Core Duo - wow

[Greg Hess]

AJ,

 

Woodcrest or conroe? If so, which board?

[AJLynn]

Allendale. It's an E6400. MSI 975X rev2.1 board with 7.1 BIOS, OCZ Platinum 6400 RAM and the same HSF you have.

 

And, yeah, turns out those OpenGL scores aren't errors, it's just really fast.

[Greg Hess]

Any reason you stopped running 2.8? Was going to give you some random things to try if you wanted to try going a bit higher, or pass a dual prime95 test at 2.8.

 

If you can modify it, try adjusting the MCH voltage to 1.65, and the FSB voltage to 1.27. Raising the pci-e to 105 megahertz also seems to do wonders.

 

Also make sure to check your memory megahertz via cpu-z or pc-wiz to make sure you aren't getting too high. 1.35 is the voltage spec for the core2duo's according to intel...but their E6700 shipped at 1.4. So they've got a bit of room if you just want to bump it a weee bit.

 

Watch out though, it gets addictive REALLY fast.

[AJLynn]

I know what you mean about addictive. I did putz with it some more and got it to almost stable at 3.12 (stable enough to run Cinebench but not to stay up and non-glitchy very long, and it still never went above 50C core temp). I have it at 2.8 (350x8) with memory at 1:1 and 4-4-4-15, where it's very comfortable though I may try for 4-4-4-12 (does that actually make much difference?)

 

I think my problem is psychological. It hits speeds that I don't think should be possible, so I don't believe in them Also, it's so easy that it makes me nervous, like I must be doing something wrong because it shouldn't work on the first try, it should take more effort than that, so I'm waiting for something to break.

[Greg Hess]

AJLynn,

 

Haha! Nice! No I don't think 4-4-4-15 vs 4-4-4-12 makes a big difference. The difference between 3-3-3-10 and 4-4-4-12 wasn't very noticable, so I doubt changing a single value will.

 

The temp ceiling for these processors is actually quite high, so 50C is actually pretty good. Mine idles around 44-46C, and spikes to 55-58C (depends on room ambient) during rendering or gaming tasks.

 

Under priming though, it actually got all the way up to 68C, which is still quite good, as it appears my particular chipset "overreports" the temperature by nearly 10C on purpose. (To try and keep people from going nuts). Some of these boards are also intentially undervolting the #'s inputed for overrides, to keep things more sane. I would start to worry (or back off), if temps start going into the 70 range. Remember that most of these boards are overreporting...but still, that number is getting to high for comfort!)

 

3.12 is awesome though! Thats fantastic!

 

What I would do is try bumping the cpu voltage up 0.05 to 0.1 (I'd be wary of going too far past 1.45 volts, with 1.5 around the max)...up to 1.4 is quite safe (same as my E6700 was stock). And use dual prime95's as a stress test.

 

http://mersenne.org/gimps/p95v2414.zip

 

Unzip this twice on your desktop. In each folder, run prime95, say "just stress testing", then select processor affinity and set one prime to core 1, and the other prime to core 2. Then start "torture test" and choose "lots of ram tested" for one of the cores, and "maximum heat/work" for the other core.

 

Let run for roughly 12-48 hours, as time permits. If it detects an error on either run, you need to adjust voltages or frequency to stabilize the machine. It is VERY hard to pass a dual prime95, as it won't allow an error from ram, or either of the two cpu cores...so if there is any type of possible stability problem, it'll show up during this stress test. You'll also see MUCH higher temps then rendering, nearly 10-12C higher...it really pushes the comp to the limit.

 

Basically if it passes that, the computer won't ever have any stability issues, as long as you don't mess with it ever again .

 

(I have an E6600 @ 3.333 @ 1.5v [13 hours dual prime95 stable...i got tired of waiting and stopped it at that point...lol]) Can't quite get to 3.4! (3.333 is 1033 on the cinebench)

 

Whats really odd, is the E6700 I bought for work, can't do more the 3.2 without getting excessively hot. I think the heatspreader on the cpu isn't 100% flat =/.

[AJLynn]

I'll try those things, thanks.

[tecton3d]

nice Andrew... I feel like that little kid in the 'hood without a nice, new, and shiny bike to ride with the rest of you while you ride around me and poke fun;)

 

thought I'd share that some Kentsfield eng. samples are out and are stomping most other systems out there... see the max scores:

http://www.tomshardware.com/2006/09/10/four_cores_on_the_rampage/page11.html

 

of all the forums out there, xtremesystems.org seems to be on top of things regarding Conroe and the next gen Kentsfield/Clovertown ... I didn't need to sleep tonight anyway:p

[AJLynn]

If you're reading and writing about computer parts that you can't get yet at 4AM on a Monday morning, you might be a grad student.

[tecton3d]

If you're reading and writing about computer parts that you can't get yet at 4AM on a Monday morning, you might be a grad student.

 

LMAO!

 

There's a chap over at 2cpu.com that's saying Kentsfield (quadcore Core2 Duo desktop) and Clovertown (quadcore Xeon workstation/server) chips are drop in cabable with current 975X and current Woodcrest, including MacPro, mobos... He seems to know well enough what he's talking about...

 

"Kentsfield and Clovertown are Q4 06. Kentsfield is apparently October with press launch at the end of this month."

 

if the release of Kentsfield is so close, I may go with a low end 6300 + a water cooling setup for the time being and snag a quadcore when they arrive in Oct/Nov.

 

here's quadcore Kentfield @ 3.6ghz!!

[Greg Hess]

I can only hope that it does run...though you are correct that it seems that conroe support = quad core support.

 

Man would that be a nice upgrade route.

[AJLynn]

Heh. That HSF installation with the 4 posts you have to push in was such a pain, I wouldn't call replacing the chip an easy upgrade at all

[Greg Hess]

Did you put the hsf on before or after installing the mboard?? My case actually has a removable mboard tray, so it wasn't too difficult to swap it on and off repeatedly.

 

If you think that hsf install was bad, you should see the thermaltake big typhoon...I don't recommend that to anyone! The install is horrific. Well maybe not horrific...just really annoying and time consuming.

 

Also was able to run the max8 demo and vray to complete the benchmark. Here was my score at 3.3!

 

[AJLynn]

3.3 - nice. Damn that's a fast render time. I never had Vray 1.47 on this so I can't do the same benchmark - 1.5 gives different results.

 

The case does have a pull out MB tray, which makes it easier - I can't imagine doing one of these things in the case - but it's still a pain. It was my first time using one of this type so it wasn't easy figuring out whether the pins were in correctly and how to get them in enough. It helped that I could see in the gap between the board and the tray and see how far they went through.

[Greg Hess]

Well you should know that the artic freezer's mounting system is identical to the stock intel hsf .

 

You are "supposed" to be able to just press it in, then click each point down. Of course that only works if they are in the correct position to be clicked. Took me awhile to figure that as well, as this was my first socket 775 system as well. (I haven't built an intel machine since socket 478 cause everything sucked. )

[tecton3d]

Heh. That HSF installation with the 4 posts you have to push in was such a pain, I wouldn't call replacing the chip an easy upgrade at all

 

"easy" only considering you'd be doubling your # of cores for a little bit 'O labor:p

[tedesco]

FUCK, just got a PentiumD 950 and was thinking I was the bomb....

 

Well, got 06:38

 

Pentium D 950

 

3Gb RAM

 

Will buy a Dual Core next time.

[Chad Warner]

I just got my Dell Dual 3 Ghz 5160's with 4 gb of ram and the test scene rendered in 1:56! I am so excited right now.

[AJLynn]

Tedesco-

 

The Pentium D is a dual core... which makes the Netburst architecture look really bad in comparison.

 

I feel your pain. My firm bought me a brand new Pentium D 2.8GHz a few weeks ago, just before I showed up. I'm trying to figure out how to spend as little time as posible setting up renders, so I can concentrate on design, so I'm thinking Maxwell - and Maxwell and slow CPUs just don't mix.

[aaron-cds]

I just got my Dell Dual 3 Ghz 5160's with 4 gb of ram and the test scene rendered in 1:56! I am so excited right now.

 

That's with xp 64, right?

[Chad Warner]

Yeah.

[cullen]

has anyone had problems opening the benchmark file in VIZ?

I'm running VIZ 2006 and get the following error on open:

[Zdravko Barisic]

Yeah.

 

Yu use x64? Any issues?

How much memory MAX can handle?

More then 2GB? 3-3.5GB?

Works fine just in 32bit OS?

[Chad Warner]

Yu use x64? Any issues?

How much memory MAX can handle?

More then 2GB? 3-3.5GB?

Works fine just in 32bit OS?

 

Yes I'm using x64. No major issues, just can't find some drivers.

I don't know what the maximum amount of ram that max can handle, but the machine i have uses 4 GB.

[frezer]

ok, probably all of u guys know about that, but if somebody didn't read that, here's the real bomb... check the max render tests:

 

http://tomshardware.co.uk/2006/09/10/four_cores_on_the_rampage_uk/

[Greg Hess]

Not to take away from your post...but...

 

Tomshardware works in the same building as intel employees, and generally works with, and even hangs out with said employees. Thus there is always a very heavy pro-intel bias over at toms, up to the point of benchmark manipulation in some previous articles (to showcase intel over amd). Thus I'd generally recommend you use other websites to look at the performance of cpu's, as the main reason toms tends to be first with cpu comparisons, is because they practically work in the same building as intel .