Help with G.SKILL DDR500 (barely overclocked) and Prime95

[bebop06]

Yesterday I got my Opteron 165 up to 2.52 at 1.375V, 24 hours dual Prime stable. I have a DFI Lanparty nF4 Ultra-D, and I'm trying to run my 2x1 GB G.SKILL DDR500 with a 188 (9/10) divider, which puts it at DDR504 at the stock timings (3-4-4-8, T1). It passed one hour worth of tests in MemTest (3+ passes of all tests), but is having a problem with dual Prime with large FFTs (or small FTTs--whichever is labeled "stresses RAM"). If I set each instance to one core and hit the Core 0 instance about half a second before the Core 1, I get an error on Core 0 after exactly 6 seconds each time. If I start Core 1 the half-second ahead, I don't get an error (at least after a few minutes). I tried running at the BIOS' default volts of 2.6, then down to 2.5 (which I heard the G.SKILL like), then up to 2.7, and even 2.8. All give me the same error.

The BIOS' reading of the DRAM voltage is always .6v above what I set it at--which seems odd--and MBM5 backs this up. It seems that the RAM idles at .6v above the voltage I set, and then approaches the correct voltage under load (according to MBM5). Not sure if this is relevant.

Also, when I first installed everything the other day, the BIOS began freezing up randomly (before overclocking). I tried flashing the BIOS, but would still freeze up. I swapped out the G.SKILL and replaced them with my old 2x512MB Corsair sticks, and the freezing stopped. After I put the G.SKILL back in, the freezing didn't come back. After this I overclocked my Opteron, and the BIOS remained fine. But when I changed the DRAM voltage to 2.5, my BIOS froze up again. Not sure if this is relevant, either.

Thanks in advance for your help!

 

[Ricemarine]

Try running your g.skill ram at stock speeds, you might have a faulty stick possibly (or 1 stick probably can't go up to 250 Mhz)..

 

[buzzsaw13]

Try running Memtest longer than an hour

 

[nealh]

I suspect this related ot vcore not ram

First test the ram at 250mhz in memtest86...set a lower multiplier 8 x250..run memtest for 6-8 hrs to confirm no bad sticks

Next what bios are you using?, what slots are the ram in?

I will assume memtest 86 is ok

when OCing the opterons the mem. controller is on the cpu and it may not be able to run mem. at 250 mhz at 1.38 vcore..bump to 1.4v bios setting..ignore winxp readings

what you set in the bios is what you get at idle and at load board overvolts about 0.03v

I have opterons run 15 small fft dual prime then fail large fft at same settings..needs more vcore

from what I hear Gskill is not lover of vdimm so stay at 2.6v to 2.7v

 

[bebop06]

nealh:

As of now, MemTest has been running for three hours (4 passes) with no problems. I'll leave it on for another 5 hours just to make sure.

When I first set things up, I flashed to the 11/24/2005 BIOS to see if that would stop the freezing. When it didn't, I flashed to the 6/23/2005 BIOS, and that's what I'm currently using. The people at DFI-Street said the only difference between the two was support for an FX-60, and recommended I stay with the 6/23/2005.

I knew that the memory controller is on the CPU, but I didn't realize that it needed more current to run faster RAM speeds. How many volts does the memory controller usually require for running at 250 MHz? (I guess I'm asking if I might have to bump up vcore past 1.4. At what point can I tell that the memory controller is getting as much juice as it wants to?)

Thanks so much for the reply.

 

[5t3v0]

bebop, I have the same ram and can confirm that these do not like higher volts. I found 2.6v gave the best results. Have you tried doing as nealh suggests & setting cpu multi lower to test if this is a ram or cpu issue. I would suspect that to get the 165 stable at 2.5GHz, will take 1.4-1.5v vcore. My 144 needs 1.55v.

Also forget memtest86. It is totally inferior to Prime95. The blend test is the one that stresses ram the most.

I also run my cpu at 9x280 but dont have the 183 divider option so run the ram with the 166 divider at 229MHz on 3-3-3-6. I can supply my timings if you're interested.

 

[nealh]

my point with memtest was not to test stablity but confirm the ram is not defective at 250mhz

Dual Prime 95 will test stablity..once we know the ram is ok..now we need to check the cpu

1. small fft to test your overclock ..165@2.5 may be ok at 1.4v..at least run it for few hrs

large ftt will stress cpu/ram and creates alot of heat..so it often needs more vcore

2.test large fft...if small fft is ok and large ftt fails usually this says you need more vcore to run cpu and ram at given speed or you need to loosen memory timings

3. blend is an excellent test of system, cpu and ram stablity...

I have found that if I can run small and large fft for 12hrs..I have no issues gaming, encoding etc

I know many will say you need 24 hrs of blend to be stable

You need Prime95 to be stable as long as you feel necessary for your needs..I personally would accept no less than 8hrs ..small and large fft

Remember so many factors will affect this..vcore, vdimm,cpu speed, ram speed,winxp, dll's..vidoe/sound, software programs..etc

the later statement is why I do have same rigid feel as some..24hrs of blend is great but any little software crap can crash stuff

You will see many say but if Prime 95 fails it is not stable...

I will reiterate a statement I made before on this..I have 2 PCs at different office thatdo "mission critical" work built by a company that sets up the system and configs hardware via DELL...nonOCed..they crash all the time, run stock and "tweaked " to run Win2000 and optimized for certain software....guess what there is no way they pass Prime95( 100% sure)...I have done 0 tinkering as I want no chance I am the reason....

So you have to set your parameters and stick to it and be happy

So have found P95 stable but the system crashes in games....so knows what is best....

I have found the 623-2 bios works best with my crucial ram(OCZ ram tech said they found 623-2 bios and yellow slots best for most ram types..ie BH5, TCCD/5, UCC etc)..I am getting some GSkill as well DDR500..monday..hoping this will work better for..kinda stupid purchase but could not pass up the price

Also you may need to switch ram slots from orange to yellow or vice versa....

Many have found orange to work better than yellow...

623-2 or 623-3 bios are very good

 

[abaez]

I have this mem and I can't even boot at 200 mhz stock timings. I am 3, 5, 5, 10 at 185 mhz. Definitely got some bad sticks. Did you order from newegg?

 

[nealh]

Originally posted by: abaez
I have this mem and I can't even boot at 200 mhz stock timings. I am 3, 5, 5, 10 at 185 mhz. Definitely got some bad sticks. Did you order from newegg?

HMM..did you get yours from newegg..I ordered from them on Wed....

 

[bebop06]

OK, I don't think vcore is the issue. I ran MemTest for 8 hours without issue. I tried to run two instances of Prime95, one for each core, and I get the same error after 6 seconds when I begin both instances within one second of each other. I tried upping the vcore to 1.4 and then to 1.425, but I got the identical error after 6 seconds. I tried lowing the memory speed to 140 MHz (100 divider) with the vcore at 1.425--same error. But I looked at the log and it said this: "FATAL ERROR: Reading from temp file." Is it possible that both instances were trying to access the temp file simultaneously, and that that's the only problem? Because that would imply that my memory is fine...

Anyone know if I'm just being stupid in trying to run two simultaneous instances of Prime95 with large FFTs? Maybe that simply doesn't work.

Otherwise, what could the issue possibly be?

 

[bebop06]

OK I'm just a moron. When reading up on how to run dual Prime, I saw some people say you needed two folders, some say just hit the shortcut twice. I didn't bother with two folders, and small FTTs worked perfectly. With large FTTs I got this FATAL ERROR after 6 seconds.. and I just made two folders, and the error is gone!

And yes, I did order them from Newegg last Friday morning.

 

[abaez]

Yes I ordered them on wednesday nealh, it also looked like the package was opened and resealed.

 

[nealh]

Originally posted by: abaez
Yes I ordered them on wednesday nealh, it also looked like the package was opened and resealed.

Ugh...so the seal on thememory was broken on delivery?

 

[abaez]

*On* the memory no, but there is no seal on the package. The bottom part seemed to be like if someone opened and then closed it but didn't close it right. I'm RMA'ing no biggie but its pretty annoying.

 

[nealh]

Originally posted by: abaez
*On* the memory no, but there is no seal on the package. The bottom part seemed to be like if someone opened and then closed it but didn't close it right. I'm RMA'ing no biggie but its pretty annoying.

no seal on the package? if the ram was sealed it was never used..

 

[TecHNooB]

I'm having similar problems. Overclocks went to hell after I put in the 2GB G.Skills. I posted about it in this thread.

 

[nealh]

Originally posted by: TecHNooB
I'm having similar problems. Overclocks went to hell after I put in the 2GB G.Skills. I posted about it in this thread.

I posted in your thread..I think you need to realize the memory controller is have to deal with 2GB of ram ..so I would expect issues vs less ram..at least this is my opinion

 

[Blooz]

You may want to take a look at the info in this thread.
http://www.dfi-street.com/forum/showthread.php?t=41953

It's feedback from many DFI users on their results with the G.Skill (and other brands) using the UCCC chips. If you d/l the spreadsheet, it's got complete settings used by the posters.

It should give you a better idea of the voltages people are using with this ram.

 

[TheRyuu]

What kind of memory does the G.Skill use? the TCCD's? UCCC?

 

[5t3v0]

UCCC. Review here on madshrimps

 

[abaez]

I was finally able to get these working at 260mhz 3-4-3-7. Very nice but very very finicky. One small setting change will not boot them.

 

[5t3v0]

what were the rest of your settings out of interest? I may have another go at overclocking mine again.

I run mine at 3-3-3-6 on 229MHz but was able to set lots of other timings quite aggressively at this frequency. All my timings are posted here

 

[abaez]

2.6 volts
CAS latency control (tCL) 3
RAS# to CAS# delay (tRCD) 4
Min RAS# Active Timing (tRAS) 7
Row Precharge timing (tRP) 3
Row cycle time (tRC) 7
Row refresh cycle time (tRFC) 13
Row to row delay (tRRD) 2
Write recovery time (tWR) 2
Write to read delay (tWTR) 1
Read to write delay (tRTW) 3
Refresh period (tREF) 3684
Write CAS# latency (tWCL) 1
DRAM bank interleave enable
DQS skew control auto
DQS skew value 0
DRAM drive strength auto
DRAM data drive strength auto
Max Async Latency 7
**DRAM response time normal
Read Preamble time 5
Idle cycle limit 256
Dynamic counter enable
R/W queue bypass 16x
Bypass max 7x
32 byte granulation disable

 

[Zim]

Have a read thru this thread.

 

[burney]

u should atleast do 250 on stock voltage and timings ... thats why it is ddr 500