Ryzen 7 7840HS is not running at full speed

[inthemirror]

Hi!

I've got an issue with this laptop. Its CPU, Ryzen 7 7840HS (edit: 7940HS is also affected) never reaches its full advertised boost frequency of 5.1GHz and it's instead limited to 4.5GHz. I've verified the issue under Windows 10 LTSC and Fedora Linux. It can be seen in Task Manager, HWiNFO64 or CPU-Z.

I've already created a support case with HP but I wonder if it's a known widespread issue and there is a workaround.

I've not found any performance related options in BIOS - I'm running the latest available version for my laptop.

Please advise.

[banhien]

@inthemirror 

CPU can run at its peak for few nano seconds, it won't stay there for long.

Regards.

[inthemirror]

This is simply not true.

The CPU can run at boost speeds for as long as possible if it's within its thermal/temperature constraints.

[TzortzisG]

Hi inthemirror.

Nicely put. Thermal/temperature constraints, but you're forgetting one other thing. The power constraints. Also, these constraints aren't the ones set by the CPU designers but instead, are enforced by the laptop manufacturers. The CPU designers give the MAX specs, and then the laptop companies use these CPUs anyway they want to, as long as they don't exceed the original specifications.

This is what your manual has to say about your CPU.

And this is what AMD has to say about it.

HP is using the configurable TDP (cTDP) option to set the default TDP 10W lower (in complete accordance to AMD designer's specs). For short durations, a higher TDP limit will be used (though I doubt it will be 54W) and when pushed the CPU will be power throttled (probably to the base 35W TDP).

Also, there is the problem of thermal throttling that occurs when reaching TJMax, but you have to understand that laptop manufacturers include an offset that lowers slightly -or even a lot- the 100 degrees temp limit. HP usually has a 97-98 degrees thermal limit.

Unfortunately, neither the TDP nor the Tjmax would cause the ""problem"" you're seeing on your system. Or the "nano-second" theory for that matter..... The problem you're seeing comes from the fact that laptop manufacturers nowadays have set a third limit to their CPUs, via what is called Embedded controllers, and this limit lowers the "Max Boost Clock" to a (more appropriate) lower multiplier, thus avoiding having their laptops run back and forth between power limits and Tjmax limits.

I haven't seen any reviews/testing on the ElitebookG10s yet, but the previous gen also had a lower max clock. You can read the review of the 845G9 which also had a powerful CPU (AMD Ryzen 9 Pro), this generation didn't have cTDP yet, but even so the max speed of this Ryzen chip was 3.85GHz instead of 4.9GHZ.

What can you do? Nothin'. This is the way laptops are designed these days.

[ffgroup]

Hi!

In HWiNFO64, CPU-Z, Geekbench 6 (HP ZBook Firefly 14 inch G10) never show 5,1GHz. Max. frequency is around 4,4919GHz. So laptop is noticeable slower. Other HP laptop with similar configuration go 5,090GHz (HP EliteBook 845 14 inch G10 Notebook PC), which is weird 😞

[inthemirror]

Thank God I'm not the only affected person.

Hopefully if there are enough of us they will notice and fix the issue.

[inthemirror]

You're right about power constraints. At no point I've hit them.

 I've left HWiNFO64 running for days doing absolutely light work, not stressing the CPU at all, not hitting TDP/TTP ever. My CPU never went above 4.5GHz.

As an experiment open HWiNFO64 on your laptop. You'll instantly see your CPU boosting to max frequencies all the time.

I've already tried to contact AMD but they are not trying to help. "Please provide an msinfo32  report". Why? What will it tell them?

HP support has recommended me to install ... drivers. Drivers for the CPU? Um, sorry, there's nothing on the website.

This is all quite ugly to be honest. HP support even wanted to replace my unit. Why would it help if firmware is broken?

Too bad HP engineers never visit these forums, so this topic is completely in vain.

Please check the replies from the HWiNFO64 developer who knows a bit or two about x86 CPUs:

https://www.hwinfo.com/forum/threads/frequency-limit-global-for-ryzen-7-7840hs-zen-4-phoenix-cpu.904...

---

@TzortzisG wrote:

I haven't seen any reviews/testing on the ElitebookG10s yet, but the previous gen also had a lower max clock. You can read the review of the 845G9 which also had a powerful CPU (AMD Ryzen 9 Pro), this generation didn't have cTDP yet, but even so the max speed of this Ryzen chip was 3.85GHz instead of 4.9GHZ.

Let me quote the review:  "The Ryzen 9 PRO 6950HS very briefly runs at 50 watts and 3.85 GHz under load - on all cores."

My CPU doesn't hit the maximum advertised frequency when I'm using a single core. I've never talked about muli-threaded workloads.

[inthemirror]

Here, 7840HS hitting 5.1GHz:

https://www.reddit.com/r/AMDLaptops/comments/14au0b5/ryzen_7_7840hs_lenovo_loq_15aph8_cpu_benchmark/

Also notice that L3 Cache frequency is reported correctly unlike with my system.

[TzortzisG]

 Yes, inthemirror, you're correct about the review. I had noticed the difference, it's just that reviewers never tell the whole story, sometimes you have to read between the lines, sometimes they completely refrain from mentioning stuff.

 I don't see where we're disagreeing. Your AMD CPU SHOULD (definitely) reach its maximum frequency (especially in idle), and I've never doubted it doesn't on your laptop. One way to look at this is as a BIOS bug. In that case 2 options are available: HP Support and their examination of your system, possible replacement of mobo. And second option to wait for a fixed BIOS to be uploaded by HP. You can definitely do either one of those things. Will they work? Imho, I don't think so. I have to admit I'm not an AMD or Linux kind of guy, my knowledge of either subject is just enough to get me by. I do use Linux distros, not a big fan though. But I DO know about Intels and the M$ OS. The thing with Intels is that there's this Windows program called Throttlestop (you've probably heard about it) that can read and write registers directly on the CPU. And the program's creator (Kevin Glynn) spends a lot of his free time trying to solve other people's throttling issues. Recently, just about half of the posts on his forum are throttling issues having to do with EC registers, which have very complicated parameters and enforce various throttling methods (on a BIOS firmware level). It goes without saying you can't get rid of them without modifying your UEFI firmware.

 Let me tell you an interesting story. Up until lately, many  HP gaming laptops' threads (mostly Intel) were about the high temperatures. They were basically complaints and worried users about the high 90C degrees or even 100C during gaming. There was no thermal throttling, they just wanted to know if there was a problem. These last months, these different kinds of questions started appearing here about gaming laptops, and they were essentially about the same thing: Why right after a BIOS update their CPU no longer had its max frequency, but instead (even when idle) had a much lower max turbo boost speed. Coincidence? Broken firmware? Bad drivers/wrong Windows settings? Did it have sth to do with all those customers nagging about their multicore CPUs with weak cooling running at Gazillion Hz and turning their systems into mini ovens? You tell me....

[inthemirror]

Funny you say that, I'm a Linux user myself but I attached Windows screenshots because CPU/GPU/HW monitoring in Linux is extremely limited but the issue is reproducible under Linux as well. It's as easy as typing in one console:

7z -b -mmt1

and

watch 'cat /proc/cpuinfo | awk "/MHz/{print \$4}" | sort -n'

in another one. That again gets me exactly 4500MHz.

Applying to HP has proved useless: the first line of support refuses to raise the issue, so I'm stuck sending them useless debug logs. I cannot replace the device because it's not officially sold in my country and there's no HP support or retailers here. Shipping it back to the county where I bought it will take months to complete and is generally quite risky.

Waiting for updated firmware (BIOS) sounds like a good idea except someone actually needs to get in touch with proper HP support to report the bug. I've been unable to do so.

Speaking of fine tuning AMD CPUs, there's a RyzenAdj utility but it doesn't yet properly support Phoenix APUs, secondly it doesn't work at all once you enable Secure Boot:

# sudo ryzenadj --power-saving
pcilib: sysfs_write: write failed: Operation not permitted
pcilib: sysfs_write: write failed: Operation not permitted
pcilib: sysfs_write: write failed: Operation not permitted
pcilib: sysfs_write: write failed: Operation not permitted
pcilib: sysfs_write: write failed: Operation not permitted
PCI Bus is not writeable, check secure boot
Faild to get SMU, SMU_TYPE: 0
Unable to get MP1 SMU Obj
Unable to init ryzenadj

 I prefer to run with Secure Boot enabled. Fedora supports it perfectly.