Core Performance Boost Contributes 14% to Ryzen 5 7600X Cinebench R23 Score

[btarunr]

AMD Ryzen 5 7600X "Zen 4" 6-core/12-thread processor is shaping up to be a speed-demon for purely gaming builds, with the company claiming higher gaming performance than Intel current flagship Core i9-12900K. A combination of high clock speeds (4.70 GHz nominal, 5.30 GHz max boost), high power limits from 105 W TDP (130 W limit), the "Zen 4" IPC, and the fact that all that power headroom is available to just 6 cores, means that the chip is able to sustain boost frequencies better. But what when Core Performance Boost (CPB) is disabled? VideoCardz scored screenshots of a Cinebench R23 run to answer just that.

With CPB disabled (in the motherboard BIOS), the Ryzen 5 7600X scores 1681 points in the single-threaded test, and 13003 points in the multi-threaded one. With CPB enabled (which is the default setting), the 7600X bags 1920 points single-threaded, and 14767 points multi-threaded, which is a 14% performance increase just from the processor's boosting algo. Disabling CPB is generally seen as a silver-bullet against high temperatures for AMD processors, and even here, we see the chip running under 60°C, and pulling 60.2 W peak, as measured by HWinfo; whereas with CPB enabled, the chip can run as hot as 92.1°C, pulling up to 110 W, pushing clock speeds up to 4.45 GHz.



View at TechPowerUp Main Site | Source

 

[Daven]

Am I missing something? AMD claims Zen 4 had a 9% IPC increase over Zen 3 in 1T Cinebench. But this benchmark run only shows a 14% increase with a 15% increase in clock speed.

 

[AusWolf]

I wonder what cooler they used. I don't like that 92.1 °C with only 110 W one bit. This is why I hated the R5 3600 that I gave away after a week or so. Either the IHS design hasn't improved despite the switch to LGA, or chiplets will always run hot.

 

[ZetZet]

I wonder what cooler they used. I don't like that 92.1 °C with only 110 W one bit. This is why I hated the R5 3600 that I gave away after a week or so. Either the IHS design hasn't improved despite the switch to LGA, or chiplets will always run hot.

Can you give one reason why you care about what Celsius your CPU runs at?

 

[AusWolf]

Can you give one reason why you care about what Celsius your CPU runs at?

I like small form factor builds. If a CPU runs hot on a test bench, or in a tower case with adequate airflow, then it will definitely throttle in a SFF system.

 

[ZetZet]

I like small form factor builds. If a CPU runs hot on a test bench, or in a tower case with adequate airflow, then it will definitely throttle in a SFF system.

Celsius is not related to the amount of heat. If a CPU runs at 60c with 110W and 90c 110W the rest of your system only sees 110W. It's the same thing.

 

[Xeanoa]

Am I the only one that's more impressed by the 60 W score than the 110 W score?
10 W per core, 2166 points per 10 W

 

[ZoneDymo]

Celsius is not related to the amount of heat. If a CPU runs at 60c with 110W and 90c 110W the rest of your system only sees 110W. It's the same thing.

im sorry what? you think thermal throtteling is connected to watt usage instead of..ya know...temperature?

 

[AusWolf]

Celsius is not related to the amount of heat. If a CPU runs at 60c with 110W and 90c 110W the rest of your system only sees 110W. It's the same thing.

I'm not worried about the rest of the system. I'm worried about the CPU itself. Smaller case, smaller cooler and less airflow all contribute to it running hotter. If it runs at 92 C on a test bench, it will die in a tiny box.

 

[ZetZet]

im sorry what? you think thermal throtteling is connected to watt usage instead of..ya know...temperature?

If you're able to cool 110W then it won't throttle. Like I said temperature is unrelated. If the person thinks 110W is unacceptable for CPUs when we are heading towards 200W then sure, but again nothing to do with the temperature.

I'm not worried about the rest of the system. I'm worried about the CPU itself. Smaller case, smaller cooler and less airflow all contribute to it running hotter. If it runs at 92 C on a test bench, it will die in a tiny box.

So the problem is cooling 110W. What is the alternative? Intel CPUs use even more power. You could undervolt this CPU and it would use 90W or less. R5 3600 was a 65W TDP and that somehow was a problem for you too.

 

[ARF]

Can you give one reason why you care about what Celsius your CPU runs at?

Because the CPU dies when the Celsius is more

""""you can easily identify by seeing your behaviour of computer whether your CPU is heating or not

1. You are CPU fan rotating very fastly that can create louder noise than usual.

2. Your computer process becomes slow.

3. Some hanging like problem

4. Automatically shut down (this is the most common one when CPU get overheated).

Sometime the CPU get burn if it will overheated.

Caution!!!!

If this kind of problem occur chances that you have to contact computer technician. Before you fixing that kind of problem. Because overheating of cpus can burn your cabinet with motherboard.""""

What happens when the CPU temperature is too high? - Quora

 

[AusWolf]

So the problem is cooling 110W. What is the alternative? Intel CPUs use even more power. You could undervolt this CPU and it would use 90W or less.

Funny enough, I've found Intel CPUs to be much easier to cool when they're set up to use the same power. If you want to know more, read the first post in my build log (link in my signature).

TLDR: My R5 3600 got to throttling temperature at around 80 W while my Core i7 11700 was happy at 120-130 W with the same cooler, in the same chassis. Power consumption and heat dissipation are two entirely different things. And again: I'm not worried about power consumption here. I'm worried about heat dissipation. Why is this so hard to understand?

 

[Pepamami]

I'm not worried about the rest of the system. I'm worried about the CPU itself. Smaller case, smaller cooler and less airflow all contribute to it running hotter. If it runs at 92 C on a test bench, it will die in a tiny box.

It wont, it will drop performance to stay at 95C, maybe its gonna be 90W or 100W, depending.

If u have a tiny/small system, just limit cpu power to 90W, it wont effect single core performance, only cut some multicore performance.

 

[AusWolf]

It wont, it will drop performance to stay at 95C, maybe its gonna be 90W or 100W, depending.

Exactly. I wouldn't like to run my CPU at its max. operating temperature, and even drop performance doing so. This is why my experience with the R5 3600 was such a huge disappointment.

 

[Dirt Chip]

So the problem is cooling 110W. What is the alternative? Intel CPUs use even more power. You could undervolt this CPU and it would use 90W or less. R5 3600 was a 65W TDP and that somehow was a problem for you too.

You need to factor the CPU core topology and size that can drasticly impcat your cooling performance.
ZEN4 have a very samll size core area (a thing AMD is very proud of) so it might be harder to cool.
The same W concentrated to a smaller area so a greater heat is generated.
Dealing with the same 110W (or any other W value) can differ greatly, no matter the cooling solution (as long it is the same)

 

[Valantar]

Am I the only one that's more impressed by the 60 W score than the 110 W score?
10 W per core, 2166 points per 10 W

Same here. And, remember, that's at 4.7GHz! That's almost where Zen3 peaked! Also, that power number includes uncore/IF power, so per-core power is less than 10W, whereas Zen3 at those clocks is more like 17-18W/core. If that is accurate, that is an astounding improvement. Comparisons with RPL notwithstanding, this looks like Zen4 will be a SFF/mobile beast - though I really, really hope we see more cache on the mobile chips this time around. Just don't necessarily expect these chips to be easy to cool at full power, due to their 2x density over RPL.

 

[The King]

pulling up to 110 W, pushing clock speeds up to 4.45 GHz.

5.45Ghz? @btarunr

 

[AusWolf]

this looks like Zen4 will be a SFF/mobile beast

I have my doubts because...

Just don't necessarily expect these chips to be easy to cool at full power, due to their 2x density over RPL.

This. Heat dissipation is crucial in a SFF build. That 92 °C at full power (supposedly on a test bench) looks menacing.

 

[phanbuey]

They're going to struggle against Raptor Lake in the mid range -- if Cinebench is indicative of overall performance (probably isn't).

ALS-S @ 5.2 6/4 ecore - the 13600K should easily stomp on this score - if it's priced around the 7600x it's going to be a huge difference (in this benchmark).

92C is bush league compared what ADL-S hits in cinebench lol.

 

[Valantar]

I have my doubts because...

This. Heat dissipation is crucial in a SFF build. That 92 °C at full power (supposedly on a test bench) looks menacing.

True, but in that class of SFF undervolting and light underclocking are par for the course, and this shows some amazing clock scaling compared to Zen3. Zen's density is indeed a problem point for cooling, but even an L9A should handle those 60W in a case with some access to cool air, and maybe even more.

 

[Pepamami]

Exactly. I wouldn't like to run my CPU at its max. operating temperature, and even drop performance doing so. This is why my experience with the R5 3600 was such a huge disappointment.

ofc if u charge same power to smaller size it will heat more, just use normal power levels, or use a cooler with bigger "inertia"
or go for two CCX cpu, where power will spread between two CCX, if u want 120W so much.

 

[Punkenjoy]

So few things here:

1. You dissipate heat (energy). The number of watt the CPU use is the amount of watt you have to remove from your cpu and your case. No matter what is the running temperature of your CPU, if one use 50 watt and the other 250 w, you should output 5 time more heat in the second case.

After that, depending on your cooling solution, on how the cpu is build etc, how the case is ventilated, what is the ambiant temperature, the fan curve of the cooler, the type of cooler, it's quite possible that the 250 watt end up operating at a lower temperature than the 50 watt cpu. But it will still output 5 time more heat in your case and in your room. If you want to build a SSF, it's the output heat that matter, not as much as temperature.

But having super high temperature at low wattage is not that great either. It add stress to component each thermal cycles, and yes you might boost slightly lower. That lead to my second point.


2. There are 2 types of CPU Throttling.
- The CPU reduce it's operating frequency bellow it's base for few moment to cool itself.
- The CPU generate empty cycle (0) but keep it's frequency. In this case clock might remain high but performance will suffer. (But generally less drastically than by reducing clock under base frequency.

If a CPU have a base of 3.4 GHz and just boost to 3.8 instead of 4.2 GHz, it's not throttling. It's just not boosting as high because of thermal. The boosting in a CPU is there to extract as much performance as possible from it's environment.

In the end, the results is the same as throttling under the base frequency, but you still get way more performance and above what is the base guarantee. So you are in spec.

 

[AnotherReader]

Am I missing something? AMD claims Zen 4 had a 9% IPC increase over Zen 3 in 1T Cinebench. But this benchmark run only shows a 14% increase with a 15% increase in clock speed.

The IPC increase is not uniform, but depends upon the application. AMD's own slides note a 1% increase for Cinebench R23. It isn't even a surprise as AMD hasn't mentioned any improvements to the backend. As it isn't limited by the frontend or the cache, Cinebench won't benefit from the improvements that Zen 4 has compared to an equivalently clocked Zen 3.

 

[phanbuey]

True, but in that class of SFF undervolting and light underclocking are par for the course, and this shows some amazing clock scaling compared to Zen3. Zen's density is indeed a problem point for cooling, but even an L9A should handle those 60W in a case with some access to cool air, and maybe even more.

I run a 180-200W chip in an SFF build - many modern SFF cases (meshilicious, sliger cases, nr200P, etc) can dissipate this heat + 350W/400W gfx cards without any issues. The main problem here isn't the heat generated (at all 110W is very reasonable) -- it's the dissipation capability. If it's running at 92C at 110W that means it would absolutely melt at 150W never-mind the 241W/288W that intel chips suck down.

They might be thermally limited on performance with this node if they're limited to 100W.

 

[Oberon]

Am I missing something? AMD claims Zen 4 had a 9% IPC increase over Zen 3 in 1T Cinebench. But this benchmark run only shows a 14% increase with a 15% increase in clock speed.

That's 14% over the CPB-off behavior, not over a 5600X. A stock 5600X (4.6 GHz max boost) scores 1540 points in the same test vs 1681 @ 4.7 GHz for the 7600X, which puts the PPC increase at right around the quoted 9%. Actual ST performance increase (not iso-clock) comes in at right around 25% vs the 5600X.