P35 Motherboards and DPC Latency Spiking

[badonkadonk]

Re: P35 Motherboards and DPC Latency Spiking

Edit: Oh umm even that is not great. They should Ideally be the same as the smallest ones without any spikes..

Also if you look in Vista performance logs there are many critical errors complaining of system degradation, windows components taking too long to start up, slow boot degradation...wtf is going on?? There is one common component, Gigabyte boards. Asus doesnt have this issue.

This is good:

[Lsdmeasap]

Re: P35 Motherboards and DPC Latency Spiking

Well, I dont see any real effects. I watch streaming video and music just fine, but that is not my main thing. I dont make or watch that type of thing daily so I may never notice it I guess.

Overall my system runs great, very great

[badonkadonk]

Re: P35 Motherboards and DPC Latency Spiking

Yes alot of people dont notice it but its still not good for performance as is not normal. The main guys that notice it are people doing audio editing etc, running audio hardware and software and there are some mega threads about it around the net. I have poor performance that may or may not be related to this issue and its very frustrating that i cant solve it even with fresh format etc. I have been troubleshooting for a couple of months on and off. This is my second X38 DQ6 and it has the same problem.

Gigabyte are providing bios fixes for p35 boards so they acknowledge there is an issue , but you have to ask them for the bios thru tech support. I am waiting on a reply from Gigabyte about X38DQ6 bios fix but i feel like giving up on this board and going back to Asus because they dont have this issue.

I feel it has something to do with the AHCI bios on this board that seems to be buggy.

[Lsdmeasap]

Re: P35 Motherboards and DPC Latency Spiking

Well have you tried not using AHCI?


Gigabyte has been great to me for the past 2 years since I switched, but I dont do anything like you mentioned above so I would likely never notice that issue

[aspenleaf]

Re: P35 Motherboards and DPC Latency Spiking

Disabling AHCI in the BIOS has no effect on the DPC spiking, at least on my system. I have the beta BIOS, and with all the same hardware and software, including AHCI, there are no spikes. The only problem I've seen with the beta BIOS is that I cannot change the CPU (vcore) voltage. I would think that would be a necessary and easy fix for the official release.
I think that proves it is a BIOS issue. And it does make a large difference in audio performance with ASIO drivers. I went from not being able to use a buffer of 64 samples on my Delta soundcards with the last official BIOS, to being able to use it with the beta BIOS with room to spare. For low latency recording, that is very important.

[Lsdmeasap]

Re: P35 Motherboards and DPC Latency Spiking

Well be sure to provide them feedback about the BIOS at the email which you got it so they know your thoughts. Only people who actually test the BIOS's and provide feedback can help so you are helping as I see it. And yeah that does make it sound BIOS related for sure

[aspenleaf]

Re: P35 Motherboards and DPC Latency Spiking

I've been corresponding with tech support and have kept them updated with my results when using the beta BIOS. I do appreciate that they are working on a fix for the problem. It will make this otherwise very fine motherboard into a great motherboard for audio workstations.

[b_rubenstein]

Re: P35 Motherboards and DPC Latency Spiking

FWIW: I just got a EP35-DS3R (v2.1) with a E8400 and I'm running it at FSB of 400 for a CPU speed of 3.6 GHz. 4 gig memory, 2 x 2, DDR2- 6400. Default memory timings and Performance Enhance set to Standard. BIOS is F2. OS is XP SP2.

DPC gives me a "staircase": bar at just under 1000ms, a couple low at 30us, one at 750us, low, 400us, low, spike back up to 1000us. On occasion a yellow spike. Highest latency I've seen was 1160us.

[Dewdman42]

Re: P35 Motherboards and DPC Latency Spiking

LsdMeasap,

This is indeed a troubling situation. Its true that many typical users will not notice the difference. You aren't going to notice a problem from watching a video or listening to an MP3. However users who are dealing with Digital Audio Workstations, video editing suites, etc...we are the ones that are trying to wring every ounce of performance out of these machines. If not, I would still be using my P4, not other reason to upgrade.

The problem is that this DPC latency problem is in fact reducing over all audio workstation performance. With an audio workstation you often need to reduce the size of audio buffers in order to get low latency response while using software based synthesizers and such. This makes the system extra susceptible to DPC latency spikes, which will manifest themselves as clicks and pops and dropouts in the audio while trying to record music tracks. That is not good.

Other mobo's are showing the low DPC latency results shown in that one post with VERY little DPC latency happening in an idle system. The Gigabyte boards are showing horrendous DPC latency problems and the users reporting the problems are also having lots of audio performance loss, compared to other mobo's with the same CPU. Furthermore, the beta bios that has been sent to a few guys has proven to eliminate the DPC latency spikes and drastically increase the performance for audio workstation use. Unfortunately, not everyone is able to get this beta bios for all Gigabyte boards having the problem.....yet. And as well it still has some bugs in it. So its not over yet.

At this point, I have to say that the Gigabyte mobo's should not be recommended at all for Digital Audio Workstation applications unless Gigabyte officially corrects the problems in their BIOS. I'm sad that I am one of the unlucky people to find this out after I already purchased my mobo, but I am hoping that these beta bios fixes will make it into a bug free official release and I will be able to march forward without a problem. Otherwise, I will swap out for another known mobo that does not have this issue, likely Asus. It makes no difference to me who owns what companies, the Asus boards are not having this problem now, Gigabyte boards are.

The one thing I like about my GA-P35-DS4 is that it has an LPT port, which almost no new gen mobo's have anymore. And I really need that. So for that reason alone I am hanging on to see if we can get this issue resolved. But make no mistake, this is a very real issue. Anyone dealing with real time intensive applications would run into this problem.

[Lsdmeasap]

Re: P35 Motherboards and DPC Latency Spiking

I am sure since they have fixed it in the BETA BIOS you have they will get it fixed fully and then release it to the public

Hope so for you users who are having these issues, but from what you say they seemed to have found the problem, and will likely soon release a fully functional BIOS for you guys, and everyone I guess since my board seems to have the issue as well now that I have ran that app

[Dewdman42]

Re: P35 Motherboards and DPC Latency Spiking

Hopefully you are right. I want to keep my gigabyte board. I really do. Its the nicest motherboard I have ever worked with to date, not counting this DPC issue. One of the things that someone commented through all of this was that they fixed this problem by recompiling some "old code", whatever that means. That implies that they didn't really figure out the problem per say, but just went back to an old release that was known to not have the problem. That's fine, except that this beta bios does not have bug fixes and such that are in the "new code", if I understand the situation correctly. They would need to back port all of the various feature enhancements and bug fixes that went into all of the releases since the old code, back into the old code base, in order to have a true release.... OR....better.....figure out exactly why the problem is occurring and fix the new code.

There is one known bug in this beta bios for example, in that you can't change the vcore. So overclocking is not technically possible with this beta bios. Though some people are saying that if they set the vcore with a production bios, then change the bios to the beta bios, the vcore sticks to whatever it was set at with the production bios.

Anyway, we are very happy to see Gigabyte working on this and will await in hope that they follow through. Apparently there were similar posts related to this issue about a year ago, but it never got to production. Maybe we are making enough noise this time that it will be fixed. I really hope so.

I actually am not sure I have a beta bios yet or not. Janus sent me an F12, but its not clear to me that its any different from the F12 that is posted on the gigabyte website. I haven't tried it yet. Someone else got an F13b from some Tech support source that was for the GA-P35-DS4(rev 2.1). That has been tested by him with the results mentioned earlier. I am on rev2.0 and not sure if I can use this particular beta bios or not, but that is probably the closest one to try.

In any case, I will feel a lot better when the whole thing is fixed in the official release, using the new code rolling forward and Gigabyte gives us some indication that they know what the problem is and have solved it. I will give them a few months to figure it out. After that, I will have to switch to another mobo if they haven't figured it out by then.

[badonkadonk]

Re: P35 Motherboards and DPC Latency Spiking

Hmm, i mentioned my suspicion about AHCI bios being the culprit, well it seems Asus have completely removed AHCI as a bios option on new p5k boards as of very recently.

What ever the issue is, i have spent 2 months head butting the walls trying to find out what the heck is wrong with my PC. Its only in the last week i learned about this DPC issue.

I guess i expect better from a top of the range product like the X38, not to mention the several hundred i have spent on different ram, PSU and other parts that i didnt need to spend. I just want a PC that works properly i dont care who makes it.

Can i suggest to any guys that are just discovering this problem to email Gigabyte Tech support with a screenshot of DPC latency checker. It looks like they are working on it but if more people report the problem then they might look into it more seriously.

Oh and i dont know if i mentioned it but if you look in Vista Event viewer performance diagnostics logs, you can see a bunch of errors , warnings about system degredation, boot degradation and various other anomolies caused by this issue.

Not to be confused with the normal event viewer make sure you look under "applications and services logs"

They are hard to find but open Event Viewer, then on the left open "applications and services logs" then "microsoft/ windows/diagnostics-performance/operational.

Hopefully Gigabyte can sort it out soon.

Hmmm i wonder if all this complaining about Vista being slow and unresponsive etc is really caused by this issue?

[AiRiCk]

Re: P35 Motherboards and DPC Latency Spiking

Sad to say that X48-DQ6 has the same DPC Latency problems (F6 Bios here). Crackles and stuff all happening at latencies my previous Asus NCCH-DL could handle without any problems. whadda you mean by UPGRADE ?

Hope Gigabyte sorts this out quickly..very quickly..

[lelcuk]

Re: P35 Motherboards and DPC Latency Spiking

Another GA-X48-DQ6 F6 BIOS DPC issue here. DPC latency was over 16000 with the supplied intel ich9 drivers, after upgrading to the latest intel driver latency fell back to "normal" spikes at about 2500 usec. Disabling ich9 drivers lower it under 2000 but not under 1000. Might be another issue with my nVidia 9800GT card (running latest rel driver). Overall super annoying ordeal considering the outrages price for this board (needed one in express this was the only one in stock).

Thats is for now, keep you posted

[Dewdman42]

Re: P35 Motherboards and DPC Latency Spiking

Just to give you guys an idea, here are some screen shots from running DPC latency checker on my 5 year old P4 2.53ghz machine. This is a slow old dog compared to my new Gigabyte-based quad. Its an Intel P4PE mobo. Some unused ports are turned off in bios(built in sound, serial ports, firewire), but otherwise, its a normal XP Pro SP3 install, nothing fancy done. There are two PCI cards(Matrox P650 and MOTU PCI-424 for audio. There is also an LPT based motu midi interface plugged in, but I was not playing on it for this test. Mouse and keyboard and plugged in via USB, through the USB hub in my monitor. It also has the mobo raid interface enabled in addition to standard IDE.

The first screen shot is idle.

The second screen shot is when I am moving the mouse around constantly, which seems to have the greatest effect on DPC latency here.

The 3rd screen shot is when I am playing back a Sonar project with an audio buffer size of 256, and video playback also happening at the same time.







As you can see, nothing even passes 500 in terms of latency and when I am not moving the mouse around it doesn't go much past 150.

Getting values over 2000, in my mind means something needs to be fixed. Even over 1000 is questionable. There is a reason the 1000 line is colored yellow and the 2000 line is colored red. See the other screenshots from people with more modern mobos getting much lower latencies than I am showing here. That is possible. All kinds of things can cause large latencies to happen, including mis configured systems, crappy 3rd party device drivers, etc. However, we are seeing here a very consistent trend that the Gigabyte mobo users are seeing really large latencies and people with other boards are consistently bragging about how they are not seeing the latencies...

I'm not meaning to slam on gigabyte here(other than perhaps pouring a little gas on the fire), but more I am just pointing to your 2500 and 1000 reported latencies and suggesting, they are way too high for you. Keep looking for better performance.