alexruiz]

Not sure about the first question but I did install the latest AMD chipset drivers.

UEFI version meant BIOS version.
Have you reflashed a new one, or no?

 

[lopri]

What I want to know is why is it a big deal to hit over 5.1 even all-core boost ? Mine run at 4.8 ghz with all 32 threads@100% all day, and they just kick butt.

Oh, and that with CO at -25 and 142 watt usage. And still kicks the 5950x right where it hurts, but like 30-50% based on the exact load.

You need to calm down. I did not say it's a big deal nor it's any kind of deal. It's a piece of information put out there by someone who thinks it's worth spending time making a video on. I just relayed it here because it seems related to the thread at hand. You sound as though someone insulted your CPU.

 

[Timmah!]

What I want to know is why is it a big deal to hit over 5.1 even all-core boost ? Mine run at 4.8 ghz with all 32 threads@100% all day, and they just kick butt.

Oh, and that with CO at -25 and 142 watt usage. And still kicks the 5950x right where it hurts, but like 30-50% based on the exact load.

Aesthetics mostly, 5,1 looks sexier than 4,8, even if the performance difference is negligible (at least compared to power consumption difference). There were plenty CPUs in the past that could clock at 4,8GHz all-core, but 5GHz+ is kinda new and still magical.
Other than that, while the clock numbers between reviews varied, most of them afaik claimed the CPU works at 5.7 single core and about 5,1 all-core on average. So one kinda wants too see those numbers on his own particular chip and it can be kinda disappointing if thats not the case. I know i would be somewhat - even if from performance/usage point there is very little difference.

 

[Steelbom]

de8auer's 7950X does 5.4 GHz with a custom water setup. 5.45 GHz causes a crash. (around 7:20)

At around 10:50 into the video there is a temperature scaling chart. You need a peak temperature of 70C for 5.5 GHz all-core boost. Not sure if it's doable with air or water.

Ahhh that's pretty nice. Yeah, I haven't ever tried a water setup... only air. Pretty happy with it though. Performance is great. Just wish I could get my hands on DRR5 128/256GB @ 4800MHz+

What I want to know is why is it a big deal to hit over 5.1 even all-core boost ? Mine run at 4.8 ghz with all 32 threads@100% all day, and they just kick butt.

Oh, and that with CO at -25 and 142 watt usage. And still kicks the 5950x right where it hurts, but like 30-50% based on the exact load.

Probs not a big deal but yeah I'd seen it mentioned that it can do 5.1. Thought I may have something misconfigured. What's "CO"?

UEFI version meant BIOS version.
Have you reflashed a new one, or no?

Ah gotcha. Yep, I did update to the latest BIOS available.

Aesthetics mostly, 5,1 looks sexier than 4,8, even if the performance difference is negligible (at least compared to power consumption difference). There were plenty CPUs in the past that could clock at 4,8GHz all-core, but 5GHz+ is kinda new and still magical.
Other than that, while the clock numbers between reviews varied, most of them afaik claimed the CPU works at 5.7 single core and about 5,1 all-core on average. So one kinda wants too see those numbers on his own particular chip and it can be kinda disappointing if thats not the case. I know i would be somewhat - even if from performance/usage point there is very little difference.

It is nice hitting 5GHz+. Like you mentioned I have seen a bunch of reviews say that it can hit 5.7 or even 5.8 single core and 5.1GHz all-core. I thought I may have something misconfigured if temps (single-core) are under 95 but I'm getting limited at 5.5GHz single-core. Also my multithreaded cinebench was about 6k points lower than in reviews, but I did have a bunch of programs running so that probably wasn't a great benchmark.

>>>>

On a similar note, how does the behaviour change if you turn on PBO? Does it try to push the clocks even higher? It seems like they are kind of designed to do that these days out of the box.
I also noticed the Curve Optimizer which sounded pretty cool but didn't understand what it does exactly.

 

[Markfw]

Ahhh that's pretty nice. Yeah, I haven't ever tried a water setup... only air. Pretty happy with it though. Performance is great. Just wish I could get my hands on DRR5 128/256GB @ 4800MHz+

Probs not a big deal but yeah I'd seen it mentioned that it can do 5.1. Thought I may have something misconfigured. What's "CO"?

Ah gotcha. Yep, I did update to the latest BIOS available.

It is nice hitting 5GHz+. Like you mentioned I have seen a bunch of reviews say that it can hit 5.7 or even 5.8 single core and 5.1GHz all-core. I thought I may have something misconfigured if temps (single-core) are under 95 but I'm getting limited at 5.5GHz single-core. Also my multithreaded cinebench was about 6k points lower than in reviews, but I did have a bunch of programs running so that probably wasn't a great benchmark.

>>>>

On a similar note, how does the behaviour change if you turn on PBO? Does it try to push the clocks even higher? It seems like they are kind of designed to do that these days out of the box.
I also noticed the Curve Optimizer which sounded pretty cool but didn't understand what it does exactly.

CO is an option of PBO. Instead of overclocking, it sets the "curve" in my case lowed. So the max power goes from 230 (or more) to about 142 watts, but the performance is almost the same, but the clocks are lower.

 

[Steelbom]

CO is an option of PBO. Instead of overclocking, it sets the "curve" in my case lowed. So the max power goes from 230 (or more) to about 142 watts, but the performance is almost the same, but the clocks are lower.

Ahhh gotcha, thanks. That's pretty impressive power savings!