Dodgy CPU?

devnull

Hi All

I'm having a weird issue with my i5 CPU on my Spectre X360, in that I am having on and off CPU issues with relation to performance and throttling which makes little sense and is resulting in some corruption of files at times and various odd things.

The machine is prone to killing it's own performance on a regular basis where the CPU appears to be running at 100% but the fans are not even coming on and any diagnostic tools I use suggest that the CPU is running somewhere between 1.1ghz and 1.5ghz and bouncing between the two even when the CPU is maxing out.

This happened the last few days regardless if i am on mains power or high performance and the general performance and sluggishness of the machine is truly shocking, using an external monitor with a laptop screen for example is no longer possible because of the fact it exhausts the CPU to say play a game on one and watch a video on the other when it was possible previously.

The CPU is rated at 2.2ghz turbo boost to 2.7ghz and obviously I'm certainly not getting the power that I should be right now and everywhere I look on the machine suggests there is no reason why this should be the case and has only happened in the last few days, I have tried disabling speedstep in the bios, turning fan to always on etc, still no help in performance, besides the CPU is so cold it doesn't need the fan anyway.

HP seem to think that it's a CPU fault because no matter what I do with the machine, I just can't get the thing to heat up or go above 1.6ghz even if I throw ridiculously demanding tasks at it. Interestingly they mentioned that despite being out of warranty, if it was a fault on Intel's side, they may be able to replace the CPU for free?

In addition I just saved a game in Football manager before quitting and restarted the machine since I got fed up with the performance and guess what? The game was corrupt as was the word file I saved 10 minutes ago and a notepad document also became corrupted that I save which is certainly odd.

Now I've reinstalled windows and I still have performance issues which begins to point towards hardware. Now an Intel CPU fault is exceptionally rare in my experience, but what else can it possibly be? Note this is intermittent, if I restart the machine and leave it off it performs well for an hour before the issue starts again.

devnull

Just had a hard freeze just now.

Was running nicely at 2.2ghz and turboboosting up to 2.7ghz when required, then after about half an hour slowly started to drift downwards, first to 1.8ghz, 1.6ghz, before going to 1.1ghz and finally 900mhz whilst the machine got colder and colder and the cursor got stuck and nothing would respond, had to do a hard reset of it.

As the speed went down the CPU usage went up to 100% and became less and less responsive but the tempreture got lower and lower so it's not heat. Restart and we're up to perfect performance again, until it happens the next time.

Alanstrainor

How old is the machine? If you have been in touch with HP, and they are suggesting a CPU fault I wonder if you're not the only one with the problem.

If the machine is relatively new I would probably bring it back to the shop and see what they say. If they offer a replacement (unlikely...) then that would be ideal.

devnull

Machine was bought in March 2015 so out of warranty sadly.

I'm starting to get blue screens as well now which is a red flag which are showing a DPC Watchdog Violation which normally occur following a 30 second freeze where everything stops responding and the lights on the keyboard go out and it drops USB power to all devices but oddly the fan keeps whiring away at the same speed for 30 seconds even when it has no need to (machine becomes very cold!), essentially the machine stops making all decisions in this time about its fan control as well. However blue screens always happen without a performance drop off in the time before them.

The hard freezes are always following a drop off in performance over the previous hour which goes hand in hand with lower clock speeds, higher CPU usage and ever decreasing temperatures and they don't blue screen, they will sit there for 5 minutes and not move and the only way to restart the machine is to power it off. Nothing in the error logs apart from the fact that the machine had an unexpected shut down. This appears to be simply the result of the system choking, a constantly reducing clock speed and the same workload leads to the CPU maxing out at 100% pretty much the whole time.

I've ran SSD drive checkers, RAM checkers and Intel's own CPU check utility and nothing turns up. It's quite weird because this machine was fine until yesterday and I simply cannot trust it now to be stable. The most bizarre thing is looking at CPU clock speed utilities like CPU-Z, shows that when I have nothing more than a browser open the CPU stays at 2.2ghz all of the time.

So it appears to be throttling that is causing the slowdowns and hard freezes, since I noticed under load is when the CPU frequency can drop off by up to 60%. However the fan is very effective and the temperature doesn't lend itself to be at a point where throttling is needed. Something I've seen in the last hour is like these figures.

CPU - Temp - Speed
100% - 48C - 1.4ghz (Machine CRAWLS)
72% - 50C - 1.9ghz
50% - 51C - 2.5hz
30% - 52C - 2.7ghz

That seems to be throttling in reverse to what it should be? Clock speed of the CPU is 2.2ghz, note how Turboboost gets invoked when it's not needed?

Just before the last crash i saw CPU 100% Temp 42C and speed 1.1ghz.

devnull

Here's a graph of CPU Temp, Clock Speed and CPU Load using Intels own diagnostic software from very first boot.



Blue = CPU Load
Green = Temp (Min 49, Average 53, peaks at 57)
Red = CPU Clock (Min 1.1ghz, max 2.7ghz)

You can see for the first 25% things look pretty normal although towards the end of it you can start to see some throttling with higher clock speeds starting to become less and less often when the CPU is under load.

For the second quarter this pattern becomes more obvious with the CPU now barely able to get about 1.5ghz, however in the third quarter ti does bounce back a few times when the load goes down.

The fourth quarter is Armageddon as far as the performance is concerned and by this point the CPU is permanently stuck around 1.1ghz and does not increase even when load drops and a restart is required in order to restore performance.

The important thing to say is during the 30 minute block that this comes from, the machine was doing repetitive tasks over and over again to simulate this performance drop off, with all background processes disabled so I was not subjecting it to different tasks.

The higher CPU load is not related to me doing more tasks on the machine, but reflected in the fact the lower the CPU speed becomes, the higher respective load of the CPU to do the same tasks.

The fan at all times is not audible as the performance drops off. This is totally absurd. If the fan was maxing out and the machine was getting too hot I'd totally understand this throttling, but it doesn't make sense to throttle the CPU speed to less than half of it's clock and slow the machine down so it's virtually unusable, when it's not warm and the fan is barely being used.

TerrorFirmer

It sounds unlikely to me that the CPU is causing many of the issues you've describing. File corruption would not be caused by the CPU, and you've described the machine as 'crawling' at 1.4Ghz. At that speed, an i5 is still lightning fast for normal desktop use.

Throttling would not cause noticeable slowdown in desktop/casual use and it wouldn't cause hard lock ups either unless the PC was actually overheating, which is not the case going from those reported temperatures or anything close to it.

When it's running at 2.2Ghz all the time, I'd check the task manager and see what's eating into the CPU cycle. It's quite possibly some Windows service, a lot of people with laptops find Windows Updates service running in the background keeps their CPU constantly at it's maximum state and often fans too. Try disabling updates and seeing what happens.

Overall, sounds more to me like a bad motherboard, impossible to say conclusively but CPU seems less likely to be to blame.

At the end of the day, a PC should last for more than that length of time so I'd bring it back to wherever you bought it from and outline your issues, if they steadfastly refuse and give you the '12 month warranty' spiel, you can tell them you'll take it to the Small Claims Court. A decent, relatively pricey laptop should last longer than 18 months.

Explosive_Cornflake

Is there any BIOS updates available?
Sounds like a heat/CPU issue. It's throttling itself due to heat, but the fans are set to come on after it's reached that temp.
A BIOS update would be good if the above were true.

devnull

TerrorFirmer wrote: »

It sounds unlikely to me that the CPU is causing many of the issues you've describing. File corruption would not be caused by the CPU, and you've described the machine as 'crawling' at 1.4Ghz. At that speed, an i5 is still lightning fast for normal desktop use.

Crawling I mean when using two screens, previousy I could use one for light gaming (football manager) or HD video and use the other screen for browsing with no slowdown. Hell I could even play FM and watch youtube at the same time with no problems in the past also.

For that kind of usage scenario the CPU should be going to a higher clock speed, yet when I close down the game or Youtube in HD, the clock speed goes up and the CPU usage goes down, it's odd because it's the more power I need, the less I seem to get.

Throttling would not cause noticeable slowdown in desktop/casual use and it wouldn't cause hard lock ups either unless the PC was actually overheating, which is not the case going from those reported temperatures or anything close to it.

When it's running at 2.2Ghz all the time, I'd check the task manager and see what's eating into the CPU cycle. It's quite possibly some Windows service, a lot of people with laptops find Windows Updates service running in the background keeps their CPU constantly at it's maximum state and often fans too. Try disabling updates and seeing what happens.

When it's running at 2.2ghz all of the time, that is when the machine performs very well and can mutli-task without any problems, however it only lasts for about half an hour after each boot before it resorts to crappy performance mode. All of those background services are already disabled. I'm not complaining about it running at 2.2ghz all of the time because this is the holy grail for me, this is when the machine is usable.

However, I have just spoke to HP on the phone who have given me an interesting explanation.

Explosive_Cornflake wrote: »

Is there any BIOS updates available?

Currently on the latest BIOS which is dated June 2016.

Sounds like a heat/CPU issue. It's throttling itself due to heat, but the fans are set to come on after it's reached that temp.

I agree. So I changed the setting in the BIOS to fans always on, which made very little difference apart from fans being always on from when the moment I booted up.

The problem is that apart from the first half hour after turning the machine on, the fans never go above humming. Now I know from past experience that the fans can go very fast and much louder, but after the first half an hour where I get good performance and clock speeds, this never happens.

The fans are barely being worked, I've noticed if I run a benchmark on the CPU after booting the fans can reach their full RPM from diagnostic software. However after this when the CPU starts throttling to 1.4ghz and won't go above, the fans never go about half their design RPM.

It's a very weird issue, the CPU is throttling itself when it's not even that hot, without even running the fans on more than half power to address the issue it thinks it has (but actually doesn't have)

Now to me it sounds like something has changed the threshold that makes the throttling kick in, without adjusting the temperature at which the fans should increase their speed so like you say the throttling kicks in at a low temp trigger without giving the fans a chance to cool it.

Incidentally I attached a graph from 3 weeks ago below, this is far more in line with what I would expect. Notice that load and temperature correlate in a way you would expect, with no drop off in CPU clock speed.

devnull

Some test results from running a few tests, gets more and more wierd.

CPU Stress Test - 30 mins
Temp got as hot as 85C
CPU was at least at 95% throughout
Not a single bit of throttling

Followed by 30 mins normal usage
After this I used the machine for half an hour and not a single bit of throttling and temperature varied between 48C and 64C depending on what I was doing and system ran flawlessly. Cpu Clock was mostly at 2.2ghz but boosted to 2.7ghz when needed as I'd expect.

Slowdown
Then suddenly Youtube started stuttering and the CPU temp had dropped to 42C, CPU load 100% and speed 1.2Ghz. So I thought lets go for another stress test here, the results were truly shocking

second stress test
Temp never went above 43C
CPU was at 100% throughout
Throttled as low as 900mhz, never above 1.2ghz

I don't understand it? Again it's an hour after a start up performance has dropped off a cliff.

Explosive_Cornflake

The way it happens after half an hour is odd.
Have you tried a fesh OS, as it could possibly be the power utilities acting up.
Even booting from a Linux USB might be useful.

devnull wrote: »

Some test results from running a few tests, gets more and more wierd.

CPU Stress Test - 30 mins
Temp got as hot as 85C
CPU was at least at 95% throughout
Not a single bit of throttling

Followed by 30 mins normal usage
After this I used the machine for half an hour and not a single bit of throttling and temperature varied between 48C and 64C depending on what I was doing and system ran flawlessly. Cpu Clock was mostly at 2.2ghz but boosted to 2.7ghz when needed as I'd expect.

Slowdown
Then suddenly Youtube started stuttering and the CPU temp had dropped to 42C, CPU load 100% and speed 1.2Ghz. So I thought lets go for another stress test here, the results were truly shocking

second stress test
Temp never went above 43C
CPU was at 100% throughout
Throttled as low as 900mhz, never above 1.2ghz

I don't understand it? Again it's an hour after a start up performance has dropped off a cliff.

devnull

I'm using third party software now to fix the CPU frequency which has resolved the problem, I have to deal with marginally hotter running, reaching early 70s under load as opposed to mid 60s, and reduced battery life by about 10% but it's much better.

Apparently there is a BIOS update coming soon "In light of changes related to recent operating system updates" I was told on Friday by HP, which suggests that it's related to a Windows update.

ED E

You're using the integrated GPU Yeah?

This sounds like the GPU is throttling the CPU as its shared die?

I'd definitely try a ubuntu live boot off a fast key and just use it for a day to ensure its not a HP utility f'cking with the power states though.

devnull

Updated BIOS fixed it.

Yeah, it's built in HD5500, won't get such a dedicated chip in an ultrabook.

devnull

And now a further windows update has broke it again. The white line denotes the time the update was installed after reboot.

Notice how once again we go from a varying CPU Usage / Processor Speed / Temp readings which are completely normal, adjusting to usage to one where CPU usage in consistently high, CPU speed is very low and temp barely changes.

The Green temp at the end Hovers between 54-56C. Note following the drop of frequency it's very stable which suggests something is telling the machine to keep the temp between those two values.

At the end of the graph, the machine becomes unusable with persistently high CPU load and just 1.1ghz of clock speed despite being cool and with no fan audible and requires a reboot, just to use again for anything other than basic web browsing. Even after closing all apps the CPU stays at 1.lghz and right at the end of the graph, it hits 966mhz before freezing up totally.

Back to using third party tools to manually force the CPU speed again, may even dust off my old desktop soon, far less powerful processor but at least it doesn't cut it's power by over 50% constantly.

The irony is, the CPU with the combination of the fans is capable of operating at a much higher speed than it constantly caps itself on, what annoys the hell out of me is seeing a machine run cool limiting itself for no reason, which is effectively what happens here. May as well not have a fan since I rarely see it get used.

BTW, I know it's a shared die, but according to Intel tools guide, the temp it takes is the "package" temp so that should include both no?

devnull

HP Update installed a new updates, everything worked well, then following a restart windows started messing around and "updating" drivers as well.Pathetic game of cat and mouse going on with Microsoft and HP updates here it seems.

Now when I go into device manager and look at CPU I see:
"A driver (service) for this device has been disabled. An alternative driver may be providing this functionality. (Code 32)" and I keep getting annoying notifications in the bottom right hand corner to tell me something isn't functioning (but it's working fine)

So it seems I need to start circumventing a few things in Windows
1) Prevent ALL updates because they break what HP keeps fixing.
2) Stop those stupid notifications in the bottom right.

How do I do it?

ED E

If you have W10 Pro you can mostly nuke that behaviour with a GP setting. Cant be done on the home editions which is what you get on most retail laptops.


Honestly, if HP arent validating their drivers with the W10 driver program then I'd be installing Windows 8 or returning it and purchasing something different. HP are a company still in a bit of turmoil and I think this is probably just one of the symptoms. Had my hands on last years and this years Spectres during the week, lovely hardware but if you're gonna have hassle like this it just doesnt fly with the ultrabook market.

devnull

The drivers that are showing as not validated are actually Microsoft ones in Device Manager. The machine has been flawless for the 18 months I had it until a few weeks ago, just to try and see if it did anything I cleaned the whole thing out with insane amounts of compressed air to get everything as clean as possible incase it was contributing with little to no difference. Still I see needless throttle.

Incidentally the CPU core temps I see idling at about 35C and I've rarely seen it go above 65C even under load. The tempretures that i have quoted are "package temps" The Package temps are always 20C over the actual core temps.

Is that normal?

Under Load Max:
CoreTemp: 62C
Package Temp: 82C

Idle Avg
CoreTemp: 40C
Package Temp: 62c

Not sure if that helps to point out where the issue is, and why the package temp is so far higher than the core?

devnull

To let you know what happened with this, HP took the laptop back, replaced the I5 CPU with an i7-5500U) an gave all the components a clean and installed a new more powerful fan and gave me a 6 month warranty on the work.

Faulty CPU.

devnull

Unfortunately, the problem has now returned again w- surely it cannot be another CPU failure? It was flawless for the last two and a half months, but I once again have similar issue. No Bios Updates or driver updates from checking logs.

There seems to be a pre-set routine to thermally throttle the CPU so it does not go above 60C once more. The moment it hits 60C it will throttle the CPU more and more and more till it gets down to below 60 then it will allow it to power up again. What has triggered this **** again?

I have had a temp monitor monitoring for the last 2 months and all behaviour was normal UNTIL TODAY with throttling only taking place when on battery to save power and only other time it was throttling was when temp hit above 80C, which I'd expect.

Today we're back to constant throttle to the point where anything mildly demanding results in the throttling kicking in.

The question is why the sudden change from normal, back to this stupidity?