Hitting wall at 3.05GHz??

[Zim]

This seems to be a characteristic of all A64 chips I have tested. They hit a wall and after that it's a case of rapidly diminishing returns. It may be interesting to see just how far you can push a chip, but in order to actually use it you should stay on the low side of the wall. 3GHz is the limit you should expect to run that chip day-to-day.

 

[Elfear]

Originally posted by: Zim
This seems to be a characteristic of all A64 chips I have tested. They hit a wall and after that it's a case of rapidly diminishing returns. It may be interesting to see just how far you can push a chip, but in order to actually use it you should stay on the low side of the wall. 3GHz is the limit you should expect to run that chip day-to-day.

I agree. It's Prime stable at 2.95GHz@1.408V so that's probably what I'll use for everyday use. Just looking for a nice oc for benching/gaming.

Originally posted by: JustAnAverageGuy
You got over a 50% overclock. That's impressive. Don't know what you're complaining about.

Sorry, I'm not complaining. 3.0Ghz is great for a $175 chip but overclocking for me is about finding the limits so I'm asking for assistance from you guys to help me find it.

 

[Big Lar]

Try memtest test 8, it seems to sort out memory problems better than test 5.
On my chip, I need to add an extra 3% to the cpu voltage to stabilize it, to just get a few more mhz out of it. I think it is the memory controller, and the extra juice helps it some. However I don't run it at max all the time, I just tone it down to 3080 with less volts.

 

[Elfear]

Originally posted by: Big Lar
Try memtest test 8, it seems to sort out memory problems better than test 5.
On my chip, I need to add an extra 3% to the cpu voltage to stabilize it, to just get a few more mhz out of it. I think it is the memory controller, and the extra juice helps it some. However I don't run it at max all the time, I just tone it down to 3080 with less volts.

I originally thought I could just add some more volts but I can't go any higher than 3.05GHz even with 1.6V. I'll give test #8 a try and see if that picks anything up.

 

[Elfear]

Ok, I tried MAL = 9, Preamble = 7, Drive Strength = Weak, Chipset Voltage = 1.8V, and LDT Voltage = 1.4V. It seems to be the most stable at those settings with the rest being the same as this screenshot. By "most stable" I mean it will pass 7 minutes of SuperPi 16M. SuperPi 8M seems to be hit and miss at those settings, sometimes it passes sometimes not. This is at 310x10 1.55V. More vcore doesn't seem to do anything.

I also tried the following with a subsequent loss of stability:

1) 133 divider (310x10)
2) 2T command rate
3) Bypass Max = 7x
4) Dram voltage = 2.8V
5) Idle Cycle Limit = 16 clks
6) Cycle Counter = Enabled
7) Trfc = 15

I seem to be a bit limited on my HT headroom too. With no real load on the cpu (8x multi) I can get the HT up to 325MHz (OCCT stable) before it poops out. 330MHz is completely unstable. I always thought the DFI boards were good at 350+ but maybe with the extra load SLI puts on the chipset I'm being limited.

 

[Elfear]

Bump.

 

[Assimilator1]

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).

Maybe so far ,but it'll cause crashes sooner or latter ,maybe when you install new software or a new game.
You really need it to be Prime passable for it to be properly stable & reliable

 

[Leper Messiah]

Originally posted by: Elfear
Ok, I tried MAL = 9, Preamble = 7, Drive Strength = Weak, Chipset Voltage = 1.8V, and LDT Voltage = 1.4V. It seems to be the most stable at those settings with the rest being the same as this screenshot. By "most stable" I mean it will pass 7 minutes of SuperPi 16M. SuperPi 8M seems to be hit and miss at those settings, sometimes it passes sometimes not. This is at 310x10 1.55V. More vcore doesn't seem to do anything.

I also tried the following with a subsequent loss of stability:

1) 133 divider (310x10)
2) 2T command rate
3) Bypass Max = 7x
4) Dram voltage = 2.8V
5) Idle Cycle Limit = 16 clks
6) Cycle Counter = Enabled
7) Trfc = 15

I seem to be a bit limited on my HT headroom too. With no real load on the cpu (8x multi) I can get the HT up to 325MHz (OCCT stable) before it poops out. 330MHz is completely unstable. I always thought the DFI boards were good at 350+ but maybe with the extra load SLI puts on the chipset I'm being limited.

have you tried it w/o SLi? It does place more stress on a system, I'd take 3.0 with SLi vs. 3.2 w/o any day, but if you're curious as to what the max is, its worth a try.

 

[Elfear]

Originally posted by: Leper Messiah

have you tried it w/o SLi? It does place more stress on a system, I'd take 3.0 with SLi vs. 3.2 w/o any day, but if you're curious as to what the max is, its worth a try.

I haven't yet only because it's a huge pain to pull one of the cards with the H2O tubing. I just built a rig for my brother with a DFI motherboard. I think I'll try it out on his board and see what happens.

 

[Elfear]

Well here is a brief update. I can get the thing SuperPi 16M stable at 310x10 but I have to use the following settings:

Vcore - 1.6V
LDT Voltage - 1.5V
Northbridge Voltage - 1.7V

Seems like a fairly huge jump in voltage for an extra 50mhz and less stability. The weird thing is that my other Opteron was barely bench stable at 3.15GHZ on this board, so I know it's possible. That was at 1.65V though and this particular chip runs a lot hotter (48C@1.6V under load) so I'm hesitant to up the vcore any more. Any other suggestions before I just chalk it up as the cpu's max?