01-14-2020 01:58 AM
Hi Anon, that's unfortunate. Maybe you can try keeping HwInfo open and check motherboard temps. In the old (good) F22, the fans where linked to this temperature instead of the CPU: as soon as it reached 40°C, fans did start. It was not optimal, but was acceptable. In later BIOS revisions it requires a much higher temperature to allow the fans to start (above 65°C according to my tests, which is way too high and they never spin, leading to a throttled scorching CPU).
Do some test keeping an eye to it, you may find the same bahaviour as well as the trigger temperature. This would explain why the performance drops on fresh start: unit is cool = fans off, cpu overheat and goes in throttling. While internally it reaches 90+ degrees, it takes time to propagate the temperature to other components, especially with the fan turned off.
Try a game too. When the dedicated GPU is running, fans should start and the cpu should run full speed.
Check also the throttling source: power or temps? In all my tests, it's always caused by temps (power limit can lock max frequency to 3.6GHz, no lower).
A precise tracking of the problem may help when communicating with support....no hopes honestly.
I might check out HWInfo because so far, other than by ear, I've not found anything else that will tell me the fan status. None of the monitoring utilities that I've tried seem to have been able to recognise the fans.
But really, there's no point me doing further tests.
The tests this weekend have proved conclusively that it is completely inconsistent.
At least twice over the weekend, I was able to put the CPU @ 100% and it maintained 3.6GHz for an hour. That is plenty long enough to be thermally stable. On these occasions the fans kicked in exactly as one would expect, and maintained the CPU well above the base frequency at a level I would be very happy with if it would do this reliably and repeatedly on other occasions.
But then yesterday, in exactly the same configuration, repeating exactly the same scenario - (boot up, leave idle for 20 to 30mins, then put under load) - instead of the fans coming on, the CPU speed drops instead, and performance dies
This is now leading me to believe this probably is a genuine hardware fault.
Either that, or the BIOS is looking at sensors / inputs that are on a knife-edge in relation to each other or their trigger thresholds, and perhaps by chance sometimes it sees one situation first which triggers the CPU throttling, and on other occasions it sees the other side of the knife edge first which triggers the fans first. (And then which ever is triggered first, that then works to prevent the other triggering).
But that's speculation.
I've had enough - 4 months now premium support have been looking at this.
I have to take it at face value - they've had 4 months to fix this, and they haven't been able to.
That is now way more than reasonable time.
Perhaps it is a BIOS issue, or perhaps it isn't.
That's now irrelevant, because either way HP have been unable to fix it.
If Premium Support had been working proactively with me, such that I could see what they were doing, and see that what they're doing makes sense, etc, and see that genuine progress is being made towards a solution, I might have been inclined to keep working with them.
But they lie to me. They drag their heels. They don't tell me what's going on (at least not anything I believe anymore).
I get the impression they are out of ideas, and just hoping for a miracle - or they're hoping for me to get bored and just quietly disappear and leave them alone.
4 months on - they haven't fixed it nor show any sign of being able to fix it.
If I give them another 4 months, I have no confidence that in 4 months time I'll be any nearer a properly working PC.
I have to take this on face value as a 'generic' customer - they've had ample opportunity to fix this, and they haven't.
I paid for a high performance PC that I actually wanted to use.
I didn't purchase the computer so that I could sit there and monitor it's operation, and try to diagnose why it's not performing, etc. (I only spotted the performance issue when I tried doing productive work and it became impossibly slow to carry on - it was only then that I felt it necessary to investigate further - I just want to use the -bleeping- thing, not diagnose it)
I either want a working PC that I paid for, or my money back.
No more fixes, no more repairs, no more diagnostics (by me! HP support don't seem particularly to be doing much diagnosis themselves), no more spending weeks waiting for a promised BIOS that when it arrives makes no claims to fix any related problem.
HP simply aren't able to fix it.
That leaves either replacement or refund.
If HP aren't just fobbing me off, and genuinely do want to keep their customers happy, then I hope they to will also accept that the time for repairs is over.
It's time for them to be pro-actively offering me a replacement - after this length of time, I really shouldn't even need to ask them!
01-14-2020 04:23 AM - edited 01-14-2020 04:28 AM
I understand your feelings and i'm sorry. The point is: doing tests is in your interest (and mine) more than anyone else.
If the problem is, as i suspect, originated from BIOS, then there is no hardware faults. You can keep swapping motherboards/units for months without resolving the problem. It's their job, you are right. They can't care less of resolving this problem due to limited reports; it's a fact we learned on our skin.
This notebook has gimped performance due to thermal throttling, which is originated from the fan not spinning.
Those fans have never been controlled by CPU temperature, never. It's not common (and plain stupid from my point of view), but it's how it works. So, the fans do start based on an internal temperature threshold: 40°C for BIOS F22.
With recent BIOS revisions, they have raised this threshold in order to accomodate 90% of people who browse internet and don't want fans to kick in.
That's my conclusion. I may be wrong of course, that's why i'm sharing those informations with you: i'd like to come to a proper, verified conclusion.
If i'm correct, no swap/replacement will solve your (my) problem.
I'd like to know the BIOS revision of the new unit, in order to evaluate if this different behaviour you are experiencing may be correlated to a different BIOS revision.
I'd also appreciate if you can confirm or negate my suspects by launching a stress test with HwInfo open and checking if you resemble the same behaviour i described above. You will only see the MoBo temperature, the fans are not detected, but you can easily ear them spinning.
I'll not stress you on doing those tests anymore. I respect your feelings.
Do as you please, i'll be gratefull if we can work together to understand and confirm the problem, otherwise i'll post here if/when i come to a solution.
I am convinced that identifying the problem is essential for moving properly.
01-14-2020 05:26 AM
>> "The point is: doing tests is in your interest (and mine) more than anyone else."
I'm not sure how.
I can do all the tests in the world, but I don't have access to the BIOS source code, so if it is a BIOS problem then that knowledge from those test results doesn't help me if HP support aren't sincerely engaging with the issue. And for 4 weeks prior to christmas I was told second (and even at one point "3rd") line support were actively developing the new BIOS F40 to resolve this issue... yet BIOS F40 arrives and neither fixes the issue, nor in the release notes does it claim to fix the issue. So I'm now left to conclude that HP support are not sincerely interested in resolving it, if it is a BIOS issue. I feel I was being strung along. So I feel that if it is a BIOS issue, it's just going to end up a dead end, irrespective of what info I give them or what tests I've run..
That said, the results over the weekend and monday evening, lead me to think it very well might not be a BIOS issue.
If it were a BIOS issue - i.e. with flawed logic - I would expect it to be at least consistently flawed.
So when on Sunday I was twice able to achieve sustained good performance by first leaving the unit idle (CPU @ 0%) for around 15 to 20 minutes, I thought great - I now have a workaround, albeit with the inconvenience of having to idle it for 20 minutes every time I take the computer out.
But then on Monday - with the same BIOS, running the same configuration (keyboard back), repeating the same sequence - first leaving it idle for 30minutes before loading up the CPU... this time I simply could not get sustained good performance.
This time, when the unit had been stressed for a few minutes, instead of turning on the fans, it instead throttled the CPU.
On the same scenario on Sunday, it turned on the fans at this stage.
I tried at least twice on Monday evening, once with at least 10 minutes idle, then another with at least 40minutes idle (while I made dinner). After neither of these occasions was I able to repeat the solid 3.6GHz @ 100% CPU for extended periods that I had seen on Sunday after similar periods idle.
I now suspect that when I though I had good performance on F28 back in July, that might have been a similar fluke.
But How Should it Work?
One worry I've had all along is that once it gets to someone in the know, they turn round and say "hang on - it's working how it's supposed to - it isn't supposed to work at high performance with the keyboard back"
I sincerely hope that's not the case, because I bought the x360 solely with the intention of using an external keyboard while doing productive work and nothing that I saw in the sales literature warned that the non-regular-clamshell configurations would be seriously crippled performance (below the advertised base performance).
My previous computer was a Surface Pro 4 - this is smaller, thinner and lighter and is a tablet configuration only, yet I've run tests and that can sustain around comfortably above it's base speed of 2.2GHz when put under sustained load (3GHz+).
So I have no reason to believe that this bigger, heaver, fatter and at least 3 yrs newer HP x360 shouldn't also be able to sustain above it's (same 2.2GHz) base frequency in any of it's configurations.
If the Surface Pro 4 can do it, then something the size of this HP should have no problem - even in it's full tablet configuration with keyboard and screen back-to-back!
And I haven't seen anything from HP that cautions otherwise!
On Sunday, I've now seen it perform solidly @ 3.6Ghz @ 100% CPU utilisation in the Backward-L configuration. So I do now know that it certainly is capable of operating at high performance in this configuration - if it wants to!
What I can now conclude though, is that whatever the BIOS is basing its decision on, it isn't consistent.
And if it isn't consistent, that points to at least probably some kind of hardware fault.
Perhaps a sensor isn't properly registering the status of the screen relative to the keyboard. Or perhaps the action of rotating the keyboard back is knocking or shifting something inside - perhaps a loose wire somewhere.
And perhaps that is what is causing the inconsistent behaviour from the firmware.
While it feels good to now have seen high performance in the backwards-L configuration - so I now feel reassured it is capable of performing in this config - I would still be happier if anyone else with an x360 could confirm if they are able to obtain sustained high performance (i.e. CPU speed remaining at or above the base frequency when put under sustained 100% CPU load) in the Backward-L configuration (same performance as in the normal clamshell configuration).
If someone could confirm that, it would at least give me confidence that a replacement should fix the issue, and to stick with the x360..
If on the other hand, this is genuinely an intended limitation of the x360 and that it really is designed to throttle down to 0.8GHz when under load in backward-L configuration, then I would consider that I've been mis-sold / the unit has been falsely advertised, and I would really then be looking for a refund (using UK consumer rights if necessary, on the basis it isn't performing as described).
As an aside note - I just saw earlier today that the Microsoft minimum specification for running windows 10 is for a 1GHz CPU. So if this unit really is designed to throttle CPU speed down to 0.8GHz under load, then it isn't appropriate for windows 10 that it's delivered with!
But, from what I've seen I don't believe this is the case - I don't believe it's supposed to throttle down, and add into the pot that there doesn't seem to be a ground swell of people complaining about this, I suspect it now probably is something peculiar with the unit I have and not a broad issue with all units.
In which case, there is at least some hope that a replacement might resolve this problem - assuming HP will provide a replacement.
I just spoke with HP support again a short while ago before typing this, and they are now escalating this to HP customer services (or something similar, I forget the exact name) who they say are the department that can consider replacements or refunds and that I should expect to hear from them in the next couple of days.
We'll see how this goes...
01-14-2020 05:39 AM
BTW to answer your questions - it came back on BIOS F32
I was careful to test "as-delivered" and confirmed the issue still existed on the as-delivered unit before trying any updates.
Once confirmed, I then tried to update to F40, and also update the Intel Thermal Framework stuff which others have suggested (incl HP support just before it got shipped off).
But neither of these updates changed the issue - it was the same before as it was after.
One thing I would be curious - could you try the same test that I did over the weekend...
- Restart the computer.
- Leave idle for 20 - 30 minutes (with CPU 0%, so close anything using any residual CPU, like browsers, etc)
- Then try to stress test it
(e.g. with https://cpux.net/cpu-stress-test-online setting the threads to 12 or more, and power to 100% - that should then fully load all 12 logical processors to 100%)
- Monitor how it behaves with task manager using the "Performance" tab - does the CPU frequency remain above 2.2GHz?
- Leave it under that load for an extended period - e.g. 20 to 30 minutes.
When I tried the above on Sunday, it maintained 3.6GHz for 1hour @ 100% CPU utilisation, using the fans as needed (mostly continuously, which is what I'd expect) to control the thermals while still maintaining performance.
When I tried to repeat this on Monday, after around 5 minutes, the CPU dropped to ~1GHz and the fans remained off.
01-14-2020 05:57 AM
It's funny, i too bought the spectre in order to avoid the necessity of upgrading my Surface Pro 4.
The throttling is indeed designed by intel and expected, but it's a case-limit scenario. Sometimes, in particular conditions is acceptable; as a regular behaviour it's not.
Answering your questions, i have the same performance in all shapes so, at least in my case, there is no correlation from the throttling and the way you open it. I also found some inconsistency in the way the fans were controlled. After many tests, i think i learned to "read" key values to identify a consistent (stupid) behaviour which control the fans based on mobo temps.
I'll do the test in the evening
01-14-2020 07:44 AM
I did the test.
There is no difference between running it immediately after a restart or after 20+ min. in idle.
In both cases:
-The CPU start at 3.8Ghz on all cores @54w, fan off
-CPU throttle down to 3.4Ghz in power limit @40w after the short boost timeout. Fan off and power limit active.
-After less than a minute it reaches 98+°C, then it start decreasing CPU frequency and thermal throttling is triggered on 3 cores, fan off
-Thermal throttling flag turn on on all cores, frequency keep lowering and stabilize between 1.4-2Ghz going a little up and a little down. Fan off
-When the PCH temperature reaches 65°C the fans start and lower PCH temp to 60°C, then stop. About 20" of cooling.
-While fan spins frequency progressively rise to about 3Ghz, then it lower again as soon as the fans stops.
When PCH reaches 65°C again this cycle repeats.
I did not test past this point (about 5 minutes, more or less depending on initial temperatures) because i'm not confortable in keeping it at such high temperatures. I got a BSOD 2 times after CPU@98+°C for about 2 minutes.
This is with F32 firmware, a little different from my tests with F31. Slightly better from a performance viewpoint, still unacceptable when taking into account temperatures. I really cannot accept that the fans do not spin and i cannot force them to run in neither of the 3 profiles they provided.
I hope this will help you. My problem is not exactly the same as yours it seems.
01-15-2020 03:37 PM
The other day I said...
"I just spoke with HP support again a short while ago before typing this, and they are now escalating this to HP customer services (or something similar, I forget the exact name) who they say are the department that can consider replacements or refunds and that I should expect to hear from them in the next couple of days.
We'll see how this goes..."
Well, I just spoke to them today...
Summary : Absolutely Dreadful.
Turns out I was talking to "HP Customer Relations Team UK & Ireland"
And to be quite honest, the call couldn't have gone any worse.
They completely weren't interested in consumer rights.
He basically said that because their (failed) attempt at repair had taken so long that it was now outside the 6 month window of the consumer rights act, any rights I may have had specific to that first 6 month window no longer applied.
He was adamant that they have the right to make 2 repair attempts.
This does not correlate with the consumer advice here in the UK (for example https://www.which.co.uk/consumer-rights/regulation/consumer-rights-act ) that you only have to give the retailer one attempt to repair a fault that occurs >30days and <6 months, after which if the fault still remains, the consumer has the right to ask for a full refund.
Anyone thinking of purchasing an HP PC in the UK thinking they will have protections from the UK Consumer Rights Act, be aware that you may have a battle on your hands.
He was also adamant that 4 months attempting to fix it (and failing) was still a perfectly reasonable time frame, and that therefore I had no grounds to escalate, nor grounds to move onto insisting instead on a replacement, on the basis that it is now taking an unreasonable length of time
Anyone thinking of purchasing HP products might want to bear this in mind before they part with their money.
He also claimed that they had so far had only 1 attempt at repair.
What I infer from that, therefore, is that he considers that ..
- all the phone calls prior,
- all the remote connections from HP Premium support, and
- all the BIOS updates..
.... were not actually sincere attempts at fixing the problem.
The Upshot of this...
Here I am, 4 months after notifying them of a fault (within 3 months of purchase - well within the 6 months window of the consumer rights act), that I have now given them multiple opportunities to try to fix...
And he told me quite bluntly the only option he was going to provide was to send it for repair again.
When I asked what they would fix that they haven't been able to identify in the past 4 months, he just insisted "it is for the engineers to decide".
Quite why I should believe the engineers will now be able to solve the problem when they've failed consistently over the past 4 months!
There you have it - the HP experience.
01-26-2020 05:13 AM
This morning i took courage and updated to BIOS F40, then i made some tests. Something has changed (in better) and i'll share my thoughts below:
Settings: i enabled "FAN ALWAYS ON" in BIOS and selected COMFORT profile in HP COMMAND CENTER.
Running some CPU stress test lead to the usual 97°C and throttling, BUT:
1- The fans started to spin on PCH TEMPERATURE of 45°C (i used HwInfo to see this value, before was 65°C) and the cpu raised again to 3.5-3.6Ghz
2- The fans continued to spin and 3.5Ghz was mantained for the remaining test duration (about 2 min.)
3- The fans continued to spin even after the test finished and stopped only once the PCH temp lowered under 40°C
4- The fans started to spin again during normal windows usage, which NEVER HAPPENED since BIOS F32
5- The measured CPU performance dropped by 26% compared to BIOS F22
So there is indeed some hidden changes in F40 release. This does not resolve the problem, anyway i should be able now to use this laptop without having it burning in my hands. The CPU will still burn and throttle, but a lot less than before and internal notebook temperatures should be kept within 45°C; a big step since 60+°C reached and mantained in previous releases.
F22 was perfect for me, this is not, but i'll try to use it in order to evaluate if i can work on this unit again. I'll do this test within few weeks to see if this behaviour is reliable and maintained every time.
01-26-2020 12:23 PM
If you're curious, I sent the PC again for repair - after all, they didn't give me any alternative, as mentioned.
Have now received it back, and same fault still persists.
I've taken a video of it and put it on youtube if you're interested...
01-26-2020 03:46 PM - edited 01-26-2020 03:49 PM
I see your CPU temp is 65-75°C, so it's not the same problem as mine.
Is your power limit flag triggered? In this case try to raise the threshold in XTU or throttlestop and try again. It doesn't seem to be related to temp throttling like in my case. I suppose you already tried to set windows power plan to max performance. Assuming the pc is properly configured (which i do), if it's not throttling due to temperature, i can't think of another reason other than power throttling.
Anyway, i got sustained performance with F40; still to be confirmed outside of the quick test i did today, but i suggest you try the same settings i used (see above post) + throttlestop in performance mode (or XTU with raised power limit).
I know you are pissed off by months of useless tests/support, but you have a new hardware and i'd try in every possible way to see if i can patch it somehow before contacting the support again.