Low 7800x3d speed in cinebench r23

wannabe.mixer

Member
CORPORAL
Joined
Aug 10, 2015
Messages
36
My speed is at 4.550 ghz during a Cinebench run with pbo enabled. I sent the PC back to the store that built it, and they said they tried putting in a new CPU and putting the CPU on another motherboard with the same issue, so they called it normal. Are they wrong? I also tried the curve optimizer and got 4.6 MHz then. The motherboard is a b650 tomahawk max Temperature 78c

Here is what the store said:

Speed on the processor momentarily peaks at around 5 GHz across all cores upon arrival. All-core and stress tests usually settle around 4.55 GHz. Tested with another processor, and the processor was tested on another motherboard with equivalent results in terms of frequency. Different frequencies on all-core with different bios: the newer beta version gave high clock speeds but a low Cinebenchscore of 15500 points; the older beta bios from last spring peaked at around 17450 with Expo enabled. The current BIOS 1.70 gave 17280 points in the last test and was stable at start-up with us. Longer startup/boot, but nothing beyond what we normally see at 6000 MHz and AMD, tight settings, and memory training are done at every startup. The processor cooler was adjusted to "offset mode" for better cooling. No impact in terms of frequency or points was noticed from this.



If i can get someone from msi that can say that this is not normal i can use that as proof to the store
 
I would try older BIOS versions first. You are not terribly far off from the norm with your score. In several reviews like here, it hit slightly above 18000 points in CB23 Multi. So you're around 4% off. Then we got some like here where you'd be bang on the money. Then we even got one here where you're performing better!

And of course, most of those reviews were done on the launch BIOS versions. The BIOS versions used for the reviews, that just added support for a new CPU, are usually the best-performing versions. Later they have to implement fixes and focus the optimizations more on the stability and reliability, so it will tend to perform a bit worse in later versions, but it will also tend to run with less problems.

In trying some older BIOS versions (any from 7D75v16 onwards), you are making sure it's actually the CPU that doesn't want to clock higher, rather than the BIOS messing with it. Because the BIOS has more or less full control on how the CPU behaves.
I did try the bios that was after the vsoc fix and it had the same issue you think i should send it back?
 
It's up to you, do you want to put your foot down on this over what's at maximum a 4% deviation from the norm in performance, but is in the same ballpark? I mean, you are seeing the impact of the below-average boost frequency right there in the scores, it results in 4% lower scores than a bunch of reviews. But other reviews got a lower score as well, one even lower than yours. So it might just be normal variation still.

Have you watched the video from here? https://forum-en.msi.com/index.php?threads/low-7800x3d-speed-in-cinebench-r23.392672/#post-2235722
There's normal differences between each and every CPU of a specific model. That's the silicon lottery. So as far as clocks and performance, especially with AMD, it seems to be enough that you get "around" the same numbers, not necessarily exactly. Personally, i would investigate further. Replacing the CPU is such a big hassle that i wouldn't do it over small things.
 
I'd check the BIOS for anything that could help, check some other benchmarks, compare CPU voltages, power consumption and temperature numbers with what they got in the reviews, and so on and so forth. Basically have a better understanding of where my particular CPU stands amongst other ones of this model. If it's all within less than 5% - and for the power consumption i'm actually expecting more than 5% less (which would be good) - then i don't think i would make that big of a deal out of it.

Now that i've just randomly googled for people with the same CPU behaviour, it was quite easy to find them, like here, here, here, here, the list goes on. Something i also just remembered, simply having HWinfo Sensors open in the background can already cost you around 300 points. So anyway, it looks like this can be more or less a normal occurrence with this CPU, and you're far from being the only one. So then i wouldn't try to have it replaced, chances are it won't change too much.
 
Why don‘t you post the hwinfo screenshot, so we can check for limitations.
Your cpu should allways try to hit 80°C in c23, no matter what your cooling looks like. It‘s more about how good it can boost while hitting it.
 
Yes, that's a fair point. If the cooling is insufficient, the CPU will notice it, and may just clock lower because it sees that there's no "temperature budget" left to exploit. What's your CPU cooler model?
 
Yes, that's a fair point. If the cooling is insufficient, the CPU will notice it, and may just clock lower because it sees that there's no "temperature budget" left to exploit. What's your CPU cooler model?
i will do a cinebenchr23 run now and send you all the pictures but its an arctic freezer 360
 
FORMSI.png



it runs at 4.6 during r23 with my curve optimizer settings the spikes u see to 5.0 is when i start hwinfo64 before i start the benchmark
 
Looks pretty good, your system temps are very good even, you must have good airflow through the system, but i guess the CPU is quite efficient as well. And with a 360mm AIO, we can assume it has nothing to do with a lack of CPU cooling.

What we haven't even tackled yet, it might have something do with Windows. If you have a spare SSD or something available, try a fresh Windows installation on there, with the latest AMD chipset drivers and AMD GPU drivers, nothing much changed otherwise.
 
Looks pretty good, your system temps are very good even, you must have good airflow through the system, but i guess the CPU is quite efficient as well. And with a 360mm AIO, we can assume it has nothing to do with a lack of CPU cooling.

What we haven't even tackled yet, it might have something do with Windows. If you have a spare SSD or something available, try a fresh Windows installation on there, with the latest AMD chipset drivers and AMD GPU drivers, nothing much changed otherwise.
its a fresh install with the latest chipset and gpu drivers
 
Did you mount the Artic Freezer with the AM5 offset?
You can also check the behavior of the frequencies, while running cinebench. Like, do they hit 4800 Mhz once at 80°C and getting pushed back to 4775 Mhz and deeper, or can the cpu hold the full boost most of the time?
And last is to check for Clock Stretching, thats why I was asking for a screenshot near the end of the benchmark.

I got the same components, i would guess even the same ram. With CO allcores -24 I‘m running 18250 Points on Cinebench R23. Just to give you one more Data.
 
Did you mount the Artic Freezer with the AM5 offset?
You can also check the behavior of the frequencies, while running cinebench. Like, do they hit 4800 Mhz once at 80°C and getting pushed back to 4775 Mhz and deeper, or can the cpu hold the full boost most of the time?
And last is to check for Clock Stretching, thats why I was asking for a screenshot near the end of the benchmark.

I got the same components, i would guess even the same ram. With CO allcores -24 I‘m running 18250 Points on Cinebench R23. Just to give you one more Data.
The store mounted it in offset mode after I asked them to do so; they stay at 4.6 all the time. i will send another screenshot what do you want a screenshot of?
 
Ok then it's probably not that. Oh well. Like i said, you can find a number of people that have a similar phenomenon with their 7800X3D. So this seems to be a CPU model where there is quite some variation in the behaviour between the specimen, similar to the video i linked before. So if you are completely unsatisfied with this, you can play the "silicon lottery" and demand a new CPU. That might then boost higher, with any luck. But as i said, it's quite some hassle for a couple percent difference in the end, so it's all up to you at this point.
 
Is there thermal limit set in pbo to 80c ?
Your exact 80.0 max temperature on CPU package looks like it hitting thermal limit wall.
 
The store mounted it in offset mode after I asked them to do so; they stay at 4.6 all the time. i will send another screenshot what do you want a screenshot of?
Ah ok.
Because my system behaves slightly different.
The CPU hits 4875Mhz, then hits the 80°C and from then on it moves between 78°C to 80,5°C and 4850 to 4875Mhz.
They might be some potential in the cooling.
 
I would say yes, you slightly lost the silicon lottery, you can play it again if you demand the CPU to be replaced. There's no guarantees for anything though. Plus the store has to play ball. Even AMD don't guarantee any specific boost numbers. They guarantee a base clock, and a maximum boost clock, both of which you are ultimately hitting.

amd.png
 
Hmm no. I could imagine you could get similar Points like mine, but it depends on ambient temperature and maybe the thermal paste and how the fans are working. But I would say that your Artic freezer is powered to 100% at 70°C. Even 85% should do it.

See. It‘s not easy to push the 7800x3d far above 80°C even tjmax is 89°, because at 80°C the cpu is starting to adjust the clock speed a little. And then this happens what you see. You are staying on 80°C on the full load, but the clocks are getting down to keep the cpu cool.
 
Hmm no. I could imagine you could get similar Points like mine, but it depends on ambient temperature and maybe the thermal paste and how the fans are working. But I would say that your Artic freezer is powered to 100% at 70°C. Even 85% should do it.

See. It‘s not easy to push the 7800x3d far above 80°C even tjmax is 89°, because at 80°C the cpu is starting to adjust the clock speed a little. And then this happens what you see. You are staying on 80°C on the full load, but the clocks are getting down to keep the cpu cool.
when im watching the test even when its at 75C or 78C its still at 4.6
 
Back
Top