NEW Benchmark scene (Max 8 Vray RC3 required)

[n6]

n6 is correct there are also X3xx quad core xeons too.

But I think they are just limited to 1 CPU per mobo?

Whats your temps like n6?

I current have my E6420 @ 3.2ghz and temps are low to mid 30's idle and mid 40's under load. I would imagine yours would be around 10C higher due to the extra 2 cores.
Im using a thermalright ultra 120 extreme too.

yep only 1 CPU but trust me : now, dual mobos are old fashion things... they are very expansive (the mobo it self, heatsinks; the PSU etc...), not easily overclockable... and I m a dual processor geak since I was running a dual Pentium 166Mhz... so trust me

Intel is curently working on a 80 cores processor...

Regarding my setup,
With stock heatsinks; its near impossible to increase the Vcore.
Mine is at the upper Intel's limit : something near 1.38V and the temp at this voltage rize extremely quickly...and high : near 75°C... but think its 4 processors in the same place!
All you need is cooling, strong cooling and Vlado to release... the next release

[DaForce]

Yeah im well aware of the expensive xeon boards.. they cant overclock much either.

Hence me getting a Asus p5b and a E6420 till the quads drop price in july.

wow.. high temps. Are you going water cooling or a good air cooler like the one I have?

Here is my heatsink, look how small it is compared to the memory.. and thats corsair dominator.. a fair bit taller than normal memory.
http://www.apex.net.au/~daforce/vray/DSC01854Large.JPG

[teabag]

DaForce: yep lost planet is playable under the xp as well but u cant get it run under dx10 ..have seen the difference in the demo and its huge!!!

[DaForce]

what.. a visual difference? really.. i would be suprised. Looks just like the 360 version when i played the demo on xp. It just ran like crap.. 2560x1600 on a 8800gtx mind you. I doubt DX10 would make it look any better, would run worse.. given how vista zaps FPS and how bad company of heroes ran after the DX10 patch.
Are you able to post some screens..im really keen to see the difference.

[n6]

Yeah im well aware of the expensive xeon boards.. they cant overclock much either.

Hence me getting a Asus p5b and a E6420 till the quads drop price in july.

wow.. high temps. Are you going water cooling or a good air cooler like the one I have?

Here is my heatsink, look how small it is compared to the memory.. and thats corsair dominator.. a fair bit taller than normal memory.
http://www.apex.net.au/~daforce/vray/DSC01854Large.JPG

you can stay with air cooling but your overclocking and stability would not be as good as watercooling.

Asus p5b is a very good mobo but its now an "old" product the Asus p5k or p5k3 is a better choise since it would allow you to get a much higher FSB (because of the P35 chipset).

Regarding the memory, dont look at the compagny name (Corsair, OCZ, A-DATA etc...)... but at the chips producer and serie.

I have 4 gb of G-Skill PC6400 (HZ chips)that can run at the speed of PC8800...

[DaForce]

This "old product" is capable of 550FSB

[n6]

This "old product" is capable of 550FSB

cetairnly not with quad core procesoors ...

[DaForce]

Well that I cannot say, but heat is your biggest issue. Get that under control and then your golden I would think.

[n6]

But you are right : this is a very good choice compared to a QX6700 or a Q6600 that have a lower quality (need more voltage for the same speed) than the X3220 or X3210.

Hope you are right - my experience is that the X3220 runs exactly the same as my Q6600s - same voltage, same heat, same top end. Let us know how you get on.

Here is my Q6600 on a P5B deluxe at 3600 MHz running Vista 64 bit, MAX9 - just cracks the 2 minute mark.

I will be running it 24/7 at a little over 3500Mhz.



Different times on different OS / Max versions

Interesting stats on another machine Q6600 @ 3420 MHz, 1G ram

Vista64bit - Max9 64bit = 2mins 5.9 sec
Vista64bit - Max9 32bit = 2mins 12.6 sec
WinXP-32 bit - Max9 = 2mins 11.4 sec
WinXP-32 bit - Max8 = 2mins 29.9 sec

All tests best of 3 frames, same machine, same conditions, netrendering.

You should check you CPU stability and thermal-throtling because its not normal that you get slower render time compared to mine at a lower speed.

Even if VRay render this test on your machine without crashing MAX or XP, that doesnt mean that you CPU is computing without errors or big thermal throtling...

[davexl]

You should check you CPU stability and thermal-throtling because its not normal that you get slower render time compared to mine at a lower speed.

Even if VRay render this test on your machine without crashing MAX or XP, that doesnt mean that you CPU is computing without errors or big thermal throtling...

I only use settings that have passed 24+ hours of Orthos or Prime 95, and 12 hours of Memtest. They are stable.

You are using Vray RC4, I am using RC3 - look no further!

My machines are not thermally throttling - that would be pretty obvious by the times, and also you can check with Rightmark RMclock.

Dual cores throttle at 82C coretemp - 3C below the "tjunction" figure quoted for my dualcore 6600s of 85C. (personally tested on 3 machines)

Quad cores have a tjunction of 100C coretemp, so I guess they would throttle at about 97C coretemp. I have not run any quadcores that high, but I have run them stably overnight peaking Prime95 at 91C. Vray renders about 12C cooler than Prime95 8k FFTs, so if I can pass Prime95, I have a 12C buffer running VRAY.

P5B vs P5K

Nice theory n6, but practice and testing always beats theory.

As for overclocking quadcore 6600s, I have noticed no significant difference between the P5B and the P5K - both top out at between 3.5 and 3.6GHZ on air cooling. (awesome air cooling like the new Ultra-120 Extreme = what used to be watercooling territory)

The limit seems to be the CPU, not the mb. Likewise the Xeon 3220 performs the same as the Q6600s.

The P5K Deluxe is worth getting because it will run the upcoming Penryn CPUs, and it has voltage damping.

But even voltage damping can be hacked for the P5B - I discovered the pencil mod for the P5B over the weekend, which is quite insane - rub a 2B pencil over one tiny resistor and it inverts the voltage droop under load! So your voltage goes up under load, rather than drooping. (So you run at a more power efficient voltage when idle, and only use the voltage when needed)

Vdroop of 0.072 volts has now turned to a BOOST of 0.048 volts

Before pencil mod:

1.6000V BIOS = 1.552v idle = 201watts, P95 8k load = 1.480v = 359watts

After pencil mod:
1.4625V BIOS = 1.440v idle = 181watts, P95 8k load = 1.488v = 367watts


http://www.xtremesystems.org/forums/....php?p=2241519

DaForce - will post BIOS settings when I have finished testing. I tend to find my max performance voltages with a lot of things turned up, then over time start turning potentially unnecessary voltages back down to see what was really needed, so I don't want to tell you to scorch your mb if not needed

[DaForce]

Thanks David, much appreciate. GOod info there too.

My rig is stable at 3.2ghz at the moment, but I had to up my PCi-E mhz to 108 to get the PCi-E link to run at 16x instead of 1x. Seems like a fairly common problem.

I have yet to run a over night prime95 run, as I have been busy so just running at stock at the moment. My mate is able to get his E6420 to 3.6ghz stable as well. I had copied his settings and my machine failed to boot. Only differences were I have P5B deluxe he has P5B-E, I have Corsair 800mhz C4D he has just corsair 800mhz C4
So I still need to tweak the 3.6ghz bios settings a little further.

Temps are great with the ultra 120 extreme, 3.2ghz yeilded temps no higher than low to mid 40s under full load.
And as its winter now in Aus, on a cold day the other day running at stock speeds, the cores were at 20-21c and cpu at 24c. Quite outstanding.

[lukx]

Can I suggest the Xeon 3220 or Q6600 instead of the QX6700? I have 3 of them, and they all do between 3456MHz and 3546MHz on air cooling with the Ultra-120 Extreme. That is on the P5B Deluxe and the P5K deluxe, forget the P5W_DH, it hates quadcore overclocked. Which was a shame, as I had 2 of them.

hmm.. I got P5W DH and some time ago was thinking about getting P5B but read on Hard[OCP] that it's better to wait for x38 chipset. So my question is...let's say without bumping vcore how fast can run Q6600 on P5B while beeing stable? Right now I'm running my Q6600 on P5W DH as 3000Ghz stable(Scythe Ninja Idle=42 C, Full Load= 53 C vcore auto in bios. Also davexl what are your settings in bios for your 3456 speed? Did you bump Vcore and if so how much? 1.5? more? Isn't it true that those new P5K boards are really hot also?

[davexl]

hmm.. I got P5W DH and some time ago was thinking about getting P5B but read on Hard[OCP] that it's better to wait for x38 chipset.

Always something better on the horizon... but I think the x38 needs $$ DDR3.

Chipsets aren't really a big thing for me, unless they magically unlock overclocking. (rare) VRAY renders within 1% on the 965 chipset P5B and the 975 chipset P5W, so BFD

So my question is...let's say without bumping vcore how fast can run Q6600 on P5B while beeing stable? Right now I'm running my Q6600 on P5W DH as 3000Ghz stable(Scythe Ninja Idle=42 C, Full Load= 53 C vcore auto in bios.

AUTO is already overclocking your vcore- it scales it up with clock speed. Run speedfan, systool, or CPU-z to see what your vcore readings are. - they will droop under load - see my post 2 above for my settings.

But the P5W has a hard limit around 3.1GHz on quadcore, so you are basically there already.

Also davexl what are your settings in bios for your 3456 speed? Did you bump Vcore and if so how much? 1.5? more? Isn't it true that those new P5K boards are really hot also?

Full BIOS settings to follow soon, I am still testing things, which has to be done between jobs, and things are busy.

But basically, disable everything in the CPU settings, disable static read and try Vcore of between 1.5 and 1.6 in the bios, depending on your cooling, keep load coretemps under 80C at Prime95 small FFT. (get latest version of Prime95 that loads all 4 cores = v25.2 or later) Set north bridge voltage to 1.5+, and make sure you have adequate cooling of your case - like a 120mm fan blowing over the ram and northbridge. A good power supply helps.

Thread for P5B settings:

http://www.xtremesystems.org/forums/...d.php?t=115217

3.3 Ghz should be very easy on air cooling, at mild voltages.
3.4 is very doable, and good for summer
3.5 is tough but achievable outside of summer.
3.6 is about out of reach.

(I am talking Prime95/memtest stable, I can run 3.6 GHz that renders vray easily, but it is risky - I use these things to earn a living, so they have to actually work, and not blow up, or corrupt data. I won't push my godbox as much as I might a renderslave.)

Don't even think of overclocking unless you use a program like Ghost or Acronis Trueimage to back up your system partition. (I run 20-40G system partitions, which back up to about 6G)

P5K are very nice, but like all new boards they have the odd issue (which is why my P5B is my current godbox) Don't expect any difference in overclocking ability to the P5B on quadcore.

[DaForce]

Have you found with the P5B that after changing some bios settings that upon saving and exiting the bios it just sits there in the bios.
I have had that happen a few times and when i hit reset it does the old double reset then back to windows. Rather odd.

Also for my PCI-E link to be 16x I need to raise the PCI-E freq to 110mhz otherwise it stays at 1x, it used to work with 108mhz but now needs 110mhz. Annoying

So if any of you overclock with your P5B and notice slow assed video performance, check with cpu-z if your PCI-E link is on 1x instead of 16x

[lukx]

Always something better on the horizon... but I think the x38 needs $$ DDR3.

I will stick to my P5WDH till X38 hit he market in August...because it will support ddr2 http://www.theinquirer.net/default.aspx?article=40169