53 replies to this topic

[_TW_]

Mul - Let me know what sort of tests you'd like done; I'd be more than happy to compare results with my E8400.

[Mul.]

Well it's been a week and I still haven't updated you with anything substantial. Once again, I won't be because I'm still waiting on a fresh copy of XP Home to arrive in the post, however I've managed to get into Safemode with Networking with my existing copy. Other bad news is that my DDR2-1066 RAM is faulty, so I'm using my Patriot PC2-6400 CAS4 RAM. I wanted to get cracking with overclocking immediately but seeing that no temperature monitors work properly in Safe Mode, I've had to be quite conservative.

So here's some SuperPi runs. I would like to remind you all that you cannot directly compare Intel vs AMD with this app as historically it runs much better on Intel rigs when in reality the difference in all other apps are much smaller.

For now I've carried out all overclocking by increasing multipliers with the stock CPU Voltage of 1.325V. CPU-Z does not read all information in safe mode so you'll have to take my word for the settings that I've applied

Stock
200MHz Base Clock * 14 CPU Multi
2800MHz CPU
1600MHz Northbridge Frequency (8x NB Multi)
DDR2-800 4-4-4-12 1T




Overclocked

200MHz * 16 CPU Multi
3200MHz CPU
1600MHz Northbridge Frequency (8x NB Multi)
DDR2-800 4-4-4-12 1T



3200MHz CPU / 2000MHz Northbridge Frequency



3200MHz CPU / 2400MHz Northbridge Frequency



3200MHz CPU / 2600MHz Northbridge Frequency





And there we have it for now.

Overclocking from 2800MHz to 3200MHz reduces 1M time by just under 3 seconds.
What's interesting (and confirmed by overclockers of Phenom and Phenom II chips) is that these CPU's do like high NB speeds, where my Pi 1M time has decreased by about a second just by increasing NB speeds.

Obviously no conclusions can be made by this sort of testing as I haven't really done a lot. Hopefully my copy of XP will arrive on Monday/Tuesday and then the real testing will begin

Edited by Mul., 07 March 2009 - 09:10 AM.

[Jonay]

FYI

[Mul.]

This is old news! I watched the live stream of those infamous overclockers pushing those Phenom II's to 6GHz+ on extreme cooling back in January!

Edited by Mul., 07 March 2009 - 09:40 PM.

[VFR_Pierre]

Nice Mul.

Seems like a very high multiplier. Is it unlimited? I guess it must be really awesome at OCing then. But since the performance (oddly) gets better as the NB frequency gets higher, it's probably better to see how high you can go with the NB frequency, then the multiplier, right?

Sorta weird that the performance is increased just by raising the NB frequency. Why is that?

Can't wait to see how far you can push the CPU when you get your copy of XP.

How much voltage can you stuff into these CPUs to keep it safe?

Edited by VFRDUDE123, 07 March 2009 - 09:52 PM.

[_TW_]

The Black Edition CPU's from AMD have unlocked multipliers.

Mul - Nice first tests.  Would you call this a wise upgrade from your previous C2D machine?  Can't wait to see more tests; care to try 3DMark06 at all?

[Mul.]

As I say it's hard to say for now. Once all my Core 2 Duo components are sold, all in all the upgrade would've costed me about £40 less than it would have to have bought an E8400 and that's it. I suppose that's the CPU I should be comparing this platform against!

I plan on buying Flight Simulator X and Crysis to add to my set of benchmarks and findings. If you have an 8800/9800 based graphics card sitting around, your rig would be ideal for comparisons

Edited by Mul., 08 March 2009 - 12:03 AM.

[_TW_]

Not at the moment Mul; could try my best to grab one in the next weeks.

[Mul.]

I couldn't ask you to lose cash for the sake of testing but of course if you find yourself with an 8800GT/9800GT in the coming weeks do let me know and we could try and carry out a similar test to what PiP and I carried out a while back comparing Socket 939 Opteron Dual Cores with the Core 2 Quad Q6600.

[Mul.]

We're in Business!



What's shocking is quite how cool these chips run. The overclocking shall commence from now.

[VFR_Pierre]

Is that full load? That is cool! I hope you get 4.25+GHz, would be nice.

Edited by VFRDUDE123, 09 March 2009 - 04:01 PM.

[Mul.]

So far

3600MHz / 2600MHz Northbridge / DDR2-800 4-4-4-12





Single Threaded Cinebench - 3261
Multithreaded - 8930
Multicore Speedup - 2.74x



I'm going to spend a little longer trying to push it further but will be working towards getting it 3.6GHz stable overnight.

Edited by Mul., 09 March 2009 - 05:00 PM.

[VFR_Pierre]

Very nice start Mul.

I'm very interested to see how far this thing'll go. Very nice temps it seems, almost unbelievable. How come the temps of the 3 cores are all always the same?

Edited by VFRDUDE123, 09 March 2009 - 06:09 PM.

[Mul.]

I'm quite sure that it's because there's only a single temperature sensor under the heatspreader.

It's getting a bit late over here and I have coursework to get on with so I think I'm going to stay put at 3.6GHz and work on getting it stable. Will keep you all posted.

[tropicalfish]

What heatsink is this? Scythe Ninja?

[Mul.]

Yep, Scythe Ninja with it's standard 1200rpm fan.

Bad news chaps. Woke up to this



Core0 failed after 8hr59min of stress testing, while Core1 and Core2 were chugging along just fine right up to the point where I halted the test at 9hr14min. This is still inadequate, especially when I'll be folding 24/7. Shall make some slight tweaks and see what the result is when I return from lectures.

[Mul.]

Seems as though there were no hickups this time round. In other news, I was able to boot in at 4.0GHz but promptly restarted after starting Prime95 and this trend appears to be there all the way down to 3.6GHz itself so far. My prediction is that this chip will hit about 3.8GHz but any higher won't be all that possible due to the amount of volts that I'm having to put through it already. Have also had to pull back Northbridge speeds to default but will start to raise that, drop the CPU Multi and start raising the base HTT clock and getting some better performance from that perspective too.

So far, it's overclocking quite well, but this CPU in particular hasn't quite met my expectations. It was to be expected but considerable voltage increases are required upwards of 3.4GHz. I can't comment on performance just yet either.

Pricewise, this CPU's competition is:

Core 2 Duo E7400 2.8GHz $119
Core 2 Duo E7500 2.93GHz $145
Core 2 Duo E8400 3.00GHz $165

The Phenom II X3 720 has a third core going for it and the potential for it to hit as high as 4.0GHz on air. The E7000's are similar clock for clock, while the E8000's are a tad quicker and while they have one less core, they will overclock higher with 4.0GHz being somewhat guaranteed most of the time. We shall see what findings can be had after finishing up on the overclocking.

[VFR_Pierre]

Oh, alright Mul. It would have been nice to see this chip hitting 4+GHz.

I wish there was an option to OC only one core at a time, with different volts for each. Maybe some day....

[Mul.]

Actually, I am able to adjust CPU Multipliers to individual cores using AMD overdrive.

[VFR_Pierre]

Mul., on Mar 10 2009, 03:57 PM, said:

Actually, I am able to adjust CPU Multipliers to individual cores using AMD overdrive.

Oh wow, that is pretty cool.  

Couldn't you just do that for one core that is unstable?