High clocks or tighter ram timings?

booyaka

booyaka

booyaka

Soldato
Joined
19 Jan 2006
Posts
16,870
Well - which is better for F@H??

Bumped my clock up to 3.33ghz on my quad but slightly slacker timings than i had at 3.15ghz on the ram.

initial findings seem to show a slightly slower frame time than i had at the lower speeds? only inital first 3 frames or so but all the same.

nothing else major running to slow things down - will leave over night and see what happens but just wondering if anyone knew??
 
not all WU in the same family take the same time to crunch - the 1760 I'm getting at work at present are giving my 1150ppd but has been as low as 980.

Triple that to account for Hz and cores and you get a range of 2940-3450.
That's 500ppd differance - give it a few WU and see what happens.

Alturnativlely Throtleing? or a lower/less syc'd memory divider :confused:
 
shadowscotland said:
not all WU in the same family take the same time to crunch - the 1760 I'm getting at work at present are giving my 1150ppd but has been as low as 980.

Triple that to account for Hz and cores and you get a range of 2940-3450.
That's 500ppd differance - give it a few WU and see what happens.

Alturnativlely Throtleing? or a lower/less syc'd memory divider :confused:
Click to expand...

that could be a point - previously my memory was running @ 525 mhz:eek::eek: now on a divider @ 462 mhz - hmmm - speed ain't everything then!!

Will have a play with the ram

Previously 3.15ghz, ram 525mhz and only needed 1.425v through cpu
currently - 3.33ghz ram 462 on divider and needs 1.5v through cpu

might just drop back to previous setup as it was hardly a slouch then!
 
I suspect RAM timings come into play more with the SMP WUs because they have to communicate among processes, and they use more memory to start with.
 
booyaka said:
Previously 3.15ghz, ram 525mhz and only needed 1.425v through cpu
currently - 3.33ghz ram 462 on divider and needs 1.5v through cpu
Click to expand...

:confused: why those timings

3.15 = 350x9 = 700 DDR2 (at 1:1)
3.33 = 370x9 = 740 DDR2 (at 1:1)
Good quality 667 CL4 (PC5300) ram will do these easily (my geil hits 850) or

3.15 = 394x8 = 788 DDR2 (at 1:1)
3.33 = 416x8 = 812 DDR2 (at 1:1)
value 800 CL5 or quality 800 CL4 (PC6400) ram also will do this easily (my new OCZ hits 1000)

Are you running 1050 and 924 with Quality PC6400?

3.15 = 350x9 = 700 DDR2 (at 1:1) - 1050 DDR2 (at 2:3) best divider but still poor (x1.5 ie ever 3rd cycle sync's :))
or to put it another way for every two cycles the cpu does the ram does three (CPU 0, 1, 2, 3, 4 = RAM 0, .667, 1.334, 2, 2.667, 3.334, 4)

3.33 = 370x9 = 740 DDR2 (at 1:1) - 924 DDR2 (at 4:5) don't use this one (x1.25 ie ever 5th sync :()
ie for every four cycles the cpu does the ram does five (CPU 0, 1, 2, 3, 4, 5 = RAM 0, .8, 1.6, 2.4, 3.2, 5 )

would be better to use 740 (370) with tighter timings at 1:1 :D (CPU 0, 1, 2, 3, 4 = RAM 0, 1, 2, 3, 4)

Only ever use 1:1 (1:2 if I've got great ram)

Edit: For what we do - only ever use 1:1 - booyaka try running at 3.33 with your ram at 740 (4-4-4-12-20) or (4-3-3-4-12-24)
but if it's like my OCZ then (4-4-4-16-28) will have to do.

At your speeds you'll be running (5-5-5-15-30) at best
 
Last edited:
jesus - there endeth the lesson!!!:p:p

Will have a look later - it's ballsitix PC6400 which i know will do 1050 with +0.4v. however last set died after about 4 months running this speed.

Current timings are very slack 5-5-5-18 i think as i was testing the cpu's limits - will bang them to 5-5-5-15 and 1:1 later and see if win SMP likes it or not!!

Shouldn't effect the overclock but cheers for that info - very very consise!!!;)
 
booyaka said:
Current timings are very slack 5-5-5-18 i think as i was testing the cpu's limits - will bang them to 5-5-5-15 and 1:1 later and see if win SMP likes it or not!!
Click to expand...

at 1:1 you should be on 4-4-4-12-24 with ballsitix PC6400 and volts around 2.0 (2.05 max) as will not even break a sweat at 740.
at 2.2v you might even get 4-3-3-12-20

Speed is not the goal with ram - max sync at minimum timings.
at 400fsb - 800 (4-3-3-12-20) is about the same as 1000 (5-5-5-15-30) and your ram will last longer / use less power / and keep case temps lower

booyaka said:
Shouldn't effect the overclock but cheers for that info - very very consise!!!;)
Click to expand...

But it will affect your ppd and SuperPi score :D
 
Last edited:
changed ram timings to 5-5-5-15 and 1:1 @ 740 to start with and all seems fine apart from losing a WU @ 60% when i started it up! damn!
 
Your for it Shadow!!!!:mad::mad:;););)

(only kidding sir!)

Nightmare!!!

Windows did it's updates in progress etc - need to restart. Closed both (crtl +c) win smp clients down to do a manual restart rather than a forced windows one and now both WU have error and finshed early!! 71% and 55% - damnation!

oh dear - back to square one again!

[20:46:42] Completed 190000 out of 500000 steps (38 percent)
[21:03:23] Writing local files
[21:03:23] Completed 195000 out of 500000 steps (39 percent)
[21:05:54] CoreStatus = 7B (123)
[21:05:54] Client-core communications error: ERROR 0x7b
[21:05:54] Deleting current work unit & continuing...

Folding@Home Client Shutdown at user request.

Folding@Home Client Shutdown.


quoted from

http://fahwiki.net/index.php/CoreStatus_codes

Error 7B

CoreStatus = 7B (123)
Client-core communications error: ERROR 0x7b
Deleting current work unit & continuing...

This is an unknown error from the fahlog of SMP clients. Because 0x7b is not defined in the F@h client or SMP fahcore, it is believed to be a Windows or SMPD/MPICH error code. Known causes are unstable systems from too much overclocking, changing network settings while the client is running, or stopping restarting the client. The error is also caused by Windows updates that need to restart your machine. See the List of Known Issues.

Folding-community: Error 0x7b
Folding-community: Kasson' post in Error 0x7b
Folding-community: List of known issues - SMP Windows Client

Here's one more possiblity:
(0X7B Problem with Windows SMP Client solved by MichaelO)
The 0X7B Error that many folks have gotten when using the Windows SMP client is not an error from the Folding Client, but rather from Windows itself according to the Folding Forums, Wiki, etc. I had increasingly started having this problem on one of my machines. This morning I was notified that my Acronis backups had failed in the Verify Step. Upon researching this error in the Acronis forums, it was suggested that the Acronis error was caused by a memory fault. I found this a little hard to swallow at first but after having 0X7B problems this morning after rebooting to try and solve my backup issue I decided to start changing the memory, and sure enough, once I replace my 2 x 1024 memory Pair the problems in both Acronis and the Folding Client disappeared.

Based on this, I would suggest that if you start getting errors on the folding client, you should first try testing your installed memory thoroughly. Windows XP is fairly forgiving about memory faults but Vista is more stringent, and the SMP Client, especially on a dual core is also not very forgiving once the memory starts to go bad.
 
Last edited:
It shouldn't be your memory (unless the 'extra' .4v killed it) as your below stock timings
Have you memtested it - might need to go to 4-4-4-12-28 (the last one is alway a nightmare too guess without seeing the pc in action)

booyaka said:
Known causes are unstable systems from too much overclocking, changing network settings while the client is running, or stopping restarting the client. The error is also caused by Windows updates that need to restart your machine. See the List of Known Issues.
Click to expand...

Turn off auto updates now (well set it to notify only) there's been a couple of network settings type updates of late and smp doesn't like changes when running.
 
Vista updating killed off my WinSMP unit the other night - I set it to notify only and didn't really think any more of it, windows had obviously restarted as it was back at log in screen so I just put it down to a messy shutdown of the client.

Not sure what message I got at the time though I'm barely overclocking at the moment so doubt it'd be a stability thing in my case :o
 
The only update I got this week was the Defender definations.


SP1 must be ahead of the game :D
 
shadowscotland said:
It shouldn't be your memory (unless the 'extra' .4v killed it) as your below stock timings
Have you memtested it - might need to go to 4-4-4-12-28 (the last one is alway a nightmare too guess without seeing the pc in action)



Turn off auto updates now (well set it to notify only) there's been a couple of network settings type updates of late and smp doesn't like changes when running.
Click to expand...

cheers.

decided to knock it back to 3.15ghz (from 3.33ghz) and tighten up the ram timings as its about 0.75 volts less throught the cpu for that speed and about 10 degrees less heat.

not really seen any major improvements at the moment from the ppd so back down by 200mhz or so.

Cheers for the help - will probably go mad next time i have a day off and try for 3.5ghz!!:p:p
 
You must log in or register to reply here.
Back
Top Bottom