That benchmark seems kind of weird and sprinty. It doesn’t give a lot of time for steady state performance. I have a Framework 16 with a 7840 and I am seeing single threaded scores the same as yours, around 662, and I am getting multi threaded scores of ~6330 but the temps are a bit odd. Running it back to back I am seeing reported temps anywhere from 84 to 96. Framework’s pretty conservative fan profile is what I suspect is causing the higher temps, though I do not have a F13 to compare it to. When I run the CPU-Z stress test the fans don’t even really kick on until the CPU is in the 90s.
1 Like
This is TDP - max temperature (it’s same for 7840U, btw - but I have ~86C max for FW13).
What I figure out so far - I have only one Core (Core 3 reaching 100C continuously, while, for example, Core 0 is always limited to 86-88C). Looks like some issues with thermal application (not all CPU covered fully?) or defective CPU with Core 3 overheating…
Thanks for your numbers - looks like 84-96C is fine, and there some issues in thermal application for my case.
I have the R7 7840hs and for me its core4 always hitting 100C in AllCore Benchmarks. It has 15C more than the lowest Core. Framework is replacing my Mainboard as it seems thats not normal.
2 Likes
Seems to be the same for me (batch 2). I just open a support case. core#0 is the lower (about 79C° in charge) and core#3 hit 99.7C° very quickly and stay at this temperature. The other cores are between 86/96 C°.
1 Like
Mine will hit 100c under heavy load, but it doesn’t stay there. It drops down significantly under normal operation. Even when gaming (with a dGPU) the CPU is under 100c. From what I’ve read about this CPU, that’s normal. However, if someone sees 100c temperatures sustained, that is a problem and you should open a support ticket.
1 Like
Compare with Framework 13 AMD Ryzen 7 7840U (look at the third column, which is MAX, over ~10 minutes CPU-Z multicore stress test):
And similar for Framework 16 AMD Ryzen 9 7940HS (look at the third column as well as MAX - note Core 3 is 104.3C while Core 0 is 89.8C - ~7 min stress test):
What’s interested - the difference in Multicore score is just ~3% between them…
Even with ~28% CPU load it’s on ~96C…
Mine reached ~100C with CPU load as 30%+ already…
Note: there no even close such scenario for Framework 13 AMD (7840U), which I also had (as mobile workstation)
Your report on FW16 show high frequence and high Current on Core (10x more than on the FW13 report)
Look like you have 1 process that have high load / CPU usage. If this process is pinned on core3 it is normal that it have high temperature.
So look more in this case that a process heat the CPU…
20% on CPU with 100MHz is 10x lower load than 20% on 1000MHz .
That say, it look for me that without dGPU the fan speed is realy low and the idle temp is high on FW16. And it take time to rankup. If I can change fan curve I definitivly push it higer.
What drivers are you using? My Helldivers 2 is fine on my Framework 16. Actually great, averaging around 120.2 fps.
If you download the driver bundle and BIOS from the Framework forums found on this website then you should have the same Driver & Software version that I have in the image (23.40.18.02). Anything beyond that and you probably installed your drivers from the AMD website which I don’t think you’re supposed to do with your Framework.
To Recap
So far, I’m under the impression that you are not supposed to use AMD’s drivers from the AMD website, but instead use the drivers from the Framework forums. I had a lot of issues in all of my games with decreased performance and increased temperatures when using newer AMD drivers that are being recommended by the AMD auto-installer. I think that it’s downloading incorrect drivers because Framework 16’s with the GPU Addon have 2 GPUs and not just one. So you need the Framework specific drivers that recognize BOTH and not just the Drivers for the RX 7700S.
Personally, I’m not using AMD GPU (iGPU / dGPU)… I have eGPU box (Razer Core X Chroma) and have NVidia RTX there. No issues with USB4 connections - for FW13 / FW16.
However, on FW16 (R9 7940HS) I got 100C+ CPU temp with ANY gaming on eGPU, while on FW13 (R7 7840U) with same gaming I got better FPS (1-5% more, but still…) and lower temps (~65-75C).
Example with FW16 (CPU temp 100C):
And same settings FW13 (CPU temps 73C):
Ok, FW13 was in idle mode…
There another numbers for FW13 during stress test (first ~20 seconds… after that temps go slowly up to ~86-88C, like in MAX column)
FW13 temps never go 90+C, while FW16 goes almost instantly to 95C+.
Moreover, I have never seen 25C+ difference between cores on FW13, not sure if it’s OK to have one Core 25C+ compared to others (like on FW16 - it’s always Core3, which reached 100C in ~15-20s)
If you pump 80W into the 7840u in the 13 it’ll do that too, even with liquid metal XD.
There is a concept of “best core” so it would make sense that the same core gets used for very high power single threaded loads.
On stress test FW13 (7840U) utilized ~32W (spike to 52W and down to 32W) and FW16 (7940HS) utilizes ~37W (spike to 56W and down to 37W).
I can’t see any 80W, unfortunately. Also increase in performance (benchmark points) is ~3-5% and after 10 minutes FW16 actually goes down and provide less points (due to throttling, I believe). That’s my concern.
“best core” on FW13 is always 3-5C higher, not 25C+. Is that expected?
That took quite a bit of fuckery to archieve since you tend to get limited by stapm even if cooling can keep up.
the hs may well have much more aggressive single core boosting, how much power do you get during single core cinebench? On the U I only managed ~19W single core so if the hs pumps a bunch more power into a single core 20+c deltas don’t sound unexpected, silicon doesn’t have infinite thermal conductivity so pumping a bunch of w into a few mm2 would make it significantly warmer than the surroundings.
Looks like after ~10 mins the power consumption for FW16 (7940HS) is ~22.5W with ~80C for Core3 (Core0 is ~57C).
Well you are dumping a lot of power in one core if I see this right, how’s the core to core delta running all core?
When i run a cb23 SC Test, its always switching the Load between Cores, to keep the heat low out of the Throttle Area and keep the frequency high.
I am getting a Board Replacement because my Core 4 is always pegged to 100C on Medium to Full Load. And i have a TempDelta of 15C. CB R23 isAbout 14300Points MC (which is 2000 Points short of what it should get) and 1740 Points SC whats right on Spec for the 7840hs.
1 Like
