stressprime results for me new opteron 170 (not o/c'ed)

[hippy2981]

CPU testing, stressprime is good. thanks to Shimmishim for recommending it. i was able to put both cores on work with a 100% cpu utilization for the SMALL FFTs (kept prority 10 for both). the test ran for 8 hours with no errors and still runnin. i plan to let it run till 12 to 16 hours.

questions are:
what is that i should be lookin for apart from the succesful tests??
the vcore and temp?? anythin else??
also, should i running in blend mode too??

thanks!!

 

[hippy2981]

results: small FFTs, 18 hours, no errors, vcore=1.33v, 39C (idle 32C)

 

[Shimmishim]

that's good.

now up your speed!!!

try 10 x 240 at stock

 

[hippy2981]

that's on my list. i would do that sometime. i have not o/ced before. once the system gets stable @2ghz, i would read more on o/cing and then mod. for 10x240, i just need to change the MHz value to 240, that's it, right?? anything more that i should check?? like vcore??

as of now, i'm tryin to fix the problem with my GFX and dcfinal. i have posted that in a different thread. still waitin for replies .

 

[hippy2981]

results: large FFTs, 15 hours, no errors.
vdimm = 2.66v, chipset 45C;
vcore=1.33v, 39C

 

[Concillian]

There are multiple components you can overclock, most prominently, RAM and CPU.

10x240 with RAM at 166 divider will keep RAM speed at 200 MHz while CPU will be overclocked. This allows you to seperate where instabilities come from if you have issues.

For checking RAM, you would use say 6x240 so you know the CPU is okay below 2000 MHz.

You may also need to lower the HT multiplier. it's probably set to 5x, which results in 1000 MHz. for 240 it should be at 4x for 960MHz. This speed has a completely tiny impact in the overall performance, and provides headroom to increase the bus speed for higher overall CPU speeds.

The ballistix should be able to handle 240, but I would still test independently for CPU and RAM. Small FFT will give max heat load for CPU testing, but blend will do more testing of the RAM. I find that max stable Prime95 speed is still unstable in some apps. Once you find your max for P95, turn it down a few more MHz for stability. I like to say that you won't notice a 2-5% drop in CPU speed for stability, but even one random hang due to too high a speed is very, very noticeable.

 

[bjc112]

Originally posted by: hippy2981
CPU testing, stressprime is good. thanks to Shimmishim for recommending it. i was able to put both cores on work with a 100% cpu utilization for the SMALL FFTs (kept prority 10 for both). the test ran for 8 hours with no errors and still runnin. i plan to let it run till 12 to 16 hours.

questions are:
what is that i should be lookin for apart from the succesful tests??
the vcore and temp?? anythin else??
also, should i running in blend mode too??

thanks!!

Never tested a CPU @ stock speeds!

I go straight for blood..

Well at least straight for speed!

 

[hippy2981]

Originally posted by: bjc112

Never tested a CPU @ stock speeds!

I go straight for blood..

Well at least straight for speed!

hehehehe. agreed. i'm very new at this.

 

[hippy2981]

Originally posted by: Concillian
There are multiple components you can overclock, most prominently, RAM and CPU.

10x240 with RAM at 166 divider will keep RAM speed at 200 MHz while CPU will be overclocked. This allows you to seperate where instabilities come from if you have issues.

For checking RAM, you would use say 6x240 so you know the CPU is okay below 2000 MHz.

You may also need to lower the HT multiplier. it's probably set to 5x, which results in 1000 MHz. for 240 it should be at 4x for 960MHz. This speed has a completely tiny impact in the overall performance, and provides headroom to increase the bus speed for higher overall CPU speeds.

The ballistix should be able to handle 240, but I would still test independently for CPU and RAM. Small FFT will give max heat load for CPU testing, but blend will do more testing of the RAM. I find that max stable Prime95 speed is still unstable in some apps. Once you find your max for P95, turn it down a few more MHz for stability. I like to say that you won't notice a 2-5% drop in CPU speed for stability, but even one random hang due to too high a speed is very, very noticeable.

that's a nice piece of informative post. i really appreciate the help. but, as always, some dumb questions from me.

1.for CPU: with 240mhz, to keep the RAM at 200mhz, i'd have to use the DRAM Frequency to CPU Freq. Ratio as 6:5?? with x10. and the HT should be 240x4 = 960mhz. voltages would have to be uped??

2. for RAM: 240mhz, 240x8 = 1.92ghz. 1:1, so that RAM runs at higher speed. and HT's 240x4=960mhz. voltages would haveto be uped??


finally, 240mhz, x10 = 2.4ghz, RAM runs at 240mhz with 1:1, HT's x4 = 960.
OR 250mhz, x10 = 2.5ghz, RAM at 250mhz, HT's x4 = 1000mhz.

please correct me if i'm wrong!!

 

[Concillian]

Originally posted by: hippy2981
1.for CPU: with 240mhz, to keep the RAM at 200mhz, i'd have to use the DRAM Frequency to CPU Freq. Ratio as 6:5?? with x10. and the HT should be 240x4 = 960mhz. voltages would have to be uped??

Maybe voltage would have to be increased, maybe not. My proc runs 9x200 at default and 9x240 is fine with no voltage increase, in fact, 9x267 is fine with no voltage increase. This depends on your processor. This is what Prime95 is for, to tell you if your stable at your current voltage.

2. for RAM: 240mhz, 240x8 = 1.92ghz. 1:1, so that RAM runs at higher speed. and HT's 240x4=960mhz. voltages would haveto be uped??

Again voltage may or may not matter. I'm not up on Ballistix, but check out xtremesystems.org to see what other people are doing for their RAM. RAM will tend to have an optimum voltage, rather than higher almost always enabling better performance, as in CPUs.

finally, 240mhz, x10 = 2.4ghz, RAM runs at 240mhz with 1:1, HT's x4 = 960.
OR 250mhz, x10 = 2.5ghz, RAM at 250mhz, HT's x4 = 1000mhz.

Yeah, if you can get to 2500. Take steps, don't go straight for max OC, you have to go a little at a time. 200 --> 240 is usually a reasonably achievable OC on a 10x chip, but beyond that is more like uncharted territory because so much is dependent on the luck of the draw with what chip you get.

You got the idea though. Test RAM separate ... test CPU separate ... make sure HTT is not limiting ... then combine them for best performance.

 

[hippy2981]

kewl!! thanks.

btw, you had mentioned the following in the first post. so the "166 divider" is a typo?? i didnt understand that. but i read and figured that it should have been 6:5 divider. would i be correct??
"10x240 with RAM at 166 divider will keep RAM speed at 200 MHz while CPU will be overclocked. This allows you to seperate where instabilities come from if you have issues."

 

[hippy2981]

unfortunately, i wont be o/cin for atleast a month, due to the known issue with crucial ballistix 1gbx2 RAM. that way, if the sticks go bad in a month (which i have read so many reviews lately), i would be able to isolate the problem. . hope that explains the reason for testing parts at stock speeds .

 

[Concillian]

Originally posted by: hippy2981
kewl!! thanks.

btw, you had mentioned the following in the first post. so the "166 divider" is a typo?? i didnt understand that. but i read and figured that it should have been 6:5 divider. would i be correct??
"10x240 with RAM at 166 divider will keep RAM speed at 200 MHz while CPU will be overclocked. This allows you to seperate where instabilities come from if you have issues."

Most motherboards list ram divider in the BIOS in assuming a bus speed of 200 MHz.

166 MHz divider = 166.67/200 = 5/6

DFI uses both terminologies, but many boards just call it 166MHz or 133 MHz divider.

 

[hippy2981]

got it. thanks!!

 

[Zim]

Originally posted by: hippy2981

Originally posted by: bjc112

Never tested a CPU @ stock speeds!

I go straight for blood..

Well at least straight for speed!

hehehehe. agreed. i'm very new at this.

I bet you suck your M&M's too!

 

[hippy2981]

. well, i got some random reboots. and i have the **ballistix 1gbx2**. these sticks have become notorious to fail in about a month's time (reasons unknown). obviously, i have to test everything at stock if i have to RMA.

besides, i have a peculiar issue goin on too with one of them games. i'm on it quite some time now.