Hitting wall at 3.05GHz??

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Hopefully you gents can help me out here. I seem to be hitting a wall with my chip that acts like a setting I need to tweak rather than the chip hitting it's max. It's OCCT stable at 3.05GHz@1.44V (305x10) but any higher than that and OCCT errors out almost immedietely, even with 1.6V. Temps hit 38C at load running 1.44V and 46C at 1.6V. I'm running a 166 divider with my ram which puts it at ~253MHz. I've tested it up to 270MHz with no issues so I don't think that's the problem. I've tried raising the chipset and dram voltage with no effect.

I guess I should test the HT ceiling on this board but I've always heard the DFI's wouldn't have any problem at the speeds I'm running. Any ideas?
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
I just have it on Auto since that's what the guys over at DFI Street recommend. Maybe I'll try it at 3x though, just to make sure.
 

Thor86

Diamond Member
May 3, 2001
7,888
7
81
Try 9x multi and increase your FSB/HTT. You may need to run a lower memory divider.
 

thecoolnessrune

Diamond Member
Jun 8, 2005
9,673
583
126
Boo hoo, I am the euivalent of a 4.7 Ghz Intel and it can't go faster. Boo hoo. Lol. Seriously though, I would try lowering the LDT and the multi and see if you can go farther. Also maybe your ram is limiting you? What divider are you using? Perhaps you should relax your timing a bit if it is too highly overclocked. That or user a higher divider.
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Thanks for the sympathy guys.

I'm currently using a 166 divider which should put the ram where it needs to be. Like I said it tests out fine at 270MHz and I'm running it at 254Mhz. The chip doesn't seem to like using a lower divider but maybe I'll try that too. Timings are pretty relaxed at 3-4-4-8.
 

LED

Diamond Member
Oct 12, 1999
6,127
0
0
3GHz is the wall for Socket 939...that why there'll be 1 change.
 

sodcha0s

Golden Member
Jan 7, 2001
1,116
0
0
Originally posted by: Elfear
Thanks for the sympathy guys.

I'm currently using a 166 divider which should put the ram where it needs to be. Like I said it tests out fine at 270MHz and I'm running it at 254Mhz. The chip doesn't seem to like using a lower divider but maybe I'll try that too. Timings are pretty relaxed at 3-4-4-8.

The RAM is probably fine at that speed, but perhaps the memory controller on the CPU is starting to get stressed. Just take it down to 100 divider and test. If it passes at that, you know it's the controller. You have a nice chip there, don't fry it over a few Mhz... :)
 

aka1nas

Diamond Member
Aug 30, 2001
4,335
1
0
Originally posted by: Elfear
I just have it on Auto since that's what the guys over at DFI Street recommend. Maybe I'll try it at 3x though, just to make sure.

Auto will be 5x most likely, which is like 1500Mhz HT, which is unlikely to be stable even on that DFI.
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Ok tried the following settings out:

1) 3x HTT
2) 10x305 w/ 133 divider (errors out of OCCT almost immedietely)
3) 10x305 w/ 100 divider (ditto)
4) 9x327 w/ 140 divider (wouldn't even get past the 2nd boot screen)
5) 8x368 w/ 140 divider (ditto)

Seems kind of strange that the two dividers I tried were so unstable as the ram was running at fairly low speeds.

Not sure what the deal is at this point. I was hoping to be OCCT stable at 3.15-3.2GHz on water since the thing is OCCT stable at 2.975GHz@1.39V and 3.05GHz@1.44V. The other A64 chips I've used had a pretty linear rate of return with an increase in voltage (generally another 100-200MHz of headroom after max oc on 1.4-1.45V) but this chip seems to do very well at low volts up to a certain point and than no matter what vcore I run it won't go higher.
 

lopri

Elite Member
Jul 27, 2002
13,314
690
126
Which BIOS are you using? Dividers work very differently per each BIOS revision.
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Originally posted by: Assimilator1
Sounds like the limit of the chip.
Nice o/c btw :D

Thanks.

I'd be thinking the same thing if the vcore at 3.05GHz wasn't relatively low. I was pumping 1.65V through my 3700+ before I hit my max.
 

lopri

Elite Member
Jul 27, 2002
13,314
690
126
Have you tried Prime95 or Super PI? I don't like OCCT too much so I don't use it, but I try to sort the problem out with Prime95 and Super PI. Sometimes just with Prime95. For instance, if you can run Prime95 Small FFT but fails in Super PI (anything bigger than 8M), then RAM timing might be causing a problem. If you pass 32M but not in Small FFT, your CPU is not stable. I don't know how OCCT works (CPU? Memory? CPU+Memory?), but my understanding is that Prime95 Small FFT is purely CPU. (Isolating RAM from equation)

Try to test each part separately. I know it takes time but in the end it saves time if you know each part's limit. FYI, my DFI SLI-D board maxes around 335~340HTT @1.7V.

 

Leper Messiah

Banned
Dec 13, 2004
7,973
8
0
Originally posted by: Elfear
Ok tried the following settings out:

1) 3x HTT
2) 10x305 w/ 133 divider (errors out of OCCT almost immedietely)
3) 10x305 w/ 100 divider (ditto)
4) 9x327 w/ 140 divider (wouldn't even get past the 2nd boot screen)
5) 8x368 w/ 140 divider (ditto)

Seems kind of strange that the two dividers I tried were so unstable as the ram was running at fairly low speeds.

Not sure what the deal is at this point. I was hoping to be OCCT stable at 3.15-3.2GHz on water since the thing is OCCT stable at 2.975GHz@1.39V and 3.05GHz@1.44V. The other A64 chips I've used had a pretty linear rate of return with an increase in voltage (generally another 100-200MHz of headroom after max oc on 1.4-1.45V) but this chip seems to do very well at low volts up to a certain point and than no matter what vcore I run it won't go higher.


The later revisions of A64 have been really iffy on the volts. My manchester is F@H stable at 2.6 with like 1.4 vcore tops. 2.7 won't work with 1.55 vcore. Kinda like TCCD I guess.

 

Assimilator1

Elite Member
Nov 4, 1999
24,155
520
126
Originally posted by: Elfear
Originally posted by: Assimilator1
Sounds like the limit of the chip.
Nice o/c btw :D

Thanks.

I'd be thinking the same thing if the vcore at 3.05GHz wasn't relatively low. I was pumping 1.65V through my 3700+ before I hit my max.

I know what you mean ,usually the 1st few increases or so in vcore will give you a decent amount extra clockspeed.But I've been reading a large thread (45+pages) over at Overclockers UK & some people in their are reporting the same odd phenomena ,maybe as mentioned its the memory controller hitting its limit?
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Originally posted by: Assimilator1

I know what you mean ,usually the 1st few increases or so in vcore will give you a decent amount extra clockspeed.But I've been reading a large thread (45+pages) over at Overclockers UK & some people in their are reporting the same odd phenomena ,maybe as mentioned its the memory controller hitting its limit?

Interesting. Thanks for posting that. Now, if it does turn out to be the memory controller hitting it's max I won't feel too bad. I feel kind of dumb though that I sold my other Opteron to my brother. It maxed out at 2.8-2.85GHz@1.4V and was barely bench stable at 3.15GHz@1.65V. I thought for sure the chip I'm using would have more headroom since the max is so much higher at stock vcore. Live and learn I guess.

I'll keep plugging away and see if I can figure this thing out. It did pass test #5 of Memtest 35 times at 3.1GHz@1.475V with ram at 258MHz 3-4-4-8, so I'm fairly confident the ram isn't the issue.
 

the Chase

Golden Member
Sep 22, 2005
1,403
0
0
The Opterons(and maybe newer Venices too?) seem to hit the wall quickly as you try to ramp up the voltage. Getting to 2.98 stable for mine is fairly easy but to get to 3.0 STABLE takes a lot more volts. I'm running at 3.0 but with reduced volts(errors out on prime 95 within 10 mins. but is stable with everything that I run).
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Originally posted by: the Chase
The Opterons(and maybe newer Venices too?) seem to hit the wall quickly as you try to ramp up the voltage. Getting to 2.98 stable for mine is fairly easy but to get to 3.0 STABLE takes a lot more volts. I'm running at 3.0 but with reduced volts(errors out on prime 95 within 10 mins. but is stable with everything that I run).

I hear ya. I'm not really looking for Prime stable at 3.15-3.2GHz, I was just hoping it would be game/bench stable at those speeds. Maybe I'll have to do a Jedi mind trick on my brother if this cpu really maxes out at 3.05GHz.