Sign in with
Sign up | Sign in
Your question
Solved

Prime95 rounding errors at stock settings

Tags:
  • Homebuilt
  • Systems
  • Product
Last response: in Systems
Share
September 27, 2010 1:20:20 PM

I'm running:
Phenom II 965
GA-890xa-UD3
2x Corsair 2gb CMX4GX3M2A1600C9
Stock CPU cooler and pre-applied paste
Palit GeForce GTX 460 1gb
XFX 650w
WD 640GB 6 gb/s SATA
Windows 7 64
Antec 300 (two fans)

This is my first build.

This is a follow up to this post: http://www.tomshardware.com/forum/280534-30-890xa-2x2gb...

Ok, so we're back to having prime95 failures.

Since I last posted, I've done the following:

Changed RAM timing to 8 8 8 20 (still at 1333mhz)

Noticed I was failing prime95 again and so went back to testing RAM

Tried 8 8 8 24 timing:

Still failed prime95

stick#2 failed memtest, shortly after starting, while testing it by itself (memtest error: 0003efdb658 1007.8mb (good) dfdfdfdf (bad) dedfdfdf)

Stick#1 passed with those timing settings, but I only tested it for one pass.

Changed timings back to 9 9 9 24 RAM (still at 1333mhz)

Tested with memtest86+, with both sticks in dual channel mode, 4 passes (about 3 hours?) no errors


First prime95 run: Failed after 1 hour (almost exactly) with '0.5, expected no less than 0.4' error

Restart comp

Second prime95 run: Failed in less than 1 minute with '0.5, expected no less than 0.4' error.

Third prime95 run: (started immediately after second run failure- no restart) ran 4 hours (almost exactly) before worker 1 fail with '0.5, expected no less than 0.4' error

Tried upping cpu voltage .1 volts and running prime95:

Failed (don't remember the time) and was running a little hot (up to 59C)

Tried underclocking cpu, HT, and NB:

Failed prime95 in a few seconds.

I'm running cpu pll voltage: Normal (2.500v)
DRAM voltage: 1.660v (default was 1.5, which we established was too low for this RAM)
NB Voltage Control: 1.170v (Normal is 1.150)
HT Link Voltage control: Normal (1.200v)
NB/PCIe/PLL Voltage control: Normal (1.800v)
CPU voltage: Normal (1.3250v)

Ram memory clock: 1333mhz
Ram timings: 9 9 9 24

Any more advice guys? The randomness of the prime95 failures has me totally confused.

I had a delay on getting all my parts, so its been about two weeks since I placed my order and I'm getting a little antsy about possibly needing to RMA before Newegg's 30 day limit runs out.

More about : prime95 rounding errors stock settings

a b B Homebuilt system
September 27, 2010 2:48:27 PM

I hate to ask, but why aren't you troubleshooting your components at stock speeds until you stabilze the system? I highly recommend you run all at default first. If the system fails at stock, default speeds I would RMA the memory at a minimum.
m
0
l
September 27, 2010 3:45:18 PM

To explain: My initial problem was that my comp wouldn't get to windows or run stable in windows with RAM in Dual Channel mode (both sticks in slots that share the same channel). Running one stick or each stick in a seperate channel worked fine.

I was told to check my RAM voltage and that seemed to solve my issues.

I had trouble initially with the BIOS defaults because my MB default DRAM voltage was 1.54v (my RAM needs 1.65v).

It was then suggested that I change my RAM timing to 8 8 8 20, as that was supported at 1333mhz.

The only other thing changed is the NB voltage, which I upped one tick, which made the stability better- it didn't fail as quickly in Prime95.

Long story short, after further testing, I still have instability problems running prime95 and have come all the way back to setting my ram timings to 9 9 9 24, which has improved stability, but my system still isn't stable.

I'm currently running at stock speed, with the memory timing a little more loose than it should need to be for the clock speed its at (1333mhz)

The ram is 2x Corsair 2GB 1600 9 9 9 24 (That's whats on the sticker on the ram stick itself)
m
0
l
Related resources
September 27, 2010 3:52:57 PM

I've currently got both sticks of ram on separate channels (NOT in Dual Channel mode) and am running Prime95 again.

If it passes 8 hours like this in Prime95 (meaning Prime95 only fails when I run it with RAM in Dual Channel mode) should I RMA the RAM or the MB or both?

Anything else to be concerned about?

I'm running at stocks speeds.
m
0
l
a b B Homebuilt system
September 27, 2010 4:06:28 PM

There are a couple of possibilities here. One your RAM may be defective, seems from your first posting on this thread that one stick passed at default speeds and the other did not. There could also be a problem with the motherboard/memory controller sub-system.

Did you try the RAM at 9-9-9-24, 1.5v, 1333Mhz (these seems to be Corsair's settings for slower 1333mhz RAM)? Your previous testing fails at any level of OCing/tweaking it seems. I would also set the NB back to default when checking this.

Are you OCing/tweaking your CPU or GPU at all? If so, go back to default and see what happens.

If you replace the memory, I suggest the following:
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Am using in a very similar config to yours and have zero problems.

Later.
m
0
l
September 27, 2010 4:32:15 PM

Ok, I'll try my RAM and power settings at the settings you suggested.

I'm not overclocking or tweaking *anything.* I'm just trying to get my new system to run stable. This is my first time building a computer. ;) 

EDIT: Ran Prime95 for 3.5 hours with ram in different channels (NOT in Dual Channel mode) and it ran stable.

Testing now in Dual Channel, 1333mhz, 1.5v, 9 9 9 24.
m
0
l
a b B Homebuilt system
September 27, 2010 4:39:28 PM

Understood, but you have been changing mobo/NB/RAM settings and that, in effect, is OCing. The previous suggestion I sent effectively underclocks your RAM to be the same as Corsair's 1333mhz offerings to see if that will stabilize your config. If it does, then I would RMA the RAM since it won't run reliably at 1600 settings (and wasn't what you paid for, albeit only a couple more $$$).
m
0
l
September 27, 2010 7:46:42 PM

Oh, sorry about that. I guess I haven't learned all the terminology correctly yet. :D  Thanks for the clarification.

3 hours running prime95 so far and stable at the settings you provided.
m
0
l
a b B Homebuilt system
September 27, 2010 7:50:10 PM

Aries144 said:
Oh, sorry about that. I guess I haven't learned all the terminology correctly yet. :D  Thanks for the clarification.

3 hours running prime95 so far and stable at the settings you provided.

Got my fingers crossed for you. Nothing worse than an unstable system.
m
0
l
September 28, 2010 3:11:07 PM

Thanks!

I realized I need to clarify something else: I never even tried to get it stable at 1600. The default was 1333 and that's all I had been playing with. (where's the forehead slap smiley?)

Ok, it ran 5 hours stable with prime95 at the settings you provided before I ended the test. (9 9 9 24)

I then ran it overnight (11 hours) with the following settings: 1333mhz, DRAM voltage: 1.5v, Timing set to 8 8 8 20 T1.

The only thing I can think of is setting the voltage to 1.52v manually, instead of using the auto selection or running at 1.66v, must have solved the stability issue at 1333.

I played around some more with the RAM set to 1600, but haven't gotten it stable yet.

This is what I've tried: 1600mhz, voltage 1.66, Timing set to 9 9 9 24 T2.

It ran for about an hour like this and then gave a rounding error.

1. Should I try setting the NB voltage up a notch from 1.15 to 1.17?

2. Any other settings I should play with to see if I can get it stable at 1600mhz?

3. Which is actually faster; 1600 at 9 9 9 24 T2 or 1333 at 8 8 8 20 T1?


Thanks again for all the help!
m
0
l

Best solution

a b B Homebuilt system
September 29, 2010 12:54:26 PM

It ran for about an hour like this and then gave a rounding error.

1. Should I try setting the NB voltage up a notch from 1.15 to 1.17?
COLGeek: No, I don't see really doing anything to help.

2. Any other settings I should play with to see if I can get it stable at 1600mhz?
COLGeek: If the straight stock settings for 1600 aren't stable, then the RAM likely has a problem and should be returned.

3. Which is actually faster; 1600 at 9 9 9 24 T2 or 1333 at 8 8 8 20 T1?
COLGeek: The difference is largely negligible from a user perspective. You could obviously research and find that the 1600 config will have higher theoretical throughput than the 1333 settings, but you need to look at total system performance to make an effective measure. Stability is WAY more important than raw potential speed.
Share
September 29, 2010 5:14:35 PM

OK, looks like its time to start the RMA process.

Thanks for all of your help!
m
0
l
September 29, 2010 5:15:02 PM

Best answer selected by Aries144.
m
0
l
September 29, 2010 6:21:30 PM

Or not.

Apparently, newegg.com won't RMA an item if you've removed the UPC barcode from the package. This is somewhat irritating, since you have to cut out the barcode, immediately upon receiving your product, and include it with other documents in order to claim a rebate. :\

At least it's stable in its current configuration. :\
m
0
l
a b B Homebuilt system
September 29, 2010 6:31:34 PM

That seems odd. I have purchased newegg stuff and returned in non-original packing materials (meaning no barcode). I suggest you call them, if you want to push the matter. Regardless, good luck!
m
0
l
a b B Homebuilt system
September 29, 2010 6:32:05 PM

try RMA direct to the maker?
m
0
l
!