Asus P6X58E Pro - i7 990x timing issue

[jhameier]

I've recently encountered a problem with the Asus P6X58e Pro MB with the i7 990x cpu. I run some pretty intense software that requires very tight timings and I started to run a little program that essentially just prints out timing every 10ms just to see how tight the tolerances are on any give box that I put together. Here is the code that I run...


#include <stdio.h>
#include <time.h>
#include <stdint.h>

int main()
{
struct timespec s,e;
uint64_t s_ns,e_ns;
clock_t type = CLOCK_REALTIME;
while(1)
{
clock_gettime(type, &s);
s_ns = s.tv_sec*1000000000UL+s.tv_nsec;
do
{
clock_gettime(type, &e);
e_ns = e.tv_sec*1000000000UL+e.tv_nsec;
}
while ((e_ns -s_ns) < 10000000UL); //10 ms

printf("DIFF : %i ns\n",(e_ns-s_ns));
}
}

I've noticed that on any other board/cpu combination, the code prints pretty close to 10ms(just slightly off). On the P6X56/i7-990x I get a hiccup every 14ms or so where it lags for about 4ms. This wouldn't be a big deal if it was just every so often but it is pretty consistently lagging. I've also checked running my simulation software that we run pretty close to 60 frames per cycle on other board/cpu combinations and this setup I am running at around 51 or 52 (this can be increased to 57 if I shut of HyperThreading).

I've tried this timing test with several different motherboards including different OS's (currently using Fedora 16). I'm not sure if this is board issue or a CPU issue (I know this is not the place to discuss CPU's) but the latest box I've tried it on is an Asus P5TWS PRO with an i7-980x and there is no sign of lagging.

Has anybody experienced any similar issues? I like to know if this is going to be a problem (I just received a shipment of 30 boxes with this motherboard/cpu combination and I would like to get this straighten while I have them in a testing environment. I need to put these new boxes into a production environment pretty quick so any suggestions would be helpful.

 

[suteck]

Check to make sure bios are up to date - newest revision, but unless you're using the same processor and memory on the other boards you won't know for sure which component it is. Or it might just be the combination of those particular components in conjuction with each other. So a different ram might work better with you cpu/mobo combination. Also, it could be just that that os might be causing the hiccup so a reinstall or different os installation might be required.

 

[jhameier]

Yep checked all bios and also tried different combinations of external hardware and software. I have not tried either the Motherboard or the processors with other combinations of MB's & CPU's to narrow it down to the MB or the CPU causing the issue. I was hoping that I would get a bite on here before I begin ripping a bunch of boxes apart. Latest estimate is that I'm loosing about 20% of my processing time. That is just not going to be acceptable if nobody here has experenced this type of issues with this combination then I have to start looking deeper and maybe get some answers from Asus and or Intel. (here are a couple of the timing outputs from a P5TWS PRO/i7-980x and the P6X58e PRO/i7-990x)

P5TWS PRO/i7-980x

DIFF : 10000015 ns
DIFF : 10000000 ns
DIFF : 10000010 ns
DIFF : 10000000 ns
DIFF : 10000006 ns
DIFF : 10000018 ns
DIFF : 10000007 ns
DIFF : 10000007 ns
DIFF : 10000005 ns
DIFF : 10000016 ns
DIFF : 10000022 ns
DIFF : 10000018 ns
DIFF : 10000017 ns
DIFF : 10000000 ns
DIFF : 10000019 ns
DIFF : 10000014 ns
DIFF : 10000004 ns
DIFF : 10000008 ns
DIFF : 10000007 ns
DIFF : 10000001 ns
DIFF : 10000003 ns
DIFF : 10000000 ns
DIFF : 10000004 ns
DIFF : 10000009 ns
DIFF : 10000006 ns
DIFF : 10000022 ns
DIFF : 10000022 ns
DIFF : 10000017 ns
DIFF : 10000015 ns
DIFF : 10000010 ns
DIFF : 10000015 ns
DIFF : 10000006 ns
DIFF : 10000004 ns
DIFF : 10000001 ns
DIFF : 10000001 ns
DIFF : 10000017 ns
DIFF : 10000003 ns
DIFF : 10000015 ns
DIFF : 10000010 ns
DIFF : 10000008 ns
DIFF : 10000005 ns
DIFF : 10000009 ns
DIFF : 10000000 ns
DIFF : 10000017 ns
DIFF : 10000015 ns
DIFF : 10000022 ns
DIFF : 10000012 ns
DIFF : 10000006 ns
DIFF : 10000014 ns
DIFF : 10000018 ns
DIFF : 10000004 ns
DIFF : 10000000 ns
DIFF : 10000009 ns
DIFF : 10000018 ns
DIFF : 10000000 ns
DIFF : 10000010 ns
DIFF : 10000021 ns
DIFF : 10000007 ns
DIFF : 10000015 ns
DIFF : 10000003 ns


P6X58e PRO/i7-990x

DIFF : 10000011 ns
DIFF : 10000008 ns
DIFF : 10000016 ns
DIFF : 10000013 ns
DIFF : 10000006 ns
DIFF : 10000002 ns
DIFF : 10116464 ns
DIFF : 10000015 ns
DIFF : 10000009 ns
DIFF : 10000011 ns
DIFF : 10000017 ns
DIFF : 11611863 ns
DIFF : 10000000 ns
DIFF : 10000005 ns
DIFF : 10000007 ns
DIFF : 10000001 ns
DIFF : 11589626 ns
DIFF : 10000009 ns
DIFF : 10000019 ns
DIFF : 10000006 ns
DIFF : 10000006 ns
DIFF : 10000011 ns
DIFF : 10000007 ns
DIFF : 10000003 ns
DIFF : 10000012 ns
DIFF : 10000007 ns
DIFF : 10000013 ns
DIFF : 10000001 ns
DIFF : 10724913 ns

Any direction you can point me in would be great before I begin tearing these boards apart.

 

[jhameier]

OK so I contacted ASUS and they gave me the canned reply to contact the support department to get an RMA. I don't think they actually read my email which I included documentation on 3 different boxes that I ran identical tests on. Hey If they want 30+ mother boards back so be it, but I bet the first time that I talk to the support department and ask for 30+ RMA's at one time is going to throw up a BIG red flag...

Has anyone experienced a timing problem with any other ASUS mother boards?

 

[jaquith]

This might help -> http://www.linuxjournal.com/article/6900 also you're comparing a consumer MOBO (gaming) to a professional (workstation) MOBO.

 

[jhameier]

Very interesting article. Not the problem but something to think about. The latest in the timing saga...

So I called the technical support line and got to 1st Level support who suggested that the problem was to CMOS battery. While that may possibly be a cause for boot issues, I questioned the fact that this could cause a specific repeatable timing drop in completely different boxes. My latest set of tests were on 2 different series of P6X58E's (a PRO and a WS - both with different RAM and Power supplies) showed a distinct drop in timing every 230ms. This was with no other process other than the OS and a small C timing program. I pressed the fact that 30 boards all with bad CMOS batteries was unlikely but was willing to humor them.

They finally agreed to push me up to 2nd Tier Tech support where I spoke with someone who actually understood what what I was talking about. Of course they couldn't give an explanation, but very quickly sent me an email with a Technical Test Report to fill out and said it would be sent to their engineers for further testing. Biggest issue is that we can only get our physics engine to update at 51Hz (56Hz with HyperThreading turned off) yet we have 3 year old boxes with Core2Extremes that can push a full 60Hz.

We have a running bet around our office that someone might have forgot to turn off some debugging flags (oops) but only time will tell (no pun intended) .

 

[jaquith]

I'm interested -- let me know how it turns out. The consumer and WS lines are as much dissimilar as they are similar; ASUS does significantly more rigorous testing for erroneous behavior on their WS lines than their Consumer lines.

It seems like some device is polling every 0.23 seconds ; example a keyboard or particularly a mouse polls at about that rate. You can try disconnecting all external inputs and if needed script your timer application to load and run @ startup.

Finding debugging-flags should be easy to find.

 

[jhameier]

Yea we thought about that also but I would hope that a polling message wouldn't take up 4ms (that would be one big message)