Phenom II 920 owners - What are your overclocking results?

[mordant80]

Just wow, I haven't been able to do 3.7 yet myself. I'm going to give it a try when I get home. I'll have to up the voltage for sure. I'm not sure how far I can take my fsb either but 267 is good atleast. I'm runnin a 13.5 muliplier though, so I'll see if I can make it run at 14 with a lower HT and more volts. May have to back off the FSB though. I'm going to have to figure out what my memory is capable of soon.

 

[johnnyfiive]

Alright, I have some clarity on Ganged and Unganged memory modes.
Ganged mode is 1x128 Bit Dual Channel and Unganged mode is 2x64 Bit Dual Channel. Ganged sets the memory controllers to work in unison, synchronously. Ganged mode has a queue that it lines up for when data is accessed. Unganged mode separates the controllers to work independently, asynchronously. Unganged mode is faster due to how it accesses data asynchronously.

To put it in simple terms, ganged mode is like doing a science project with a partner. Instead of you and your partner splitting tasks, you both work on the same piece of the project. Unganged mode would be the same, except you split your tasks. You work on one piece, he/she works on the other.

Back to computer terms. Ganged is better for single threaded applications and Unganged is best for multi-threaded applications and multi-tasking. In general, your going to see better results in unganged. Single threaded benchmarks will show a big difference in results when ganged mode is enabled, and vice versa. As for overclocking, I've yet to find out if ganged or unganged offers better stability. Kei may know this.

 

[mordant80]

Lol, great description... I'm thinking with the way I use my computer that unganged probably would be the better option for me. Would be interesting to know if one or the other offers better stability... if there is a difference I have a feeling it'd be neglegable.

 

[johnnyfiive]

Lol, great description... I'm thinking with the way I use my computer that unganged probably would be the better option for me. Would be interesting to know if one or the other offers better stability... if there is a difference I have a feeling it'd be neglegable.

I too would like to know about the stability part. I do know that Kei runs in ganged mode according to his screenshots so maybe stability is better in ganged? Either that, or Kei loves running single threaded apps?

 

[mordant80]

No tellin until he chimes in... I did see somewhere one person was able to run tighter timing in ganged mode vs unganged. Being that's just the results from one person, it's hard to tell. Geez, can't wait to get back home and screw with this thing some more.

 

[johnnyfiive]

Me too! I'm in the middle of random WebEx sessions with a vendor and I'm typing in here due to WebEx having a fit. I'm ready to go home!

 

[Kei]

Quick answer, run unganged (AMD recommends that as well) because the Phenom loves it due to how they made it.

The screenshots with the ganged memory for mine are only there because I was running SuperPi which will give me a tick or two faster time overall, but I never run ganged mode for anything including single threaded applications. The "gain" you get from running ganged is FAR to small to worry about especially if you look at the "GAIN" you get from running unganged in everything else. Btw.....unganged really does give huge massive ridiculous gains over ganged in anything multithreaded. Just go try to encode a video or something using ganged.......I'll wait because it's gonna be a while. After that do the samething in unganged and I won't have to wait at all. Unganged is Fast with a capital "F"

Stability wise unganged again is better.

Okay I said that'd be short and it kinda is (for me), now I need food and to shower to get out of here already.

Kei

(remember ganged=fail, unganged= will get your more women)

 

[mordant80]

Nice.. noone likes to fail and more women is always a plus.. Unganged it is.

 

[johnnyfiive]

lol, more woman!

 

[johnnyfiive]

Inching towards 3.8GHz.

1.505v in BIOS, NB voltage at 1.325v in BIOS.

CPU Validation: http://valid.canardpc.com/show_oc.php?id=505562

I'm surprised by the K9A2 and gskill ddrii 800 memory (4x1GB too!)... doing a decent taking these overclocks.

 

[Flyordie]

http://valid.canardpc.com/show_oc.php?id=505843
The current 24/7 setup.
DRAM= 2.2V
CPU= 1.3V

 

[Kei]

The AM3 chips have me curious now so I'm playing around with the X2 Regor and X3 Heka settings on my 920 to see what it takes to get them stable at 3.5Ghz as well. I already know that 3.5Ghz @ 1.456v is 100% stable on four cores though I think it could probably be a little bit lower (I did just run a quick 10 minute test at 1.44v and it passed).

Now I'm gonna play around with 2 and 3 cores only to see what I can come up with. I hope I can get some really great results just like with all four cores.

Kei

 

[johnnyfiive]

Kei, how are you disabling cores? Is it a BIOS feature on your motherboard or are you using AOD?

 

[ASharp]

Kei, how are you disabling cores? Is it a BIOS feature on your motherboard or are you using AOD?

If the M3A32-MVP's BIOS is like the M3A79-T then there should be an option in the CPU Configuration menu called Processor Downcore or something which allows you to select how many cores you want active.

Looks like this:

 

[johnnyfiive]

Ahhh I see. I sadly don't have that option with the MSI K9A2 790FX Plat. That is a nice feature to have if you have no need to utilize all four cores.

 

[Kei]

Yes, ASharp the M3A32-MVP has the same CPU Downcore option as the M3A79-T boards. It's extremely useful since there aren't too many times when you actually NEED to have all four cores running so you can save on power and heat just turning off 1 or 2 of them.

So far I'm still playing with the X3 Heka settings at 3.5Ghz and I just finished a few different stability tests with it.

Results

AOD Stability: Pass
Everest Stability : Pass
SuperPi 1M: Pass
Wprime: Pass

During those tests the system was using a total of 215-218 total watts which is sweet especially when you consider that's 100% at such a high clock!

Maybe I can lower the cpu voltage another tick and it will be stable, but we'll see in a bit. I'd like to see what the 3DMark06 score will be though before I do that.

Kei

Edit: 3DMark06 score added with my standard gpu 690-1175Mhz clock.

Validation (figure I should start posting these )
http://valid.canardpc.com/show_oc.php?id=506143

 

[Kei]

Ahhh I see. I suck cuz I don't have that option with the meh okay MSI K9A2 790FX Plat. That is an awesome super feature to have since you don't need to utilize all four cores all the time.

Fixed ^

---------------------------

Told you so....you need to get an ASUS awesome board.

Kei = win

 

[Kei]

Okay and now X2

The X3 Heka setting couldn't go any further down than 1.392v but I suspect it has something to do with the board and the voltage options as I couldn't get a good slot inbetween 1.392 and 1.376v which caused a fluctuation so it wouldn't pass stability. It would run just fine but once at 100% load it would fluctuate and then kill it.

Anyway I've been testing the X2 Regor setting for a bit now and it's passed the best run stable at an awesome setting of...

3.5Ghz @ 1.312v

No problems with any stability tests and it's run everything I've thrown at it. I'll run 3DMark06 to see what the score would be with my usual gpu overclock as well. I'm very proud that it was able to keep the 3.5Ghz clock with only 1.312v!

Total system power consumption at 100% load for the X2 setting was only at 185-188 watts which is awesome for a clock speed that high! Temps well.....don't bother because I don't remember them even reaching 32C during 100% load. It was between 30-31C the whole test.

During both the X2 and X3 tests I left the fan as I run it every day on "Silent" mode so it's using roughly 570-670rpm tops and stays normally in the 600rpm range even during the stress test at 100%. I didn't bother to test at max speed (2000rpm 110cfm) because it'd be pretty pointless and very loud.

Kei

Edit: 3DMark06 score added with usual gpu overclock of 690-1175Mhz
Validation
http://valid.canardpc.com/show_oc.php?id=506227

 

[Ryo129]

3.36mhz 1.45v
stable at 7+ hours in prime95

validation:
http://valid.canardpc.com/show_oc.php?id=506233

 

[mordant80]

Woohoo! After I dropped the HT Link, I decided to give the 14x multiplier another shot, and it not only booted but made it through a 3dmark bench. Didn't have to raise the voltage anymore either. I kinda doubt it's prime stable without raising the voltage some, but here's where I'm at now. I love this F'n chip!

 

[Kei]

Final results of 3.5Ghz X2/3/4 test

Okay so the final results for the X2, X3, and X4 settings when all clocked with identical 3.5Ghz cpu clocks and other settings are as follows...

X4
3.5Ghz @ 1.440v
Wprime 32 - 11.000 sec
3DMark06 - 15,196

Total System consumption 250-253 watts

X3
3.5Ghz @ 1.392v

Wprime 32 - 14.495 sec
3DMark06 - 14,563

Total System consumption 215-218 watts

X2
3.5Ghz @ 1.312v

Wprime 32 - 21.343
3DMark06 - 13,178

Total System consumption 185-188 watts


All of these results are at the exact same settings the only setting changed was the amount of cores powered up for use. No tweaks or fine tuning was done so the scores are of course not optimized and would be higher if I was to do so. Total System consumption readings are of the TOTAL system power usage during 100% cpu load stress testing taken at the wall outlet not just the cpu usage which is far far lower.

Attatched is the cpu-z readout with all specs used for each test. All system settings shown in that shot were identical for each test with the exception of the amount of cores activated (shot shows 2).

To me it looks like either way you go, you win.

Kei

 

[mordant80]

Well I couldn't get it running prime stable at the speed I last posted. Took the voltage as high as 1.45, not sure if I really wanna go any higher than that, my goal is to find my highest that I can run stable 24/7. I can leave the multiplier at 14 and back the FSB down to see how high it will actually go and be stable. Only thing I'm not sure of is whether it would be better to run my memory in 800 mode with tighter timings which would put it somewhere between 800 and 1066, or if I should try to run it in 1066 mode with looser timings having it run over 1066. May just have to try both and test it to see which gives better performance. Another option would be to set the multiplier to 13.5 and try to raise the FSB some more, probly would have to loosen memory timings there as well... so many options. Any thoughts on which direction I should go? I probly should find out how far my FSB can even go huh?

 

[johnnyfiive]

Well I couldn't get it running prime stable at the speed I last posted. Took the voltage as high as 1.45, not sure if I really wanna go any higher than that, my goal is to find my highest that I can run stable 24/7. I can leave the multiplier at 14 and back the FSB down to see how high it will actually go and be stable. Only thing I'm not sure of is whether it would be better to run my memory in 800 mode with tighter timings which would put it somewhere between 800 and 1066, or if I should try to run it in 1066 mode with looser timings having it run over 1066. May just have to try both and test it to see which gives better performance. Another option would be to set the multiplier to 13.5 and try to raise the FSB some more, probly would have to loosen memory timings there as well... so many options. Any thoughts on which direction I should go? I probly should find out how far my FSB can even go huh?

Definitely find out how far your FSB will go. That will ultimately tell you how far you can actually go. Lower the NB, HT, and CPU multipliers and start upping that FSB until it no longer boots.

 

[johnnyfiive]

Come on 920 brethren! Lets see those results!

 

[peste19]

mordant80

how is the motherboard? i am planning on buying the same one you got. does it have an option to change the timmings of the ram in the bios?

thanks