Printable View
You guys have no idea how much help this thread has been! Thanks KTE...et all!!
Here's a screeny of what I consider decent...
http://i30.photobucket.com/albums/c3...burt/240HT.jpg
Check out the NB Speed though!! It almost seems to be tied to the HT link speed... :confused: Whats even stranger is I can't change the HT Multi in AOD either (not sure about bios)...
My NB FID=2 in p-states, that should be 9x right? Am I missing something? :shrug:
@Daveburt714
If your NB multi is at 6x then max HTT multi will be 6x because HTT can't be faster than NB....
My Phenom was benchable @240HTT, but that was far from stable....
With time I'm finding that only below 2.38GHz system is fully stable under x64 OS...
What's funny is that even if I boot @2.53GHz or 2.6GHz I can run Orthos, Prime and AOD tester for hours with no errors. Only when gaming my system is hanging up. Best test so far - Crysis in CF. If that game runs, then anything runs.
For example @2.5GHz I had freezes after just 3-120seconds of playtime :p:
@2.43GHz I could play max. 3-15 minutes
@2.4GHz Crashed within 1h
@2.38GHz NEVER crashed!
:clap: to Crytek for developing very good stability tester :p:
I simply need new CPU ;)
You're very welcome. :)
Yep, Processor Performance State NB FID 02 = 6x multiplier (240x6=1440)
As I mentioned earlier, HT speed can only be equal or below NB speed. ;)
If you want to oc, throw CPU Voltage in BIOS up to ~1.45V.
For some reason, Phenom deosn't seem to scale with voltage. Around 1.35V seems to be the limit I'm seeing with most people and it gets you around 2.47GHz stable. I'm sure this is linked to the VID. We need higher VIDs.
A very good (doable for all I reckon) setting to try 24/7 is 220x11 @ 1.33V and keep NB FID 06, HT Link 1800MHz and RAM at 1:2.
Keep an eye on AOD CPU VDDC so that it stays ≥1.33V.
Did some more testing last night KTE, and NB seems to stay stable considerably higher if I use GANGED memory mode. Perhaps relieves stress on controller or NB?
CPU @ 2706Mhz | NB @ 2200Mhz | HT @ 1000Mhz | Mem 410Mhz (820mhz) 5,5,5,18,2T
Passed '06 no probs
http://service.futuremark.com/compare?3dm06=4449632
doesn't that make your memory run in single channel mode? does mine last time I checked. :S
Yes, it does. I noticed a small drop in mem performance but this could be counter-acted with faster memory.Quote:
Originally Posted by blargman
I'm gonna pop off the NB heatsink and try cooling with my waterblock to see if there's anymore headroom left in it.
Yep. It's quicker for single threaded access and slower for multi-threaded access. Overall=>slower. ;)Quote:
Originally Posted by blargman
MR.Q: Nice testing. One thing though; have you tried 2100/2200 in unganged mode with the same settings? Is that possible?
Since you booted high IMC clocks, you can get higher CPU volts too than anyone who can't. Running sub-2.1GHz on NB, I could never get plus 1.484V available for CPU but above 2.15GHz and higher CPU volts were offered. You have 1.568V applied, I've seen 3GHz 1M at those volts possible. Tried it again yet?
Don't try that one just yet. ;)Quote:
Originally Posted by MR.Q
The NB speed = IMC speed. It's within the CPU.
That NB onboard is just the chipset and does nothing for us concerned here. :)
So the only thing which can effect IMC speed is the a) CPU cooling b) IMC volts c) CPU volts.
just a quick question:
on AM2+ motherboards, the max NB mulitplier is 9...what about on AM2 boards?
Same... if it works. ;)Quote:
Originally Posted by mrcmtl
Too many of the AM2 MBs do not support Phenom yet or at least not properly so they give false clock readings (like 3GHz/4GHz=>8GHz when you're at stock). So you have no idea how to tell what the real NB speed/multi is at that time.
Only way to judge is to run benchmarks we know the scaling of.
Well, I started out at NB 9x, seems AOD was lowering the Multi while I was upping the HT clock... I'm not sure if its just my chip (0745 MPMW) or what, but I can't seem to get any kind of stability above 1850 on the NB...:(Quote:
Originally Posted by KTE
I see all these screen shots above 2k and my chip just won't do it reliably!!
http://i30.photobucket.com/albums/c3...t/230x11x8.jpg
AOD seemed to be a little quirky too, so I just started OC'ing through the Bios.... This is with an 8x Multi on the NB (1840), and all V's at stock. Hopefully this will be a stable place to come back too!! :D
One wierd thing I noticed was that upping the V's actually caused more instability on my machine! If I even bump the voltages I seriously doubt if it would boot with these settings... Wierd... :shrug:
Oh well, Good signs everywhere!! I've been running 4 instances of Prime, typing this message, using the clipboard and paint and uploading to photobucket!!! :D
Guess I better hurry up and post this Bad Boy!!!! :rofl:
That can't happen. :p:Quote:
Originally Posted by Daveburt714
On-the-fly NB multi change is not physically possible AFAIK. You might have set NB speed too high so it lowered it at bootup but that would have to be at bootup.
I've hardly seen many yet. Around 2-2.25GHz I've seen but not stable. Also you have to feed it more volts if you want high NB speed. Like above 1.4V.Quote:
I see all these screen shots above 2k and my chip just won't do it reliably!!
:confused:Quote:
One wierd thing I noticed was that upping the V's actually caused more instability on my machine! If I even bump the voltages I seriously doubt if it would boot with these settings... Wierd... :shrug:
Try it. Go up one notch on HT ref than what you have above, don't move volts in BIOS. Does it boot? If so stability test it (P95/game)
Then reboot but set high volts at the same MHz. Any difference? If not go up one notch in AOD on HT ref and retry above testing. ;)
Just got home from a party!!! Happy F'n New Years folks!!!
I'm sure you gave me some good advice KTE... I'll check it tommorow! Hic....... :wiggle:
I couldn't wait for the BE... But this isn't a bad chip for $239!! At the last settings I posted... It will definetly hold me over till the B3 stepping!
It's the 1st here since 9hrs35mins ;)
My last reply is exactly when I got back from some celebrations too. Happy new year mate!! :D
Ok... I'm sure your right... But I set NB FID to 2, per your instuctions for the handicapped guy.... :cool: :DQuote:
Originally Posted by KTE
Just look a couple messages up.. Mr.Q is over 2100 (granted thats Ganged)..:)Quote:
Originally Posted by KTE
I'm not done messing with stuff yet KTE, But a 2.53 Stable Phenom is working pretty good for me right now!! :up:
See the thread earlier. I ran 2.64GHz CPU & 2.16GHz IMC/HT including stress testing and benching, uptime was 24/7 for over 3 weeks, bootup values. But it still wasn't stable. After 4 weeks at that the chip would never boot that again. :(
Same with 2.85GHz CPU 2.07GHz IMC/HT.
Same with 2.75GHz CPU 2GHz IMC/HT.
Same with >2.5GHz CPU >2GHz IMC/HT.
Dual channel.
On one day max fully stable was 2.85GHz.
4 days after it was 2.3GHz.
3 days after it was 2.75GHz.
2 days after it was 2.3GHz.
2 days after it was 2.64GHz.
And so on.
Until it finally reached max stable (bootable) of ~2.5GHz.
Maximum fully stable running WCG at 100% load for 24/7 was 1.98GHz IMC/HT.
People on Gigabyte boards are getting better IMC clocks and stability. On the MSI board, not much. Its a hit and miss situation.
Pick a clock speed, stress test it, bench it, run it for 3-4 weeks, bootup, boot down, hibernate, standby and the you'll know if its stable. :)
Yes KTE i agree mine has also gone down from 2.72 to 2.53 and now i can only do 2.42 stable for gaming and general use. Stress testing at 2.53 is fine but forget playing a game like Crysis for more than a hour. What you think the cause of this decrease in performance. Happy New Year all.
Well, you were right KTE.... I left it to prime overnight and it didn't make it through the night. Then it started showing instability again today....
So I did bump the v's up one notch on the Vcore and NB, and it seems to be OK again... The wide performance swings you've seen are pretty odd, hope that doesn't happen to me!
Do you think it's related to the Mobo or the processor itself?
As for upping the V's I had them both cranked while testing and didn't see any real benefit from it as far as improving my OC...
Of course.... I could have gone a little too far the other way ;)
If only the bugger would run like this consistantly :(
http://service.futuremark.com/compare?3dm06=4478136
Wow Mr Q that is a awesome speed .Keep it up. I never got close to that speed.
I tried that auto overclock thing with my Canine A2 CF and I was able to boot into windows fine @ 2640MHz. 3dmark06's CPU test froze though.
Anywhere in 2.5 seems fine with auto settings. I will try some real overclocking later.
@Mr.Q
Just curious, did you loosen up your memory timings? And if so, which ones...
Thanks.... Dave
I'll let you know the basic Phenom pattern. ;)
If you want similar overclocks to Mr.Q:
Get >1.25VID
VCore +1.5V
Low NB/HT
It should run close to 2.8-2.9GHz with those, benchable. Probably not stable though and probably not repeatedble consistently.
On 1.20VID you won't get above 2.8GHz benchable (maybe not even screenable) at best and ~1.45V. It's a lower performance state, thus lower MHz/TDP allowed. This wa the main problem with mine, it switched VID too often auto.
Max I benched was 2850 and I ran it for many days. I didn't go above 2870MHz on benching and that was only little apps. Mainly because I couldn't be bothered as it was that unstable.
Mr.Q: Yep. I thought you'd benched 3GHz too Mr.Q? :confused:
I hope those temps are chilled watercooling or they're being reported wrongly. Phenom 1.2V 1.20VID idles near 12C above ambient with the best air HSF and some deltas or with medicore WC. With voltage and VID increase TDP shoots up v.quickly and at 1.53V 1.25VID, you'll be seeing high temps. ;)
sadly no KTE. If I retain the 9x NB at boot and use unganged mode the max I can get NB speed to is 2047mhz or thereabouts :(Quote:
originally posted by KTE
MR.Q: Nice testing. One thing though; have you tried 2100/2200 in unganged mode with the same settings? Is that possible?
I set all my memory subtimings to auto for this Dave. I usually manually set them but have allowed auto to remain since my last cmos reset. I figured I'd give the memory an easier time and concentrate on the cpu/nb speeds. KTE is correct with his info below.Quote:
Originally Posted by Daveburt714
This is exactly the case as KTE has said. NB was dropped to a multi of 7 before this which allows for more headroom on the cpu fsb.Quote:
Originally Posted by KTE
Not quite 3Ghz no, I've a 3003mhz screeny from AOD but it was so unstable half the screenshot is corrupt, as in the bottom half is pure grey. Max verified so far is 2948mhz using cas 4 on mem or 2937mhz using cas 3 on this A-data DDR2-800.Quote:
Mr.Q: Yep. I thought you'd benched 3GHz too Mr.Q? :confused:
I hope those temps are chilled watercooling or they're being reported wrongly. Phenom 1.2V 1.20VID idles near 12C above ambient with the best air HSF and some deltas or with medicore WC. With voltage and VID increase TDP shoots up v.quickly and at 1.53V 1.25VID, you'll be seeing high temps. ;)
For cooling I basically have my pc outside. I've fitted intake hoses to port chilled air from outside directly onto the cpu hsf aswell as running the cpu fan on 15v which drops a further 4c from my cpu's idle temps. I've fitted another intake hose directly onto my system memory with a 12v 80mm fan fitted directly on top of the dimms, then finally my gpu is also fitted with a 12v 80mm fan drawing air directly from outside which as you know gets exhausted directly out of the case rear with g80's. This leads to my case ambient temps being very low. The highest cpu load temp I ever see with 4 cores fully loaded is 43c at 1.560v-1.580v :up:
What kind of memory speeds are typical with these Phenoms guys?
I am not even sure of how to ask that question as there are so many variables I am sure but I guess help me fill in the blanks to some degree please;)
one word, great.
Whether it's unganged mode or the fact there are two mem controllers to split the workload I am unsure.....but any memory Ive tested on this phenom rig far exceeds anything I can get with the same dimms on any other rig to-date.
For example on my 6000+ rig I had this a-data ddr2-800 (5,5,5,18,2t) running at 790mhz using 4,4,4,12,2t which I was happy with. Sticking to cas 5 allowed the dimms to reach 430mhz-ish stable. Now with the very same memory running in the phenom rig I can push 445mhz with 3,5,5,18,2t with I'm sure more headroom. I couldn't even get my M2N-E SLi deluxe with 6000+ to boot at ddr2-800 cas4:down:
If I had better memory then I'm sure I'd see even better results, but for el-cheapo mem at £17 per 1GB you really can't complain :)