amb669
New Member
- Total Posts : 92
- Reward points : 0
- Joined: 2010/03/29 19:52:23
- Status: offline
- Ribbons : 2
hello everyone i'm having issues when enabling HPET (64bit) on windows 7, i get a very noticeable loss in performance everything seems really jittery (what looks like loss of fps) and very slow load times, i noticed awhile back that dpc latency checker would show my latency at around 70-120us (with spikes of up to 1200 while opening/closing apps) with HPET enabled and around 12-15 (with spikes of up to 100 awhile in use) with HPET disabled, does anyone know what's goin on or how i can enable it without losing mad performance (i notice my mouse lag while at my desktop it's that bad) i also notice theres a device option for HPET in the device manager but when i go to properties it says "no drivers are installed for this device" could this be my issue? any help would really be appreciated
intel core i7
evga x58 3 way sli mobo (e780)
evga gtx 460 gpu
corsair hx850 watt psu
corsair 6GB ram (9-9-9-24-2t)
windows 7 ultimate
thanks
dustin
|
eppopipe
CLASSIFIED Member
- Total Posts : 2239
- Reward points : 0
- Joined: 2009/06/28 16:38:16
- Status: offline
- Ribbons : 7
Re:HPET destroying performance
2010/10/17 22:31:30
(permalink)
Maybe its just me, but its sounds like a HD problem. Test you HD for bad sectors *shrugs*
|
skimtneer
Superclocked Member
- Total Posts : 122
- Reward points : 0
- Joined: 2009/02/21 10:36:01
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2010/10/18 10:24:12
(permalink)
By the way, I remember HPET was not in the earlier BIOSes for my E758 (X58 3x SLI) board. Then somewhere over the next year it showed up as a new feature. From the description of it, it sounds like the only reason to enable it is if we're running some software that needs high precision event timing. Can anyone give examples of common games or other software that actually need this? Otherwise I'm thinking I should just turn mine off in the BIOS. Update: obviously not in great depth, but the Wikipedia article on HPET is a good first look for those who (like me) didn't know much about it. In a nutshell, it sounds like with HPET enabled your system can do a better job of synchronizing multimedia streams. http://en.wikipedia.org/wiki/High_Precision_Event_Timer
post edited by skimtneer - 2010/10/18 10:29:27
|
amb669
New Member
- Total Posts : 92
- Reward points : 0
- Joined: 2010/03/29 19:52:23
- Status: offline
- Ribbons : 2
Re:HPET destroying performance
2010/10/18 12:34:59
(permalink)
eppopipe
Maybe its just me, but its sounds like a HD problem. Test you HD for bad sectors *shrugs*
i tried this and sure enough i had an error during stage 4/5 "windows replaced bad clusters in file 57578 of name \hiberfil.sys" with HPET (64bit) enabled i don't see slow load times anymore  but i can still notice the slow/jittery mouse if i disable HPET it's nice and smooth 
|
Ruler
New Member
- Total Posts : 20
- Reward points : 0
- Joined: 2009/12/07 16:25:48
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2010/10/19 23:22:37
(permalink)
At the risk of sounding like an idiot, how do you test your HD?
Windows 7 x 64 EVGA (X58) 132-BL-E758 Motherboard Intel Core i7 Extreme 965 Corsair H70 Corsair Sandforce 120 (SSD) Patriot Warp 128gb (SSD) 500 gb Velociraptor WD 2TB Green for storage 12 GIG DDR3 (1600) Corsair Dominator Ram 2 x EVGA Nvidia GeForce GTX 295 (SLI) Thermaltake Tower 2 x 24 inch Samsung Monitors (245BW) (T240HD)
|
XrayMan
Insert Custom Title Here
- Total Posts : 63846
- Reward points : 0
- Joined: 2006/12/14 22:10:06
- Location: Santa Clarita, Ca.
- Status: offline
- Ribbons : 115
Re:HPET destroying performance
2010/10/19 23:32:55
(permalink)
My Affiliate Code: 8WEQVXMCJL Associate Code: VHKH33QN4W77V6A   
|
eppopipe
CLASSIFIED Member
- Total Posts : 2239
- Reward points : 0
- Joined: 2009/06/28 16:38:16
- Status: offline
- Ribbons : 7
Re:HPET destroying performance
2010/10/19 23:34:29
(permalink)
|
Ruler
New Member
- Total Posts : 20
- Reward points : 0
- Joined: 2009/12/07 16:25:48
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2010/10/20 12:08:23
(permalink)
Thanks, The HD Tune Pro is easy to use. I don't know what all the graphs mean, but my error scan was in the green. I'll spend a little more time trying to figure it out.
Windows 7 x 64 EVGA (X58) 132-BL-E758 Motherboard Intel Core i7 Extreme 965 Corsair H70 Corsair Sandforce 120 (SSD) Patriot Warp 128gb (SSD) 500 gb Velociraptor WD 2TB Green for storage 12 GIG DDR3 (1600) Corsair Dominator Ram 2 x EVGA Nvidia GeForce GTX 295 (SLI) Thermaltake Tower 2 x 24 inch Samsung Monitors (245BW) (T240HD)
|
zalbard
Superclocked Member
- Total Posts : 194
- Reward points : 0
- Joined: 2009/12/14 02:30:24
- Location: Germany
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 10:15:46
(permalink)
Well, I figured it out. This is a BIOS bug. I owned EVGA X58 Classified E760, now I moved to Classified E762. With pretty much bare bones fresh install and every single non-vital device disabled, I was getting both "DPC routine execution time" and "ISR routine execution time" of around 130-150 µs, with occasional spikes to over 200 µs. Way too high since most people are getting single or double digit results (see here). I kind of gave up on it for a while... Recently, while being bored, I downloaded and installed LatencyMon to monitor DPC and ISR Latency. Quite a handy tool, which shows processes that create DPC and ISR delays. So, the problem ended up being "ntoskrnl.exe" - Windows kernel (core)... Which means that no connected device is responsible, but the system itself. So, I went into BIOS, and disabled pretty much everything I could. Long story short: disable HPET (entirely) and disable C6 State (C3 and other ones are OK). My DPC is now down to 10-14 µs, same goes for ISR. Occasional spikes to around 50-100 (thanks to HDD activity, I guess, could not figure it out yet... they only happen once every 10-30s, though, so not such a big deal). Around 10 times lower on average. EVGA, please fix HPET and C6... It's not supposed to happen. For everyone who wants to check their systems out, download DPC Latency Checker (the best tool for DPC monitoring, IMO) and LatencyMon (monitors both DPC and ISR, and also shows the processes and the delays they are causing).
post edited by zalbard - 2011/01/20 10:24:10
|
zadinex
New Member
- Total Posts : 95
- Reward points : 0
- Joined: 2006/05/07 12:20:33
- Status: offline
- Ribbons : 1
Re:HPET destroying performance
2011/01/20 11:52:20
(permalink)
zalbard
Well, I figured it out. This is a BIOS bug.
I owned EVGA X58 Classified E760, now I moved to Classified E762.
With pretty much bare bones fresh install and every single non-vital device disabled, I was getting both "DPC routine execution time" and "ISR routine execution time" of around 130-150 µs, with occasional spikes to over 200 µs. Way too high since most people are getting single or double digit results (see here).
I kind of gave up on it for a while...
Recently, while being bored, I downloaded and installed LatencyMon to monitor DPC and ISR Latency. Quite a handy tool, which shows processes that create DPC and ISR delays.
So, the problem ended up being "ntoskrnl.exe" - Windows kernel (core)... Which means that no connected device is responsible, but the system itself.
So, I went into BIOS, and disabled pretty much everything I could.
Long story short: disable HPET (entirely) and disable C6 State (C3 and other ones are OK).
My DPC is now down to 10-14 µs, same goes for ISR. Occasional spikes to around 50-100 (thanks to HDD activity, I guess, could not figure it out yet... they only happen once every 10-30s, though, so not such a big deal).
Around 10 times lower on average.
EVGA, please fix HPET and C6... It's not supposed to happen.
For everyone who wants to check their systems out, download DPC Latency Checker (the best tool for DPC monitoring, IMO) and LatencyMon (monitors both DPC and ISR, and also shows the processes and the delays they are causing).
This perhaps explains what has been happening to my system. With a i7-980X, E770, and all bios available for it. When I enable C6, system performance becomes jittery, sluggish, slower performance in general as if additional latency was applied. I also have HPET enabled in the bios. When I disable C6, normal performance resumes. My system is running at stock settings, I'll have to give the latency checker,monitor a shot.
i7-980X, EVGA Classified 3, Asus GTX 1080 Strix OC, 24GB G.Skill, Samsung 840 Pro 250GB, Antec TPQ 1200 Power Supply
|
rafale
FTW Member
- Total Posts : 1554
- Reward points : 0
- Joined: 2010/01/09 20:11:47
- Location: Portland, OR
- Status: offline
- Ribbons : 6
Re:HPET destroying performance
2011/01/20 12:00:23
(permalink)
Interesting. Will try this as well.
|
5thduke
CLASSIFIED Member
- Total Posts : 2726
- Reward points : 0
- Joined: 2008/04/12 19:25:57
- Location: ociania
- Status: offline
- Ribbons : 3
Re:HPET destroying performance
2011/01/20 13:02:23
(permalink)
 interesting thread i checked mine was getting 220us disabled HPET
Attached Image(s)
never attempt to remove a fly from your friends head with a meat cleaver .
|
HalloweenWeed
CLASSIFIED Member
- Total Posts : 4792
- Reward points : 0
- Joined: 2009/11/13 03:32:55
- Location: Maine, USA
- Status: offline
- Ribbons : 37
Re:HPET destroying performance
2011/01/20 13:32:40
(permalink)
I didn't see anyone here asking you if you installed 64-bit Windows. Even if you think you did, you may have grabbed the wrong disc. R-click on Computer and look at "System type" to make sure. Even if you are running 64-bit, you installed with HPET set to 32 so you may need to reinstall.
Main (gaming) rig: i7-3930K; Asus Rampage IV Extreme; H100 W/p-p Excaliburs, AS5; MSI 7870 2GD5/OC; Crucial M4 SSD 256GB. See my primary ModsRigs: Adam for the rest, and I have a second (wife's) rig Asus Rampage III Extreme & 960: Eve.   Overclocking is useless to me if it is not rock stable.
|
Brazen_NL
FTW Member
- Total Posts : 1227
- Reward points : 0
- Joined: 2010/12/01 18:12:19
- Location: The Hague, The Netherlands
- Status: offline
- Ribbons : 5
Re:HPET destroying performance
2011/01/20 13:49:19
(permalink)
If I enable HPET, resume from sleep isn't reliable. I also haven't seen HPET as a requirement for anything I do so I just went with it.
|
Crusheddream
FTW Member
- Total Posts : 1985
- Reward points : 0
- Joined: 2009/11/06 00:58:49
- Location: Central NY
- Status: offline
- Ribbons : 12
Re:HPET destroying performance
2011/01/20 14:28:45
(permalink)
So whats the verdict on HPET does it lower performance? Would disabling it increase performance any?
|
zalbard
Superclocked Member
- Total Posts : 194
- Reward points : 0
- Joined: 2009/12/14 02:30:24
- Location: Germany
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 14:31:16
(permalink)
It does cause lag, yes. It shouldn't, though... Guess there's something wrong with the current implementation... Same goes for C6.
|
5thduke
CLASSIFIED Member
- Total Posts : 2726
- Reward points : 0
- Joined: 2008/04/12 19:25:57
- Location: ociania
- Status: offline
- Ribbons : 3
Re:HPET destroying performance
2011/01/20 14:33:52
(permalink)
crusheddream
So whats the verdict on HPET does it lower performance? Would disabling it increase performance any?
i have not noticed any differences with my performance with it disabled just the graph saying my latency is a lot better.
never attempt to remove a fly from your friends head with a meat cleaver .
|
Serben
New Member
- Total Posts : 10
- Reward points : 0
- Joined: 2008/12/26 07:53:03
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 14:36:28
(permalink)
zalbard
Well, I figured it out. This is a BIOS bug.
I owned EVGA X58 Classified E760, now I moved to Classified E762.
With pretty much bare bones fresh install and every single non-vital device disabled, I was getting both "DPC routine execution time" and "ISR routine execution time" of around 130-150 µs, with occasional spikes to over 200 µs. Way too high since most people are getting single or double digit results (see here).
I kind of gave up on it for a while...
Recently, while being bored, I downloaded and installed LatencyMon to monitor DPC and ISR Latency. Quite a handy tool, which shows processes that create DPC and ISR delays.
So, the problem ended up being "ntoskrnl.exe" - Windows kernel (core)... Which means that no connected device is responsible, but the system itself.
So, I went into BIOS, and disabled pretty much everything I could.
Long story short: disable HPET (entirely) and disable C6 State (C3 and other ones are OK).
My DPC is now down to 10-14 µs, same goes for ISR. Occasional spikes to around 50-100 (thanks to HDD activity, I guess, could not figure it out yet... they only happen once every 10-30s, though, so not such a big deal).
Around 10 times lower on average.
EVGA, please fix HPET and C6... It's not supposed to happen. This bug was uncovered in Linux 2.6.35 and I was one of the people affected by it: With HPET enabled everything involving a graphical environment was SLOW--and not just 200us slow, 2-5 seconds slow. The developer that fixed it implied that it was a chipset bug, not a BIOS bug so you could be SOL unfortunately ... I think the workaround was to read back from the HPET where the spec didn't require it. He called HPET a 'brain dead' design, but that's by the by. EDIT: Here's the explanation, in case you're interested. EDIT 2: . . . . Further the write to the comparator register is delayed internaly up to two HPET clock cycles in certain chipsets (ATI, ICH9,10) . . . (X58 includes the ICH10)
post edited by Serben - 2011/01/20 14:50:28
|
Brazen_NL
FTW Member
- Total Posts : 1227
- Reward points : 0
- Joined: 2010/12/01 18:12:19
- Location: The Hague, The Netherlands
- Status: offline
- Ribbons : 5
Re:HPET destroying performance
2011/01/20 14:37:14
(permalink)
5thduke
crusheddream
So whats the verdict on HPET does it lower performance? Would disabling it increase performance any?
i have not noticed any differences with my performance with it disabled just the graph saying my latency is a lot better.
Same here. Just the S3 resume issue I posted about, nothing else.
|
Rgallant
SSC Member
- Total Posts : 605
- Reward points : 0
- Joined: 2007/12/22 13:56:38
- Status: offline
- Ribbons : 1
Re:HPET destroying performance
2011/01/20 15:11:47
(permalink)
-I sort of wonder if something like this needs to be turn on before the win-7-64 load.
3770k [KL370]|| 3xMCR320[basement] || ek bay res|| 655x2 ek-tops || 9xSFF21F || Vl4n MVF WC || Evga GTX 780 sc 941 sli EK block,s Sam 256gb 840 Pro|| HAF 932 || 8 gb Tridents 2400 W7HP-64 || 27"A850D || AX1200w || G15,G13,G500s SB-Z -analog >Yamaha 765 AVR, 4x PSB alpha b1 ,1 x PSB alpha C1 ,10" sub
|
JDookie
FTW Member
- Total Posts : 1813
- Reward points : 0
- Joined: 2009/06/22 10:49:19
- Status: offline
- Ribbons : 5
Re:HPET destroying performance
2011/01/20 15:34:49
(permalink)
Rgallant
-I sort of wonder if something like this needs to be turn on before the win-7-64 load.
I am going to test exactly this over the weekend. I installed Windows 7 64bit with HPET 64 enabled, so I *shouldn't* have the problems listed, but after running the test, if it shows higher than normal latency I will disable it and see what happens. Very interesting find by the way. Thanks, amb669.
i7 3770K @ 4.8GHz (Koolance CPU-380i water block) Gigabyte Z77X-UP7 2 x 8GB Corsair Dominator Platinum C9 1866MHz Dual 240GB Corsair Neutron GTX in RAID 0 EVGA GTX 780 Ti Classified K|NGP|N w/ EK water block Corsair AX1200 Dell 30" 2560x1600 U3011 Monitor Koolance ERM-2K3U (copper) external cooler Corsair SP2500 2.1 Speakers | Win7 x64
|
zalbard
Superclocked Member
- Total Posts : 194
- Reward points : 0
- Joined: 2009/12/14 02:30:24
- Location: Germany
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 15:35:38
(permalink)
It's not just HPET, it's also C6 that's at fault, and that puzzles me the most... Cause really, it's a totally standard thing, and I've never heard about it causing issues.
|
Serben
New Member
- Total Posts : 10
- Reward points : 0
- Joined: 2008/12/26 07:53:03
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 16:36:43
(permalink)
zalbard
It's not just HPET, it's also C6 that's at fault, and that puzzles me the most... Cause really, it's a totally standard thing, and I've never heard about it causing issues.
When HPET is available, the kernel uses it to calculate when to sleep. Usually HPET, a High Precision Event Timer (the name says exactly what it does), allows for more accurate decisions for when to sleep and when to run--so power management becomes more effective. If HPET it buggy (as it is on X58), then the higher precision becomes worse precision. C6 has the greatest wakeup latency, so it'd be the most noticeable. If the Windows kernel uses the same workaround that Linux used to, then the result is worse than the bog standard timer, since it's waiting to check the counter against the compare value. The kernel checks, fails, waits, checks, fails... over and over in a busy loop instead of giving the process using the timer cpu time. Hence all the lag. It's just doing a lot of nothing (or some other process is carrying on uninterrupted). In effect, HPET is of no benefit--instead it makes things worse. This is only the case on modern Intel chipsets though (apparently). Of course, in order to make sure that this is the problem, it'd be good to ask people with X58 boards from different manufacturers.
|
zalbard
Superclocked Member
- Total Posts : 194
- Reward points : 0
- Joined: 2009/12/14 02:30:24
- Location: Germany
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 17:12:50
(permalink)
Serben When HPET is available, the kernel uses it to calculate when to sleep. Usually HPET, a High Precision Event Timer (the name says exactly what it does), allows for more accurate decisions for when to sleep and when to run--so power management becomes more effective. If HPET it buggy (as it is on X58), then the higher precision becomes worse precision. C6 has the greatest wakeup latency, so it'd be the most noticeable. But why does the problem persist regardless of CPU load, then? I have high DPC latency with C6 enabled both when idle and running WCG / LinX (no time to sleep or wake up there). It just doesn't add up.
|
dejanh
FTW Member
- Total Posts : 1045
- Reward points : 0
- Joined: 2006/08/07 10:40:07
- Location: Vancouver, BC, Canada
- Status: offline
- Ribbons : 3
Re:HPET destroying performance
2011/01/20 17:20:54
(permalink)
|
Serben
New Member
- Total Posts : 10
- Reward points : 0
- Joined: 2008/12/26 07:53:03
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/20 17:43:47
(permalink)
zalbard
Serben When HPET is available, the kernel uses it to calculate when to sleep. Usually HPET, a High Precision Event Timer (the name says exactly what it does), allows for more accurate decisions for when to sleep and when to run--so power management becomes more effective. If HPET it buggy (as it is on X58), then the higher precision becomes worse precision. C6 has the greatest wakeup latency, so it'd be the most noticeable.
But why does the problem persist regardless of CPU load, then? I have high DPC latency with C6 enabled both when idle and running WCG / LinX (no time to sleep or wake up there). It just doesn't add up.
At a guess, there'll be more small timers being created and destroyed when it's idle, since all processes will only need short time 'slices'. In all honesty I am not an expert so I don't know. What I do know is that whether my computer was idle (in the sense that no processes were using cpu time) or not didn't affect whether the mouse jumped over the screen in ~2-second intervals with constant 800% load, and that disabling SpeedStep & C[X] prevented the latency problem. dejanh
It may or may not be related to HPET. See http://www.sevenforums.co...73543-dpc-latency.html
But if this is a known problem to people without Intel chipsets, then evidently something else is amiss, regardless.
|
zalbard
Superclocked Member
- Total Posts : 194
- Reward points : 0
- Joined: 2009/12/14 02:30:24
- Location: Germany
- Status: offline
- Ribbons : 0
Re:HPET destroying performance
2011/01/21 08:55:03
(permalink)
dejanh
It may or may not be related to HPET. See http://www.sevenforums.co...73543-dpc-latency.html
Thanks. Doesn't help the issue, though, at least in my case. Aleady running High Performance [power] mode with services taken care of, tried diagnostic OS mode, same DPC / ISR results.
|
dejanh
FTW Member
- Total Posts : 1045
- Reward points : 0
- Joined: 2006/08/07 10:40:07
- Location: Vancouver, BC, Canada
- Status: offline
- Ribbons : 3
Re:HPET destroying performance
2011/01/21 09:39:14
(permalink)
zalbard
dejanh
It may or may not be related to HPET. See http://www.sevenforums.co...73543-dpc-latency.html
Thanks. Doesn't help the issue, though, at least in my case. Aleady running High Performance [power] mode with services taken care of, tried diagnostic OS mode, same DPC / ISR results.
The thing is that the more your read about DPC latency issues you could be looking at pretty much anything causing delays. It could be software, it could be drivers, it could be hardware, it could be hardware in certain combinations, could be conflicts, etc. It really is very hard to diagnose the source. That's part of the reason why I posted the link earlier as well, to give people an idea that it may not necessarily be related to HPET.
|