Coming back to Munich, Intel provided us three hours on a Conroe based machine, using what it described as a mid-range processor running at 2.66GHz. I wonâ€™t go in to too much detail as this will be covered at a later point, but this was on a 1066MHz bus with 4MB of shared Level 2 cache. One Intel representative mentioned how they had a chip overclocked to 4GHz - considering how much faster Pentium M is clock for clock when compared to Pentium 4 â€“ this is an exciting time indeed.
The machine itself was built by Intel with a fairly standard set of games and benchmarks. I was initially surprised at how noisy the machine was, but upon closer inspection there was indeed a CrossFire setup inside. Out of the frying pan and into the fire.. There was even an AMD FX60 machine overclocked to 2.8GHz for comparison.
I am a generally a complete cynic and naturally suspicious of anything that is put in front of me. So when presented with these two machines, there wasnâ€™t a part of me that didnâ€™t say â€œyeah rightâ€. However, as I was there, I took the opportunity to do a bit of games testing. I recorded timedemos in both Quake 4 and Counter-Strike: Source, using the same levels as I do in SpodeMark, while also attempting to replicate the demos path. In reality, this is not dissimilar to other websites methods of benchmarking by recording frame rates using Fraps and playing the same section of game over and over.
The results I achieved were very high â€“ too high. I copied these timedemos on to a USB stick and ran the same demos on a 3.8GHz Pentium 4 machine with the same memory speed and timings and the difference was a lot greater than Iâ€™d like to believe. Compared to the AMD machine on site, the difference was actually believable. But I would be doing AMD no favours by quoting benchmarks compared to a machine built by Intel. No matter how much I inspected configuration files and BIOS settings, and no matter how much I try and justify it, at the end of the day both machines are built by Intel. That is why I wonâ€™t be posting a single result until I get a machine in labs.
However, based on my experience, I do believe that Intel is on to a winner and the performance crown will no longer belong to AMD. After all, even if Conroe turned out to be the same performance as Yonah, things would be looking good, and yet Conroe is designed to be 20 per cent faster clock for clock.
So what is AMD going to do? Considering an overclocked FX60 was struggling when compared to a mid-range Intel part, things look very bad indeed. The move to AM2 has been graceful and there is a definite improvement in performance, but I donâ€™t think the performance increase is going to be enough to compete. Even with the eventual move to 65nm and therefore increased clock speeds or even the addition of Level 3 cache, things are looking bad for AMD. During my stay in Taiwan this week for Computex, I shall be trying very hard to get some answers on what AMD does plan to do.