Testing was performed on an Intel Desktop Board D975XBX using 2GBs of Corsair CMX1024-6400PRO and a Maxtor Maxline III SATA150 250GB hard drive. The graphics card used was a Sapphire Radeon X1900 XTX.
We compared the Intel Extreme Edition 965 to the Intel Extreme Edition 3.73GHz, which is basically a single core version of this chip, based on the Prescott 2MB core. This still supports HyperThreading, so does benefit from multi-threaded applications.
The 3.7 Beta of PovRay was used in order to get SMP support and testing was done using the built in benchmark. This benchmark stresses the CPU entirely using many different ray tracing methods.
CINEBENCH 9.5 was also used, which is a benchmark based around the Cinema 4D application. Only the CPU rendering results were used.
SYSmark 2004 SE was used, which is a length vigorous test that involves applications such as Adobe Photoshop, Microsoft Office and Macromedia Dreamweaver.
On top of this, we ran Call of Duty 2, Battlefield 2 and 3DMark06 â€“ all of which take advantage of two CPUs. We ran this at 1,280 x 1,024 and 2,048 x 1,536 using 0x FSAA and trilinear filtering, and 4x FSAA with 8x AF.
Starting with PovRay, results were as we would expect with close to 100 per cent increase in performance. A similar result is seen in CINEBENCH, with an exact 100 per cent improvement. Notable is how even on the single core processor, hyper threading gave a noticeable increase in performance â€“ 17 per cent! It is also interesting to see that Presler is slightly faster than Prescott in single CPU rendering mode â€“ possibly due to the second core taking same of the overhead for the rest of the system.
SYSmark results were surprising with only a 33 per cent overall increase in performance from having a second core. In office productivity, the difference was negligible.
What this tries to illustrate is that for general use, having two processors doesnâ€™t yield much improvement. On paper, that much is true â€“ but from my own use of SMP systems, this isnâ€™t the case. It gets rid of a lot of the little hangs here and there and just generally improves multi-tasking considerably. Naturally, the applications we run arenâ€™t going to be sucking 100 per cent of the processing power all of the time, hence the benchmark results not showing as much of an improvement. But it is definitely noticeable.
Finally, we come to gaming where we see 10-15 per cent improvement at the most by going dual core. Most of this improvement is seen in the lower resolutions where anyone buying a P4EE just wouldnâ€™t be running it. In the higher resolutions, the GPU is holding back performance and the CPU makes very little performance difference. Naturally, as these are timedemos they arenâ€™t indicative of true gaming performance where AI computation plays a big part. But, I think 10-15 per cent is a fairly true to life indication of performance gain seen inside games.
The Intel Extreme Edition 965 is a fast processor and is a good note for NetBurst to leave on. However, at Â£700, this is an amazing amount of money to pay for a processor. With Conroe just around the corner, it would be a large investment in what will soon become outdated technology. This point alone should be enough to make hold on to your cash.
If you really have to spend your money now, perhaps due to looming end of tax year, and you really want to pay Â£700 for a new processor, then I'd sooner buy an Athlon 64 FX-60 for its superior memory controller and cross-core communication.