Revision E OC discussion thread

cell_491

2[H]4U
Joined
Oct 16, 2004
Messages
2,608
Since Eclipse seems to be busy I went ahead and made a discussion thread for his results thread...ENJOY!! :D
 

XeroHouR

[H]ard|Gawd
Joined
Sep 14, 2004
Messages
1,401
My 3500+ on Epox 9NDA3J+ is only doin 2.65 right now, I'm bumpin up .1V but not gettin very far. I'm not sure whats keeping me, its 35oC idle and maybe 45 load, I've got water on it. I'm not able to get it fully stable at 2.7. I accredit my RAM , its TCCD but won't do well without >2.8V. Anyone else doing anything, do I just have a dog?
 

GodsMadClown

2[H]4U
Joined
Aug 9, 2002
Messages
3,112
Yeah. I just realized thet I shouldn't have posted discussiontype stuf in there. Sorry. Feel free to delete it.

Anyhow, I do Folding@Home with my computers. They are loaded 24/7, and I like them to be very stable, but very fast. So, I'd like to see the Prime95 validation time moved to 24-48 hours.
 

GodsMadClown

2[H]4U
Joined
Aug 9, 2002
Messages
3,112
XeroHouR said:
My 3500+ on Epox 9NDA3J+ is only doin 2.65 right now, I'm bumpin up .1V but not gettin very far. I'm not sure whats keeping me, its 35oC idle and maybe 45 load, I've got water on it. I'm not able to get it fully stable at 2.7. I accredit my RAM , its TCCD but won't do well without >2.8V. Anyone else doing anything, do I just have a dog?

Use a memory divider and loose timings to eliminate the RAM as a bottleneck. When you discover the CPU limits, you can tighten the memory back to stable settings. I use Memtest86+ for this.
 

cell_491

2[H]4U
Joined
Oct 16, 2004
Messages
2,608
GodsMadClown said:
Yeah. I just realized thet I shouldn't have posted discussiontype stuf in there. Sorry. Feel free to delete it.

Anyhow, I do Folding@Home with my computers. They are loaded 24/7, and I like them to be very stable, but very fast. So, I'd like to see the Prime95 validation time moved to 24-48 hours.
well folding still doesnt run your cpu at 100%
 

GodsMadClown

2[H]4U
Joined
Aug 9, 2002
Messages
3,112
cell_491 said:
well folding still doesnt run your cpu at 100%

Yes, it actually does. I have the command line version set to run as a service with idle priority. This means that it runs on any idle time that my CPU has. My task manager graph is always pegged at %100.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
If I can run Prime or Memtest overnight, I'm satisfied...go ahead and run it for two days straight if it's what you want, but I figure if it doesn't crash under load after 6-8 hours, it's not likely to, at least not due to the overclock.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
GodsMadClown said:
Yes, it actually does. I have the command line version set to run as a service with idle priority. This means that it runs on any idle time that my CPU has. My task manager graph is always pegged at %100.

It fluxuates a little, at least everywhere I've seen it. And please don't tell me you're using FAH for load testing...
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
cell_491 said:
hey eclipse you have my oc down under air you should put it under budget water cooling or sumthing
thanks for the thread cell. and i'll change that. doing stability testing at 2.82ghz though ;)


here's what i'm at so far. i think i can take first in that 512k bench we had a few weeks ago

512k-13.593s.PNG
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Sweet. I haven't been able to break 14 yet, though I'm hoping once I get my primary hard drive cleaned up so there's more than 3% free space I'll be able to shave half a second off...
 

GodsMadClown

2[H]4U
Joined
Aug 9, 2002
Messages
3,112
ashmedai said:
It fluxuates a little, at least everywhere I've seen it. And please don't tell me you're using FAH for load testing...

Mine doesn't fluctuate. You must have been looking at the graphical version. It's less efficient.

For testing I use SuperPi for a rough stability assesment, then Prime95 (In Place FFTs) to test the CPU for 10 hours. Then I tweak out the memory speed and timings to run a error-free 24 hour loop of the extended tests of Memtest86. Finally I do a good 24-48 hours of Prime95 (blend) to test the whole package of memory and CPU.

Prime95 will sometimes error after 24-48 hours on something that I've hoped would be stable.

I do media encoding and file processing all the time on my rig and I want to to be solid enough to be on 24/7.
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
prime95 doesn't work with this cpu though. it seems that i can do anything i want at the speeds above, aside from prime. infuriates me, but whatever. i can live
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
I kept getting rounding problems, turned out to be my second stick of memory was being naughty. I usually use CPU Burn anyway, it has a timer.
 

tsuehpsyde

Supreme [H]ardness
Joined
Oct 22, 2004
Messages
6,604
GodsMadClown said:
So, I'd like to see the Prime95 validation time moved to 24-48 hours.

I agree...I personally think if anyone posts their OC without any less are lying.

My OC's Prime (not Rev. E, but I agree on the Prime95 argument):

http://www.sourcekills.com/gallery/view_photo.php?set_albumName=tsuehpsyde&id=prime_running
http://www.sourcekills.com/gallery/view_photo.php?set_albumName=tsuehpsyde&id=prime_stopped

Now....someone......please....TELL ME WHAT REV. E PROCESSORS RUN 1:1 WITH TWO STICKS! K THX. :D This leans highly if I'm going to upgrade to a Newark or to socket 939.....

/end yelling
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
P4 uses the equation FSB * Memory Divider = Memory speed. Because of the way it's designed it helps a lot if you stick with 1:1.

A64 uses the equation CPU Speed / Memory Divider = Memory speed, where the Memory Divider is determined by [CPU Multiplier / Memory Ratio] rounded up. So 1:1 on an A64 would mean the memory speed is equal to the CPU speed, which is pretty much nonsensical.
 

tsuehpsyde

Supreme [H]ardness
Joined
Oct 22, 2004
Messages
6,604
Uhm, I'm aware HOW it works....I just want to know where the memory controller shits out after 2 sticks of memory. Somewhere around 240MHz Clawhammer's memory controller flakes out with two sticks, but goes to 250MHz+ with only one stick. Looking to see what kind of gains the Revisions on the memory controller have for two sticks, as I want to run DDR500+ on my memory, running both sticks 24/7.
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
I hit DDR500+ just fine on a Winchester...for that matter my current Venice can too. What I want to know is, did they fix that "4x double sided requires 2T" thing yet?
 

tsuehpsyde

Supreme [H]ardness
Joined
Oct 22, 2004
Messages
6,604
Official word is, they will all run DDR400 four sticks, but only guarenteed at 2T....but eclipse has a link to someone that can do it at 1T. But I doubt that's the case for the entire line.
 

J-Mag

2[H]4U
Joined
Dec 21, 2004
Messages
3,640
ashmedai said:
P4 uses the equation FSB * Memory Divider = Memory speed. Because of the way it's designed it helps a lot if you stick with 1:1.

A64 uses the equation CPU Speed / Memory Divider = Memory speed, where the Memory Divider is determined by [CPU Multiplier / Memory Ratio] rounded up. So 1:1 on an A64 would mean the memory speed is equal to the CPU speed, which is pretty much nonsensical.

Agreed: Going "1:1" on a A64 is pointless unless it just so happens to be your sweetspot. There is no performance loss when using ratios other than the obvious mem clock decrease but you are overclocking that anyway...
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
unless there is when you have a weird rig like mine that refuses to run at anything less than 1:1...

well i should rephrase that. i can do 250x10 1:1.
5:6 gets me to..... 250x10! wow! i think the dfi hates ballistix
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
Oh yeah, and on the 3000+? Don't count on overclocking your memory much. Getting the CPU maxed out and stable at that high a HT is pretty much all it's going to do. I could get the memory up to about 245 for a little bit at a time, but mostly it doesn't want to go very far over 200 once you're in the neighborhood of 300 HT.
 

SGTFury

n00b
Joined
Aug 17, 2004
Messages
51
Got my 3200 Venice today :D Have to wait til the rest of my watercooling stuff comes in though, then I'll set it up and post benchies. The CPU:

ADA3200DAA4BP
LBBLE 0516EPEW
 

Sottilde

Limp Gawd
Joined
Apr 19, 2005
Messages
177
My:

ADA3000DAA4BP
LBBLE 0516CPBW

Does 2474 right now at 274 HTTx9, with a 5/6 divider. For some reason, a 2/3 divider just kills my oc... going to try looser mem timings tomorrow.
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
yeah, dividers kill me too, i think the dfi was made to run 1:1 with high scaling ram :(
 

cell_491

2[H]4U
Joined
Oct 16, 2004
Messages
2,608
(cf)Eclipse said:
yeah, dividers kill me too, i think the dfi was made to run 1:1 with high scaling ram :(
even at stock speeds my dfi board wont boot if i set it to any divider
 

ashmedai

[H]F Junkie
Joined
Feb 26, 2005
Messages
12,402
I'm running a 7/10...only quirk is that CPU-Z and A64Tweaker can't figure out what it's set at.
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
taken from the oc data thread


ProOC said:
If I dropped the HT to 3x I could hit 2.8ghz but it wasnt stable. I am going to mess with it some more by loosening ram timings, but I am thinking it may not help as my ram is rated for 433 and I'm only at like 196ish with that divider. Anyone have any ideas what might be holding me back? I'm using water and my temps dont ever go above like 40c it seems.
do you really need that much voltage? i find that when i give my core too much voltage, it becomes less stable :(



Rasha said:
i wanted to keep my ram 1 to 1.. does a divider hurt it that mutch?
http://valid.x86-secret.com/show_oc?id=16619
no, feel free to drop the divider to get more mhz out of the cpu. core clock speed is way more important than ram clock speed anyhow ;)
 

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,854
I have been considering grabbing a venice because at xs im seeing lots of them hit on average 2.7-2.9 with 3000 and 3200's. my winchester 3200+ used to be stable at 280x9, now 270x9, maybe my heatsink is dusty, problem here - I need 1.8v! This will surely kill this poor chip if I keep this up. 2.7ish would make is somewhat worth it with lower voltage, and I get to make use of this overpriced ram I foolishly bought :D I don't like to play the luck game, but thats what overclocking is, I lost to cbbid last time, should I grab a venice before they lose their oc potential?
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
2.7ghz > 2.4ghz. it's just a matter of if you think it's worth it. i don't see the oc-ability dropping off like winchester did though.. so maybe you can just wait till that chip dies :D
 

Jesus_Faction

[H]ard|Gawd
Joined
Mar 31, 2005
Messages
1,078
ashmedai said:
Oh yeah, and on the 3000+? Don't count on overclocking your memory much. Getting the CPU maxed out and stable at that high a HT is pretty much all it's going to do. I could get the memory up to about 245 for a little bit at a time, but mostly it doesn't want to go very far over 200 once you're in the neighborhood of 300 HT.

yeah mate im getting similar results with my 3000
ive been running at abut 280 htt with the ram at ddr420 and 2-2-2

i installed windows 64-bit today tho and now i cant load it up without dropping my command rate to 2T with my 2x512 twinmos
 

Jesus_Faction

[H]ard|Gawd
Joined
Mar 31, 2005
Messages
1,078
ive got the ultra-d and i just upgraded to the newest beta bios

it isnt any better than the 3/10 official release in regards to this particular issue
 

Jesus_Faction

[H]ard|Gawd
Joined
Mar 31, 2005
Messages
1,078
yup already tried that

i think ill just reinstall windows 32-bit tomorrow if i cant find anything about this tonight
 

Jiggaman2576

Limp Gawd
Joined
Oct 27, 2004
Messages
377
I got me a sandiego 3700+ and the DFI NF4-SLI-DR and I couldn't be happier. Currently running it at 300X9 1:1 with 3-4-4-8, I may try lower timings but it requires 2.9V on the ram. I have to give my CPU almost 1.7V to make it stable at 2.8 ghz though, and I wanted to know what is safe for these CPU's? I'm on water, but have crappy thermal paste in it right now (where did I put my as?!) My temps with good thermal paste rarely reach 100F full load. I dont have any screens but sisoft reports 7500 mb/s bandwidth, my CPU scores higher then the highest one by a decent margin, 29S for 1M in superpi/13 for 512K. Superpu stable up to 8M, 30,309 3dmarks (01). Screenshots take up too much bandwidth, they are nice to verify someone actually did what they say they can do, but really, unless its just a rediculous number, what is the point? Seems a lot of people can hit these kinds of numbers anymore, which is great for us!
 

ProOC

Limp Gawd
Joined
Jun 17, 2003
Messages
129
(cf)Eclipse said:
taken from the oc data thread

do you really need that much voltage? i find that when i give my core too much voltage, it becomes less stable :(

Yes I needed that much to get it stable at 2.75ghz.

I tried to loosen my ram timings even though my ram was only at 200 with the divider and did up to 230 without the divider when I was going for the 255 and more fsb, but I just cant get over 250fsb stable. Not sure what it is, but I would like to know. I tried dropping my cpu multi to see if it was the chip freaking out because it shouldn't be my ram, but my mobo told me I needed to change my settings when I rebooted, but I told it to boot anyway and it was using the lower multi but didnt take the fsb of 255 into effect and was using 200 so I had a slow clock and slow fsb for some reason, so I am at a loss of what to do next.

Not that 2.75ghz stable isn't decent or anything :p
 

(cf)Eclipse

Freelance Overclocker
Joined
Feb 18, 2003
Messages
30,027
i think you just need a new bios.. wait for msi to get off their lazy bums and write one for you ;)
 
Top