am I at the limit ? 2.4c at 3.3

[ZL1]

well I got my AS5 ...

(warning long story to follow, please forgive)

... so the first thing I wanted to do was install a new (active) northbridge hsf (I had one by ThermalTake on the shelf, yes I know I know, what was I thinking ), but when I got to it I noticed it was rather small compared to the stock one, especially the base bothered me, it was smaller, more uneven and it didnt have the padding the stock one has for the caps (then I noticed something interesting, the Tt was actually for a 845)
I decided to keep the stock one, use some AS5 and glue the 40mm fan from the Tt to the top - it worked nicely
but that didnt help with the OC though (well the fan does only push 5.1cfm)
my oc remained at 3.2 max stable, 3.3 win boot unstable and 3.4 post (I do think it improved NB temps a bit)
so I went ahead to step 2, installing the zalman 7000alcu I had waiting for as5
installed this one, with some modifications (had to bend some of the fins as they were hitting the newly glued on 40mm on the northbridge , but that ended up working just fine)
now with the as5 and the zalman on the cpu I started getting much better temps for one, it improved my load temps by about 15degrees, I could not belive it, I still cant
I think the zalman improved the northbridge temps as well, because of the zalman's construction and its proximity to the northbridge it blows alot of air on the NB and the temps sure feel lower (to the touch).
my new cpu speeds are 3.2 stable with stock voltage (much cooler), 3.3 with 1.7v bios (speedfan says its more like 1.75 and under load it drops to about 1.625-1.65 I think a droop mod is in order), 3.4 max windows (prime fails with stock voltage, but with 1.7-1.75v (1.75-1.8v speedfan) it doesnt fail but the whole system freezes a few mins into prime - am I at the limit ?) and a max of 3.6 post, very nice stuff I wonder if I could actually boot (maybe) with some better cooling

so in the end I got a cooler northbridge = longer life for the mobo, cooler cpu = longer cpu life and the hope for reaching newer speeds

my temps are:
3.2 = max 38 (38 is on occasion, its running mostly 36 under 100% load on both cpus)
3.3 = max 42 (same as above, usually 40)
3.4 = max ?? freezes (but I havent seen more than 42, was able to run superpi)
I know asus was known to show lower temps, but even if we add another 10 to all those numbers, they still look ok

so what you guys think ? am I at the limit of the cpu ?

also prime related question, I noticed that prime cant use 100% of cpu (and I couldnt run two instances for some reason, prime 2.38) so I decided to run prime + superpi, but that didnt work well since prime for some reason uses up a huge chunk of ram (about 400mb !!!) and that leaves almost nothing to the others and when I try to run superpi I get swap access and the cpu load drops
(I did run prime + toast, as toast doesnt need alot of ram to run)

voltage question, would you guys say its safe to run at 1.75 ?
also since it looks like a droop mod would be quite a good idea, anyone know a good guide for droop modding a p4c800-e-d ?
and I was thinking since Im soldering I may as well volt mod, anyone know a good guide for volt modding this mobo ? also anyone know whats the max vdimm I can get by voltmodding this thing ?

so what you guys think ?
think I could run 3.4 ?
is 3.6 possible with better cooling ?
I was thinking xp90 with a good fan (cant use xp120 wont fit, well it will but wont clear the nb with a fan on it)
or maybe mod the zalman with a higher cfm fan ? I was thinking that might be great as it blows on the NB and ram too
or a 120mm zalman and mod ?

please let me know and sorry again for the huge post and huge number of questions


Thank you
Dan

EDIT: by the way bios is 1015, I hear this is best for northwood OCing
also the ram was set to 3:2 for testing purposes, so the ram shouldnt be holding me back

EDIT2: I was able to run two instances of prime, but I got max cpu load of 1-5% the hdd was going like crazy since these two used up all the ram, anyone know what could be going on ?

 

[sxr7171]

Yes, I can help you with your Prime95 problem: set the toture test to the first option - small FFT. The third option just uses all your RAM and thrashes your hard drive around - I guess it can be useful to see if you have errors in your RAM or HD interface but the CPU <5%.

 

[ZL1]

Originally posted by: sxr7171
Yes, I can help you with your Prime95 problem: set the toture test to the first option - small FFT. The third option just uses all your RAM and thrashes your hard drive around - I guess it can be useful to see if you have errors in your RAM or HD interface but the CPU <5%.

Thank you !

running two nicely right now, will let it burn in overnight at 3.3

 

[sxr7171]

Originally posted by: ZL1

Originally posted by: sxr7171
Yes, I can help you with your Prime95 problem: set the toture test to the first option - small FFT. The third option just uses all your RAM and thrashes your hard drive around - I guess it can be useful to see if you have errors in your RAM or HD interface but the CPU <5%.

Thank you !

running two nicely right now, will let it burn in overnight at 3.3

Even though it says 2 gets hotter, I found that 1 is actually hotter. Try it and see which works better for you. If you want it hotter (do this if your temps are stable only be very careful) look for a program called "CPU Burn" it gets that thing hot but does not check for errors. Good for checking max temps.

 

[ZL1]

Originally posted by: sxr7171

Originally posted by: ZL1

Originally posted by: sxr7171
Yes, I can help you with your Prime95 problem: set the toture test to the first option - small FFT. The third option just uses all your RAM and thrashes your hard drive around - I guess it can be useful to see if you have errors in your RAM or HD interface but the CPU <5%.

Thank you !

running two nicely right now, will let it burn in overnight at 3.3

Even though it says 2 gets hotter, I found that 1 is actually hotter. Try it and see which works better for you. If you want it hotter (do this if your temps are stable only be very careful) look for a program called "CPU Burn" it gets that thing hot but does not check for errors. Good for checking max temps.

well here is what I did, I ran first instance with option #1 and second instance with option #2
running full blast, 42C at 275x12 (3.3) with 1.7v in bios (1.75v soft) 1.625 load voltage
case=34, room temp is 75F or so

Thanks
Dan

 

[sxr7171]

It doesn't really help to run 2 instances since 1 instance will max out your CPU. Just running one might make it hotter since the first and second options use different techniques to torture the CPU and now your CPU cycles are split between the two methods. Try one at a time and see if you get higher temps. You kind of want to see how high you can possibly go. But 42C is darn good.

 

[ZL1]

sxr7171 if I use just one prime it will only use 50% of cpu

ok dual prime still running
since last morning (18hours)

cpu temps are as follows
if case is 32 the cpu is 42
if case is 34 the cpu is 44
I opened window, dropped case temp to 27 and got 38 cpu
seems to follow a general rule of approx 10deg above case

so on temps Im good, but I wonder about the oc itself, what you guys think ? time for a new cpu ?


Thank you
Dan

 

[zakee00]

Originally posted by: sxr7171
It doesn't really help to run 2 instances since 1 instance will max out your CPU. Just running one might make it hotter since the first and second options use different techniques to torture the CPU and now your CPU cycles are split between the two methods. Try one at a time and see if you get higher temps. You kind of want to see how high you can possibly go. But 42C is darn good.

yes it does....he is using a HT processor, so he needs two open to use 100% of his cpu.

 

[cockeyed]

You could disable HT in the BIOS and that will give you 100% CPU usage.

 

[ZL1]

Originally posted by: cockeyed
You could disable HT in the BIOS and that will give you 100% CPU usage.

why not just run two primes ?


Thanks
Dan

 

[sxr7171]

Originally posted by: zakee00

Originally posted by: sxr7171
It doesn't really help to run 2 instances since 1 instance will max out your CPU. Just running one might make it hotter since the first and second options use different techniques to torture the CPU and now your CPU cycles are split between the two methods. Try one at a time and see if you get higher temps. You kind of want to see how high you can possibly go. But 42C is darn good.

yes it does....he is using a HT processor, so he needs two open to use 100% of his cpu.

I'm sorry I had no idea about HT processors.

 

[cockeyed]

Originally posted by: ZL1

Originally posted by: cockeyed
You could disable HT in the BIOS and that will give you 100% CPU usage.

why not just run two primes ?


Thanks
Dan

Either way works to load the CPU to 100%; take your pick. I recall seeing a discussion on the prime95 forum about this subject, it might provide you with some more info.. Prime95 forum
Select "Full View Version" at the top of the page for standard forum format view.

 

[ZL1]

I will check it out, thanks

Dan

 

[VirtualLarry]

Originally posted by: zakee00

Originally posted by: sxr7171
It doesn't really help to run 2 instances since 1 instance will max out your CPU. Just running one might make it hotter since the first and second options use different techniques to torture the CPU and now your CPU cycles are split between the two methods. Try one at a time and see if you get higher temps. You kind of want to see how high you can possibly go. But 42C is darn good.

yes it does....he is using a HT processor, so he needs two open to use 100% of his cpu.

No, that's a fallacy caused by a mis-understanding of Windows' CPU-usage graphs.

 

[ZL1]

Originally posted by: VirtualLarry

Originally posted by: zakee00

Originally posted by: sxr7171
It doesn't really help to run 2 instances since 1 instance will max out your CPU. Just running one might make it hotter since the first and second options use different techniques to torture the CPU and now your CPU cycles are split between the two methods. Try one at a time and see if you get higher temps. You kind of want to see how high you can possibly go. But 42C is darn good.

yes it does....he is using a HT processor, so he needs two open to use 100% of his cpu.

No, that's a fallacy caused by a mis-understanding of Windows' CPU-usage graphs.

wait are you saying Im not reading the cpu usage right ?


Thanks
Dan

 

[VirtualLarry]

Well, what I was commenting on was that with HT enabled, the CPU-usage reporting, will only show up to 50% on each graph, essentially. But in truth, one thread running at 100% with HT enabled, and another thread idling (most likely placed in the HALT state by the OS scheduler), will in fact max out the CPU, even though the usage graph will only show at most 50% usage. Running two instanced of Prime95 (or whatever), will show both graphs nearly maxed out, but won't max out the CPU itself significantly more. In fact, if both threads cause severe low-level CPU resource-contention, even though the graphs show more, the actual CPU usage will be overall less. (It's a bit technical and counter-intuitive, I know. Suffice to say, that with HT enabled, Windows' CPU-usage graphs are not to be trusted, because they aren't based on the physical CPU's usage, just each "virtual" thread.)

 

[ZL1]

Originally posted by: VirtualLarry
Well, what I was commenting on was that with HT enabled, the CPU-usage reporting, will only show up to 50% on each graph, essentially. But in truth, one thread running at 100% with HT enabled, and another thread idling (most likely placed in the HALT state by the OS scheduler), will in fact max out the CPU, even though the usage graph will only show at most 50% usage. Running two instanced of Prime95 (or whatever), will show both graphs nearly maxed out, but won't max out the CPU itself significantly more. In fact, if both threads cause severe low-level CPU resource-contention, even though the graphs show more, the actual CPU usage will be overall less. (It's a bit technical and counter-intuitive, I know. Suffice to say, that with HT enabled, Windows' CPU-usage graphs are not to be trusted, because they aren't based on the physical CPU's usage, just each "virtual" thread.)

actually I see one graph at low and the other at high, but overall at 50
mbm also shows it like this, most of time cpu1 is <10 and cpu2 is >90

thing is with two primes the cpu will show 100 everywhere and the temps will also raise higher, plus I can tell by monitoring vcore that cpu load is higher


Thanks
Dan

 

[rogue1979]

If you are running 1.75v on the core, your are putting the cpu at serious risk regardless of the temps.

1.70v absolute maximum on a Northwood, more like 1.65v to be safe. Raising the voltage above this just isn't worth the risk, lots of extra heat and only 100MHz more.

 

[ZL1]

Originally posted by: rogue1979
If you are running 1.75v on the core, your are putting the cpu at serious risk regardless of the temps.

1.70v absolute maximum on a Northwood, more like 1.65v to be safe. Raising the voltage above this just isn't worth the risk, lots of extra heat and only 100MHz more.

yes I was worried about that myself, see it only stays at 1.75 at idle, under load it drops to 1.625-1.65 so a droop mod should help
its back to 3.2 at stock v now, awaiting droop mod, if that doesnt help I think I'll be hunting for a new cpu


Thanks
Dan

 

[ZL1]

Originally posted by: rogue1979
If you are running 1.75v on the core, your are putting the cpu at serious risk regardless of the temps.

1.70v absolute maximum on a Northwood, more like 1.65v to be safe. Raising the voltage above this just isn't worth the risk, lots of extra heat and only 100MHz more.

ok played around with it and got it to do 1.65v, 1.575 in bios, but it boots as 1.65v (soft readings) and will drop to about 1.525-1.55 under load, I tried lower but it was failing prime when the cpu went under 1.5v
well I guess 1.65 is good for now, especially since previously when set at auto it used to boot with 1.6v and ran like that for a year, with a nice droop mod it should do 1.525-1.55


D