|
Advertisement:
|
AMD Phenom II X4 940 Black Edition and Radeon 4670 |
Join the community - in the OCAU Forums!
|
Overclocking, Sandra Benchmarks
Overclocking:
No point beating about the bush, this is what we came to see, and I hope I can give you an insight into what the Phenom II can provide for overclockers. As I mentioned previously I’ll be dipping my toes into sub-zero cooling with dry ice, and comparing that with my homebrew water-cooling setup which consists of: a Storm G4 water block, a Swiftech MCP655 pump, and a ghetto-fabulous Toyota Cressida heater-core which sits atop my Antec 900 case, with the cases 200mm fan blasting a bucket-load of air through it.
Firstly, some BIOS confusion restricted my results a little. I was instructed by AMD to use the BIOS supplied with the MSI board for testing. This particular BIOS revision (oddly named v0.0) refused to work with the Kingston HyperX and the 1:2.66 HTT:DRAM divider. After spending quite a bit of time trying various things and wondering why the OC gods had cursed me, I settled with using the 1:2 divider and hence DDR2-800.
It was only later that I tried the latest v1.4 BIOS which fixed things up and DDR2-1066 was working, albeit with a slight hitch. Both DFI and MSI boards are for some reason inclined to set particularly weak timings with DDR2-1066, namely 5-7-7-20 instead of the SPD’s programmed 5-5-5-15, and this happened with both sets of RAM. Manually changing this is not much of a hassle, but it kind of renders moot the whole point of SPD and EPP.
Apart from these hiccups, the rest of the process was quite smooth, and both boards impressed me overall. Water-cooling allowed me to hit 3.8GHz at 1.6V, and a 2.2GHz North Bridge clock with 1.3375V, which are the speeds I used for testing in this review. Temperatures were excellent, staying under 50 degrees fully loaded.
With dry ice (thanks to OCAU member aussie_revhead for the pot) I managed 4GHz at 1.65V, and cranked up the CPU’s North Bridge one notch to 2GHz. At one point I had 4.6GHz with 1.7V, but I couldn’t get it stable, or anything over 4Ghz for that matter.
After a good 6-7 hours I was finally starting to feel a bit more comfortable with sub-zero, and then I ran out of dry ice! After a fairly unspectacular session I now have a new deal of respect for all those hardcore overclockers out there like Team.AU. Go hard lads!
Sandra Lite 2009.SP2
Note that for all Sandra tests, the “Current Processor/Chipset” in red is the Phenom II X4 940 Black Edition, overclocked to 3.8GHz.
Cache and Memory:
Self explanatory really. Things sure have come quite a way since the FX-60 was the chip to lust after!
Memory Bandwidth:
The Phenom II pulls out a 17% lead over the Phenom X4, and a 27% lead over the 8750 X3. When overclocked, this balloons out to 27% and 37%, respectively. Not surprising, considering the integrated memory controller is largely dependent on clock speed.
Processor Arithmetic:
With a 15% clock speed advantage, the Phenom II generates a 21% advantage over the Phenom 9950, and an impressive 71% walloping of the Phenom 8750, in the Dhrystone Arithmetic Logic Unit test. At 3.8GHz this transforms to nearly 50% and 111%, respectively.
The Whetstone Floating Point Unit test gives the Phenom II the points by 19% over the 9950 and 73% compared to the 8750. Overclocking changes the picture to again just shy of 50% and 119%, respectively.
Processor Cryptography:
When compared to the Phenom 9950, cryptographic bandwidth favours the Phenom II by 23%, yet its hashing bandwidth lead of 17% is only marginally faster than its clock speed advantage of 15%. Crank up the Phenom II to 3.8GHz and we see a similar story, a 49% cryptographic bandwidth lead from the 46% clock speed advantage, yet hashing bandwidth yields Phenom II the advantage by 55%. No doubt there’s a valid technical explanation, but this writer isn’t educated enough to elaborate!
Processor Multimedia:
Floating point sees the Phenom II edge out the 9950 by 18% at stock speed and 49% when overclocked. The integer tests seem more closely related to clock speed, as they come in at 15% and 48%. The Phenom 8750 lags by 71% in the floating point tests and 66% in the integer tests. Having one less core and giving up 25% in clock speed really hurts it in this test.
|
|
Advertisement:
All original content copyright James Rolfe.
All rights reserved. No reproduction allowed without written permission.
Interested in advertising on OCAU? Contact us for info.
|
|