Project: 2684 (Run 2, Clone 5, Gen 10)

[DrSpalding]

PantherX,
Yes, it is OCed. I will try again with everything set to stock later today to see if it all works fine there. I have not tried a different PCIe slot and probably could do so, but only as a last resort.

The PSU is an Antec CP850 and according to the total draw while fully loaded and OCed, my UPS says it is supplying ~420W to the PSU and accessories. This should be well within the headroom of the both the PSU and the voltage rails available, since I am only using one GPU. W/o the GPU F@H client running right now, it tells me 354W and as I said, that includes a few accessories and a 21" wide screen LCD.

The one thing I may do is reduce the machine from 12GB to 6GB. I have heard that using more memory makes it harder to keep stable at higher clocks and I don't need the extra 6GB if I am not running a Linux VM for folding any more.

[Grandpa_01]

What we need is memtest86 that runs a continues loop of FAH perhaps one of the coaders that have made some of the other tools could figure out a way to do it.

[bruce]

Return the memory to Corsair and tell them that you want RAM that runs reliably at the speeds on the sticker.

[B2K24]

Return the memory to Corsair and tell them that you want RAM that runs reliably at the speeds on the sticker.

I'm not up for RMA the RAM just because bigadv fails. Everything else is stable at 7-7-7-20. IBT 20 runs maxium settings, OCCT 2 hours, LinX 20 passes
not to mention I have never had a bluescreen.

Plus at 4.0 GhZ I can't select 1600 MhZ speed only 1531 is selectable. only if I run 4.2 GhZ 1600 is selectable.
So the error I experienced with the core was due to settings that were not correct and not bad hardware.

[B2K24]

What we need is memtest86 that runs a continues loop of FAH perhaps one of the coaders that have made some of the other tools could figure out a way to do it.

What we really need is an F@H stability client that runs bigadv for 0 credit as a tester or a way to run a dummy P2685 WU.
After applying an overclock or changing settings I would like to run an "emulated" P2684 or P2685 that does not download or upload to stanford servers and of course would be 0 points. That way they are not dealing with failed bigadv's or partial ones being turned in.

You can run 20 different software's but it all boils down to we want 100% bigadv stability so there SHOULD be a way to run the exact projects as tests without any interference to Stanford.

[7im]

You can save off any project you want and use it for testing later...

The next time you get one of those projects, make a copy of the entire fah folder. Then you can use that copy as a tester.

Reconfigure the copy to ignore deadlines, that way you can use that copy indefinitely.
Also configure it to prompt before connecting, no need to try to send back a WU you already completed, nor waste yours and Stanfords bandwidth.
Any time you want to set the WU to restart at frame 0, delete the results.dat file in the Work folder.

Test away.


P.S. For simplicity sake, and portability, I'd recommend downloading the v6.29 drop in binary (which is actually v6.30 with no MPICH install.bat requirement).

[B2K24]

Thanks very much 7im, I didn't know it was possible.
I plan to backup a P2684 and P2685 then if I plan to apply a greater overclock or tighten some timings I can make sure bigadv is stable.
Thanks for the useful info.

[Grandpa_01]

B2K24

Just curious but what are your specks of the box that you have had the memory problems with. If you are running an ASUS board most of them have problems with running tfrc memory setting at anything below 60 when pushing the OC allot of the ASUS boards need a setting of 88 to be 100% stable. I do not very many other boards than ASUS very often although I do once in a while so I don't know much about the other boards. But I do know that when pushing the OC most boards will not run quoted settings very often. The only board I can ever remember doing it was an nVidia board. Some will jump in here and say they can get there memory to overclock allot but I would challenge them to take there CPU to the max and give it a try. I can OC the hell out of memory if I only take any of my 920's to 4Ghz 1600Mhz 6-6-6-21-60-1T @ 1.66v, but take it to 4.2Ghz and I am damm lucky to get the dominator's to run at 1600Mhz 7-7-7-24-88-1T @ 1.66v. Rated speeds are pretty much for stock speeds and medium OC and are good start out points. But in reality most memory will not run at stated speeds when pushing the CPU past it realistic limits.

[B2K24]

ASUS P6X58D Premium
Intel Core i7 930
6GB Corsair Dominator TR3X6G1600C7D
cosair HX 1000 wt power supply
Nvidia GTX 480
Windows 7 X64 O/S


I have an ASUS P6X58D Premium and right now I'm pushing in i7 930 at 4 GhZ 191X21
now the thing is at 4GhZ in bios dram freq. 1531MHz can be selected not 1600 or 1603 rather. (I don't know exactly why this is)
1603MHz is only selectable if running 200X21 4.2 GHz.

I don't have memory problems I just had a bigadv fail on me and I assume to this point it was a memory setting issue.

Code: Select all

[00:58:28] CoreStatus = C0000029 (-1073741783)
[00:58:28] Client-core communications error: ERROR 0xc0000029

When I encountered the error I was running 191 BCLK X21 CPU ratio settings 1531 MHz DRAM and I changed first 4 timings 7-7-7-20
I believe this is where I went wrong because when everything is AUTO the board defaults to 9-9-9-24

Now when I attempt to run 4.2GHz 200X21 with all timings AUTO the board will default 7-7-7-20

So I'm almost sure the error was for me to set 7-7-7-20 at 4 GHz created the instability. I need a better understanding of how the settings work and how they are connected. Hell I should just push 4.2 GHz and will probably do so in a few days now that I understand I can test a bigadv without sending a crappy result to Stanford.

[Grandpa_01]

set it at 200 X 20 it will give you 1603Mhz set the timings manually to 7-7-7-24-88-1T @ 1.66v to start with and see what happens you can start lowering tfrc once you are sure it is stable drop from 88 to 82 etc. If you want to run at 4.2Ghz you will most likely have to run at 7-7-7-24-88-1T or 2T @ 1.66v or possibly higher clocks. I would suggest at least 1.33V loaded to the CPU if you can handle the heat. I have not been able to get any of mine 100% stable with less than 1.33V

[B2K24]

Thanks for the advice it's much appreciated.
Right now everything is rock stable and I'm in position to complete my very first bigadv
I'm at 78% and it will complete tomorrow. Here is my HFM benchmarks of P2684 I also run GPU3 so that steals some CPU cycles.

4.0GHz 9-9-9-24 1531 MHz

Code: Select all

 Project ID: 2684
 Core: GRO-A3
 Credit: 8955
 Frames: 100


 Name: SMP
 Path: C:\FAH\
 Number of Frames Observed: 78

 Min. Time / Frame : 00:44:50 - 20,515.7 PPD
 Avg. Time / Frame : 00:45:46 - 19,891.3 PPD
 Cur. Time / Frame : 00:45:06 - 20,171.4 PPD
 R3F. Time / Frame : 00:45:33 - 19,972.1 PPD
 All  Time / Frame : 00:45:46 - 19,877.6 PPD
 Eff. Time / Frame : 00:45:54 - 19,819.8 PPD

Now if I attempt to run 4.2 GHz 1603MHz dram 7-7-7-20 I'm curious how the benchmarks would be different.
I'm not sure if the TPF improvement would be worth the trouble and higher temps/volts for 24/7 usage.

[DrSpalding]

The -bigadv WU 2684 my machine is running should be done in about 1.5 hrs so I'll take it offline then and continue testing it. If it passes memtestg80 first and then IBT + GPU F@H client at stock speeds, I'll restore my OC (4.0GHz) and then add some to the timings of the memory to see if loosening them up has any effect and if not, bump up the cpu vcore and the memory controller (uncore) voltage a notch or two.

Heat has not been an issue with the cores staying at 80°C or less and the GPU at ~87°C or less. There were some GPU WUs during the winter than would get it to 90°C but recently, even with summer heat, 87°C is about the highest I see.

[B2K24]

I'm gonna shoot for 4.2 GHz and 7-7-7-24 in about 2 days. I will post the info here.

[DrSpalding]

My machine was stable at stock clocks and auto everything, so I put back my 4.0 GHz OC, but let the BIOS deal with the memory timings itself instead of manally setting the first five settings. They were substantially smaller than what I was using previously since the memory is running under-speed (1148 vs. 1333 MHz). It got a BSOD of 124 which points to either cpu vtt or vcore to be increased. I knew my vtt was on the limit of 1.15v which needs to be within .5v of the DIMM voltage of 1.65v, so I bumped it and it was good to go running a GPU F@H client plus IBT with 8 threads and 10GB RAM for several hours.

An added bonus is that the machine is making about 4-5% more PPD on the -bigadv 2684 WU and nVidia GPU WUs. I think it has been running ragged at 4.0 GHz since I moved up from 3.8.

I wish I had known that the BIOS would probe and set memory timings appropriately for any particular speed, it would have sped up and made the OC process far more pleasant!

[B2K24]

I got this info. off overclock.net thread concerning bsod

if it bosd's:
Errorcode #101 = bump vcore
Errorcode #124 = bump qpi

As far as auto settings go I think it depends all on the board. It seems though my P6X58D Premium sets them beautifully.
I remember on old DDR2 boards I messed with in the past you always had to manually set timings or performance would go down the tubes.
DDR3 however is very new to me and so many settings in bios it's absolutely crazy. The bandwidth of DDR3 is so high that if timings are not tightest as possible I don't really think it matters. It's not worth the insecurity of instability.

Check out all the settings in my bios
I don't run these settings it's just a template I C&P

Code: Select all

1st Information: 6-7-6-18-6-60-12-8-24-0
DRAM CAS# Latency......................[]
DRAM RAS# to CAS# Delay................[]
DRAM RAS# PRE Time.....................[]
DRAM RAS# ACT Time.....................[]
DRAM RAS# to RAS# Delay................[]
DRAM REF Cycle Time....................[]
DRAM WRITE Recovery Time...............[]
DRAM READ to PRE Time..................[]
DRAM FOUR ACT WIN Time.................[]
DRAM Back-To-Back CAS# Delay...........[]

2nd Information: 1N-61-61-62
DRAM Timing Mode.......................[1N]
DRAM Round Trip Latency on CHA.........[]
DRAM Round Trip Latency on CHB.........[]
DRAM Round Trip Latency on CHC.........[]

3rd Information: 8-8-18-8-8-8-7-6-4-7-7-4
DRAM WRITE To READ Delay (DD)..........[]
DRAM WRITE To READ Delay (DR)..........[]
DRAM WRITE To READ Delay (SR)..........[]
DRAM READ To WRITE Delay (DD)..........[]
DRAM READ To WRITE Delay (DR)..........[]
DRAM READ To WRITE Delay (SR)..........[]
DRAM READ To READ Delay (DD)...........[]
DRAM READ To READ Delay (DR)...........[]
DRAM READ To READ Delay (SR)...........[]
DRAM WRITE to WRITE Delay (DD).........[]
DRAM WRITE to WRITE Delay (DR).........[]
DRAM WRITE to WRITE Delay (SR).........[]