InfiniteMac OSx86

InfiniteMac OSx86 (http://infinitemac.com/forum.php)
-   Leopard 10.5 (http://infinitemac.com/forumdisplay.php?f=85)
-   -   [Solved] AMD Dual Core Timing Issue AKA Mouse Bug (http://infinitemac.com/showthread.php?t=734)

agrafuese 03-17-2008 11:37 PM

@voyn1x: no, i don't have SMB turned on.

@farcaller: it seems i recall the same keyboard issues.

@snow: next time instead of force-restarting, do apple+shift+Q and it will allow you to log out. then you can log back in with GUI working again.

ok here's a little report for now: i've had my system up for 10 hours now without reboot. i have ianxxx's suggested memory settings: cas latency @ 3, cycle time @ 18, row refresh @ 18, HT @ 3x. things are still running fine, however, i haven't tried stressing the system out again because i don't want it to freak out and give me the 10810 error once more (thus forcing me to reboot and start my long-term testing all over again). admittedly, most of these 10 hours were spent with the system being idle because i was asleep, so only about 2-3 hours of that time was me actually using the system. so, nothing intensive was going on. just a little bit of itunes, some web browsing, and some entourage email checking.

@ianxxx: not to be a pest, but for me a 7-hour stint still isn't enough. i've had this problem happen after many more hours than that. in fact, to me it seems like it may not even be an issue of time at all, but rather just the right combination of apps and usage. for example, i've had this issue happen after using Linotype Font Explorer (it's a great app for font organization, btw, if anyone is interested. google it.). this app, i believe, it highly RAM and CPU intensive, and i did notice once or twice that my system started acting up after it had been up for an hour or so. also, as i mentioned in a previous post, i used to run Tomato Torrent (not anymore though, because it sucks in general) and that also seemed to cause the problem, especially when i'd leave it running with torrents over night. i'll run Font Explorer now for a while and see if i can get the system to act up again.

now, here's a noob-ish question: what is this talk about "spawns"? can someone explain that to me? if it is what i think it is, then it makes sense because essentially Tomato torrent has many simultaneous processes going on (connecting to many peers, gathering data, etc.)...and Font Explorer also has this same thing going on because it has to process many different fonts (gathering info on the font, displaying the name of the font with the actual typeface, etc.).

ok that's all for now. sorry for the TLDR guys, hope you actually read all of this :)

Snow 03-17-2008 11:59 PM

I stopped reading at the @ sign bro, then I went back to reading when I saw my nick mentioned ;)

Thx for the suggestion on logging out.

So to add to this, I havent got any probs so far running my RAM at 3-4-4-8 and HTx3. Before this I was running my RAM at 2,5-3-3-7 and HTx5. And this time, with relaxing my RAM timings Leo does feel a bit more responsive overall. When this turns out well I think I will bump the HT back up to x5 cause I have a feeling it might only be the RAM timings.
Let's hope it stays this way. I'm willing to cope with the lower timings if it fixes the whole crash/mouse issues.

Thx all for the input and fingers crossed!

agrafuese 03-18-2008 12:46 AM

Just a little update: I've got music playing in iTunes, a large file transfer going on in Transmit, iChat open talking to someone, Safari with multiple windows, and Firefox just sitting idle (wow, it's not crashing!). I also just opened Font Explorer up and activated an additional 60 fonts (up to about 350 fonts all together now). If this stays stable for a while, I'll add a little more stress by opening up Photoshop and Illustrator. At that point, it will be a pretty typical working scenario for me, and therefore a good test for the problem to arise. Here's one question I have though to derail this whole memory issue:

I've never bought RAM for a "real" Mac before. So how does it work? Can you just go on newegg.com and buy some Kingston RAM or Corsair, or some other fairly "standard" RAM, or do you HAVE to buy it from an Apple store? The reason I ask is because what would happen if I stuck my RAM in a real Mac? Would these problems be happening? Furthermore, can you change memory timings on a real Mac? It seems to me that if these things aren't necessary for a real Mac that there could in fact be some kind of tweaks done to Zeph's release so that we don't have to worry about these memory timing settings. Also, this kind of safeguard MUST have been implemented in my old Tubgirl version of Tiger 10.4.10 because wouldn't it also need to have specific memory timings if this weren't the case?? Don't get me wrong (Ianxxx, lol), I'm not trying to kill your theories, but I am still skeptical because it seems really ridiculous for every AMD user to have to tinker around with his system. And what about Intel PC users? Are they worrying about this? Or does Kalyway take care of it somehow? I really think we need to push this in Zeph's direction and see what he has to say. We have different systems! There are enough of us here now that this can't be small potatoes. There must be a blanket solution for us all. Anyone with me on this?

agrafuese 03-18-2008 11:40 AM

damn, where did everyone go? anyone else have any news on this?
i was running for a solid 21 hours without rebooting. no crashes, no slowdowns. then, suddenly, safari started crashing out of nowhere. the first two times it crashed, i could start it back up, but then after the third, it gave me the dreaded -10810 error on all of my apps and nothing would start anymore. if anyone cares to look through it, here is my console log:

Code:

3/17/08 11:20:18 PM ReportCrash[796] Formulating crash report for process Safari[721]
3/17/08 11:20:19 PM ReportCrash[796] Saved crashreport to /Users/agrafuese/Library/Logs/CrashReporter/Safari_2008-03-17-232013_AMD-Mak.crash using uid: 501 gid: 20, euid: 501 egid: 20
3/17/08 11:20:19 PM com.apple.launchd[95] ([0x0-0x99099].com.apple.Safari[721]) Exited abnormally: Floating point exception
3/17/08 11:20:20 PM kernel CODE SIGNING: cs_invalid_page: p=798[ATSServer] clearing CS_VALID
3/17/08 11:20:36 PM /usr/sbin/ocspd[800] starting
3/17/08 11:24:19 PM ReportCrash[802] Formulating crash report for process Safari[799]
3/17/08 11:24:20 PM ReportCrash[802] Saved crashreport to /Users/agrafuese/Library/Logs/CrashReporter/Safari_2008-03-17-232417_AMD-Mak.crash using uid: 501 gid: 20, euid: 501 egid: 20
3/17/08 11:24:20 PM com.apple.launchd[95] ([0x0-0xad0ad].com.apple.Safari[799]) Exited abnormally: Floating point exception
3/17/08 11:24:21 PM kernel CODE SIGNING: cs_invalid_page: p=804[ATSServer] clearing CS_VALID
3/17/08 11:26:58 PM /usr/sbin/ocspd[809] starting
3/17/08 11:28:39 PM ReportCrash[813] Formulating crash report for process launchd[95]
3/17/08 11:28:39 PM ReportCrash[813] Saved crashreport to /Library/Logs/CrashReporter/launchd_2008-03-17-232839_AMD-Mak.crash using uid: 0 gid: 0, euid: 0 egid: 0
3/17/08 11:28:39 PM com.apple.launchd[1] (com.apple.launchd.peruser.501[95]) Exited abnormally: Floating point exception
3/17/08 11:28:39 PM kernel CODE SIGNING: cs_invalid_page: p=814[mdworker] clearing CS_VALID
3/17/08 11:28:42 PM ReportCrash[813] Formulating crash report for process Safari[807]
3/17/08 11:28:43 PM ReportCrash[813] Saved crashreport to /Library/Logs/CrashReporter/Safari_2008-03-17-232839_AMD-Mak.crash using uid: 0 gid: 0, euid: 0 egid: 0
3/17/08 11:28:53 PM /System/Library/CoreServices/Dock.app/Contents/MacOS/Dock[107] spawn_via_launchd() failed, errno=22 label=[0x0-0xb20b2].com.apple.Safari path=/Applications/Safari.app/Contents/MacOS/Safari flags=1

the good news is, it seems the slowdowns have disappeared for now. if i could choose, i suppose i much prefer this -10810 error to the slowdowns, but the truth of the matter is i have to reboot my system either way. i still haven't run that app that Ianxxx tried from the link i posted. maybe i should take the leap now :\

Snow 03-18-2008 12:40 PM

Argh bad news, I could say it's just Saf being a dick and that real macs also have this issue but since the error is rare I think we can still conclude it's AMD related (or at least OSX86)

I need to do some more testing to post updates on this matter m8.

I'll keep you updated.

Voyn1x 03-18-2008 03:05 PM

After a lot of searching, i have discovered that Apple ensures that all memory modules meet their timing requirements by not POSTing if incompatible memory is installed in the system and that it is very important that Mac users should only use memory that's specifically made for the Mac.

Because all Macs configure their memory timings based on the SPD that Apple specifies, all memory is configured to have the same timings regardless of what the modules are capable of - in other words, all Mac memory performs the same!

This would explain why Ianxxx suggestion of playing with the ram timings seems to work.

@Agrafuse: Open up terminal and type 'top'. This gives the terminal equivalent of Activity Monitor. Keep that open and when you start to get the -10810 error have a look to see how many processes are running, you should be running 55-80 on average. If the amount has sky rocketed you should be able to see the offending process.

Ianxxx 03-18-2008 03:15 PM

Hi everyone, I'm gonna throw something new in to the mix, my broadband got disconnected last night problems with provider, so I did a test, put my memory overclocked etc.
Its still running this morning, I left it on when I came out of the house and will be back to see whats happening around dinner, but I'm begining to think that this is in someway network related.
And that the memory is just like the cpus=1 which helps but is not the cure.
for info I am using a realtek 8169 lan card with the apple kext supplied on 10.5.2 rev 1 dvd.
what configs are others using?

Snow 03-18-2008 03:25 PM

Hey Lanxxx

Actualy I was thinking the same thing yesterday. Remember the crash issues with the net on Zephs rev1 release (first one)
Mind you, I'm not using the Nforce LAN with fixed forcedeth or whatever but I'm using the MArvel Yukon (other LAN port on my ASUS A8N SLI)

And btw, by overclocked do you mean higher speed and/or tighter timings?

Cause I'm still running my ram overclocked at 230 but with lower timings 3-4-4-8 (default)
Only the timings seem to affect the issue for me so far (I'm not jumping yet)

So for me I went from HTx5 FSB 230 RAM 2,5-3-3-7 (which I have always used for such a small OC) to HTx4 FSB 230 RAM 3-4-4-8

Hope that helps with info

Ianxxx 03-18-2008 04:29 PM

my ram is now set at htx3 fsb 450 3448

The more I think about it the more i think this is a network bug. I ran my system succesfully for over 8 hours no problems and "no internet".
So the question is "us with the bug, what do we have or do different to those without.
So question to anyone who doesn't have it, did you change anything network related since you installed leopard? Different IONetworkfamily.kext? Etc?
For those who do have it, are you like me no changes to network since installation. Also any thing else anyone can think of, I'm thinking this must be something really simple and really subtle as some don't and some do and there doesn't seem any apparent link, what are we all missing?

agrafuese 03-18-2008 11:07 PM

Yes! I woke up this morning and began to believe it was network related too! Here's why: I can almost 100% guarantee that this problem will occur when I've left my computer on overnight with torrents running (this time I was using BitRocket, before it was Tomato Torrent). I have a feeling it is also upstream-related because I only had one torrent downloading. It was only around 100MB, and it finished just before I went to bed. So performance was still good when it finished and started to seed. I have an auto-stop when ratio hits 2:0, so after I seeded 200MB, there could be no more peers connecting. So this problem must happen fast, with only a few peers connecting over a short period of time. I should also say that I have Little Snitch running, and I use it in demo mode (so it stops after 3 hours into the night). I have no idea if this has anything to do with the problem at all, but it's just some more info to throw at you all. Also, one last thing I've noticed is ALWAYS associated with this problem is this console message that repeatedly occurs throughout the night:

Code:

3/18/08 8:41:32 AM mDNSResponder[17] mDNSPlatformRawTime went backwards by 4 ticks; setting correction factor to 206737245
The number of "ticks" differs, and it seems that the higher the number, the worse my mouse movements and GUI will perform. I've had it go up to 15 or 20 in the past. Also, the more frequently this message occurs, the worse performance is too. For instance, right now the mouse and GUI are slow but not as bad as usual, and I noticed that the tick numbers are all around 3-4 and the message only occurred a total of 5 times at random instances throughout the night/morning.

...now, for the cherry on top...I found this little line in there as well:

Code:

3/18/08 9:11:38 AM kernel skgeosx ERR: ouputPacket() zero segments
skgeosx is the network driver I am using.

I am using the same network setup as Snow. Marvell Yukon 88E8001. I too have the nLan adapter, but I haven't been using it because I had problems with it in the past. I am running 10.5.2 rev1 right now, so I don't know if it would work better now with this release. Maybe I'll give it another try. What network adapter/driver are you guys using?