Well my system is finally complete, overclocked and tested for stability

[BigCoolJesus]

Well first, my system is finally done, put together, installed and running at what i want it to.

3500+Venice @ 2.7GHz (45C on high loads)
Epox 9NPA+ SLI
2GB RAM (running at 193MHz, best i could do with a divider)
Geforce 6800GT (415/1120)
HDA-Mystique 7.1 Gold
Hitachi 160GB SATAII
etc etc etc

Ok, first my results

After an 8 hour straight Burn-In test (ran both CPU and RAM benchmarks a total of 90 times straight each in Sisoft) my system didnt freeze or hang, and my CPU temps never went above 47C

Aquamark - 72,900
3Dmark 03 - 13,404
3Dmark 05 - 5778
PCMark 05 - 4460



SuperPI 32M - 29 mins 23 seconds

 

[pulsedrive]

Actually that 2nd Balloon ship test is a CPU test, so you might have some instability issues with your CPU, have you run Prim95 on it yet?

 

[BigCoolJesus]

Originally posted by: pulsedrive
Actually that 2nd Balloon ship test is a CPU test, so you might have some instability issues with your CPU, have you run Prim95 on it yet?

yea, i ran Prime95 and SuperPI

Prime i was only able to run for 3 hours, but it did fine, and ive ran SuperPi about 20 times, no errors either

if its an instability problem, then why would 3Dmark 03 or PCmark crash?

 

[Pabster]

3 hours isn't enough. Sometimes 12 hours isn't even enough. I always let it go 24 hours Small FFT and 24 hours Large FFT. For me that is on both cores. If it passes both, it is solid as a rock.

It isn't uncommon to see a failure after the 12 hour point but certainly 3 hours isn't enough time.

 

[BigCoolJesus]

Originally posted by: Pabster
3 hours isn't enough. Sometimes 12 hours isn't even enough. I always let it go 24 hours Small FFT and 24 hours Large FFT. For me that is on both cores. If it passes both, it is solid as a rock.

It isn't uncommon to see a failure after the 12 hour point but certainly 3 hours isn't enough time.

See, IMHO, i dont really care for Prime95...... i was even hesitant to download it (again) for this build, but i deicded to, and i just dont really care for it

sure, it will weed out all but the most solid CPU's, but considering i only use my computer for 3-4 hour strecthes that involve intensive tasks (gaming), and that the real test is to actually use the computer to test stability, Prime95 is a waste to me

Really, the only real way to test a CPU is to use it, just like you plan on to (ie, if your only using your computer for AIM, just use it for AIM, and if it locks up or reboots or freezes randomly, then you know its instable)

thats just my 2 cents

 

[Pabster]

Fair enough. Test as you wish. I'm just adding some Prime95 methodology from my own experience

 

[BigCoolJesus]

Originally posted by: Pabster
Fair enough. Test as you wish. I'm just adding some Prime95 methodology from my own experience

i completely understand and i know where your coming from with Prime

 

[daftpunkit]

I know most boutique companies run tests for at least 24 hours. When I got my Falcon-NW 4 years ago, with the computer came the benchmarks and comparison sheets. They ran Prime95 for 23 hours and some odd minutes I remember. If I remember right it should run at least 12 hour for stability. Somewhere I read some fanatics run it for 1 week! Personally if I were to build a new computer I would want to use it at least within the first few days...

 

[Spike]

I see prime-95 for what it is: a tool that shows absolute stabity. For the longest time my machine would fail prime-95 after 35 min but crashed only 3 times over the course of a year, and all those where due to video card drivers. (I once ran the computer for 3 weeks without a crash).

Finally I upped my cpu voltage a few months ago and now I can pass prime-95 for 12+ hours (I stopped it after that amount of time because I wanted to play a game). So sure, it works but you can defintly have a stable system and crash prime-95.

As for the OP, what was your 3Dmark05 score the first time you ran it? Or is that the same as the PCmark05 score you listed? I'm curious to see how that system performs compared to an older one like mine. Thanks!

-spike

 

[BigCoolJesus]

Originally posted by: Spike
I see prime-95 for what it is: a tool that shows absolute stabity. For the longest time my machine would fail prime-95 after 35 min but crashed only 3 times over the course of a year, and all those where due to video card drivers. (I once ran the computer for 3 weeks without a crash).

Finally I upped my cpu voltage a few months ago and now I can pass prime-95 for 12+ hours (I stopped it after that amount of time because I wanted to play a game). So sure, it works but you can defintly have a stable system and crash prime-95.

As for the OP, what was your 3Dmark05 score the first time you ran it? Or is that the same as the PCmark05 score you listed? I'm curious to see how that system performs compared to an older one like mine. Thanks!

-spike

my 3Dmark 05 score yesterday was 5879 and i was running at 2.7GHz with Vcore of 1.72

Today im running 2.7GHz with Vcore of 1.7


if its due to the Vcore, ill live with it, no need to push it more then i have to just for a benchmark


Well im off to finally play BF2 (been waiting the last week to finish putting everything together, im excited)

 

[Ike0069]

Glad to hear it's up and running good.
Very nice CPU OC, and even better on the VC.

 

[GuitarDaddy]

Originally posted by: BigCoolJesus

Originally posted by: Pabster
3 hours isn't enough. Sometimes 12 hours isn't even enough. I always let it go 24 hours Small FFT and 24 hours Large FFT. For me that is on both cores. If it passes both, it is solid as a rock.

It isn't uncommon to see a failure after the 12 hour point but certainly 3 hours isn't enough time.

See, IMHO, i dont really care for Prime95...... i was even hesitant to download it (again) for this build, but i deicded to, and i just dont really care for it

sure, it will weed out all but the most solid CPU's, but considering i only use my computer for 3-4 hour strecthes that involve intensive tasks (gaming), and that the real test is to actually use the computer to test stability, Prime95 is a waste to me

Really, the only real way to test a CPU is to use it, just like you plan on to (ie, if your only using your computer for AIM, just use it for AIM, and if it locks up or reboots or freezes randomly, then you know its instable)

thats just my 2 cents

3dmark05 is a very tough test, if its locking up in the second ballon CPU test then your not completely stable. But per your above philosophy pick a game you like to play and play it for several hours and if it doesn't lock up then forget about 3dmark05 just like you forget about Prime for many hours. But don't be surprised when you suffer sporatic lock ups or crashes.

Personally I want a 100% stable system, but many people don't care. And your logic that since you don't run intense apps for more than a few hours there is no need to be 12-24hours prime stable is flawed. Stability is not time related, if your system is unstable it may crash/lockup/error at 18hrs 36mins, or it may crash in 15mins it makes no difference. It may complete 12hrs of prime one time, and crash after 2 1/2 hours the next time. Personally I don't call an overclock stable until I've run 3 or 4 successful overnight 12+ hours prime runs, and stressed the system with day to day usage for a couple of weeks.

 

[Bona Fide]

I personally am not a fan of the "Xtremesystems" stability philosophy. If it can run SuperPI 1M, it's stable. WTF kinda methodology is that...

Anyhow, if you want to finish 3dMark05, you need to stabilize your CPU. Otherwise, just forget about the test, and don't complain when your computer DOES crash.

 

[BigCoolJesus]

Originally posted by: Bona Fide
I personally am not a fan of the "Xtremesystems" stability philosophy. If it can run SuperPI 1M, it's stable. WTF kinda methodology is that...

Anyhow, if you want to finish 3dMark05, you need to stabilize your CPU. Otherwise, just forget about the test, and don't complain when your computer DOES crash.

i wont complain, thatll just let me know i need to back down the OC, which is a lot better then going by a rounding error in Prime95

 

[BigCoolJesus]

Finally got 3Dmark 05 to run, just had to bump up the Vcore to 1.72


Im gonna try Prime95 tonight, so if it will agree with me........

 

[n7]

I like to use games as my main stability test, but Prime95 should at least be run overnight.

I find that UT2k4 finds instabilities in less than half an hour of play though, rather than like 5 hours or whatever with P95, so i use it as my main measurement.

Also, man, your vcore is too high.

I really wouldn't go higher than 1.65V.

I know your temps seem good, but i don't believe they are accurate.

 

[BigCoolJesus]

Originally posted by: n7
I like to use games as my main stability test, but Prime95 should at least be run overnight.

I find that UT2k4 finds instabilities in less than half an hour of play though, rather than like 5 hours or whatever with P95, so i use it as my main measurement.

Also, man, your vcore is too high.

I really wouldn't go higher than 1.65V.

I know your temps seem good, but i don't believe they are accurate.

No, the temps are definatly accurate.... im not only going by Epox's monitor probe, but i have an actual temp probe set-up on the heatsink, and thats never gone past 34C, so i know my temps arent to high

 

[Pabster]

You just set the probe on the base of the heatsink? That isn't very accurate at all.

You'd need to drill a small hole in your heatsink and get the probe right to the core to have a semi-accurate idea of how hot the chip really is.

 

[BigCoolJesus]

Originally posted by: Pabster
You just set the probe on the base of the heatsink? That isn't very accurate at all.

You'd need to drill a small hole in your heatsink and get the probe right to the core to have a semi-accurate idea of how hot the chip really is.

Well the purpose of this probe is to see the temp of the fins, and the temp of the air around the CPU.... i really didnt need it to be accurate


I just ran another 100 cycle Sisoft Burn-In Wizard benchmark, and this time Epox says the highest temps i had were 45.0C, and the average was 44.2C, so even with a 5-7C leeway, im still well under 60C (not to mention, i doubt ill be playing BF2 for 6 hours straight)

So i know the Vcore is extremely high, but as long as the temps stay below 60C on the CPU, theres nothing to worry about (except shortage of CPU lifespan by 50%, but i wont have this CPU for more then a year, as i upgrade often)




Heres all of my temps for you to say are impossible to have with this Vcore.........

Sensor Properties:
Sensor Type Winbond W83627THF (ISA 290h)
GPU Sensor Type National LM89 (NV-I2C 4Ch)

Temperatures:
Motherboard 35 °C (95 °F)
CPU 34 °C (93 °F)
Aux 39 °C (102 °F)
GPU 48 °C (118 °F)
GPU Ambient 34 °C (93 °F)
HardDrive 31 °C (88 °F)

Voltage Values:
CPU Core 1.72 V
Aux 3.38 V
+3.3 V 1.52 V
+5 V 5.08 V
+12 V 10.88 V
+5 V Standby 5.14 V
VBAT Battery 3.12 V

 

[rise]

1.72v is alot for a venice imo.

as opposed to long runs of prime, you could try occt for a 30 min stability test or set a 1 hour time or something. if you have mbm5 or speedfan it also gives you a nice graph of your temps/volts throughout the run.

 

[Pabster]

Originally posted by: BigCoolJesus
Heres all of my temps for you to say are impossible to have with this Vcore.........

I do believe your temps are inaccurate, and, frankly, I don't see the point in measuring temps if you don't care that they're inaccurate.

Using a thermal probe inside the heatsink (as close to the core as possible) is fine; just throwing the probe "around the fins" is not giving an accurate result. Which to me equates to a big waste of time.

But, like I said, if you don't care that the temperatures are wrong, that's fine

 

[Bona Fide]

+3.3 V 1.52 V
+5 V 5.08 V
+12 V 10.88 V

RED FLAG!!! Look at the flux on your power rails. That's not good at ALL...

 

[BigCoolJesus]

Originally posted by: Bona Fide

+3.3 V 1.52 V
+5 V 5.08 V
+12 V 10.88 V

RED FLAG!!! Look at the flux on your power rails. That's not good at ALL...

No, those arent noteworthy at all
i used Everest home edition for the temps, and it reported those voltages, which are way off


My real voltages are

+3.3 @ 3.37V
+5 @ 5.1V
+12 @ 12.11V



and Pabster, the temps i reported are actually ones from the hardware monitors/probes already installed on the Epox and harddrive (ie, actual probes set up in actual areas)
I know the hardware monitor can be +/- 5 or so, so given the benefit of the doubt, my CPU is only 53C at extreme loads, nothing wrong with those temps (especially given the Vcore im using)



And rise4310, whats occt?