Can you say BS???

Before u say anything else....read this...
http://babelfish.altavista.com/babelfish/urltrurl?lp=de_en&a p;url=http%3A%2F%2Fwww.computerbase.de%2Farticle.php%3Fid 3D229%26page%3D3%26sid%3De895f31da434f784547ddc4f28cb022c

Heh... interesting.

I think www.anandtech.com did the best write up on this. HT only shows off it's power when running two completely different types of programs.

But in single application mode, the performance gain is nil, and HT can cause a performance loss simply because it will try to process 2 threads instead of one.

Let's say we have a few threads, A, B, C, and D. Thread A and B start at the same time through the P4-HT. Thread A is processed normally, but it requires the same execution slots as Thread B. Despite going in at the same time, Thread B is held over until the next set of cycles. Depending on the state of C, it may enter the processor, or may bot. If it doesn't, the processor holds it over until it's normal third turn.

If it does, three things could happen, the first being that they are completed as normal and B and C exit the processor at the same time. The second is that they once again require the same set of execution slots. They can either error, and be entirely flushed out, and started again. Or, C can be held over for the next set of cycles. It is also possible that depending on the steps inside of the processor for C to be completed and thread B held over until thread D starts.

And the same with thread D, it can be processed, error, or switch places.



This complexity of choices inside HT enabled P4's is what can hurt their performance.



Despite what Intel claims, the average home user won't see that much benifit from using HT enabled computers (the average user being one that probably subscribes to Video Professor...)



Now to deal with AMD's part.

I kinda of fault the benchmarks on the benchmarks themselves. One of the things that really grates on my nerves is the boastments of what score was achieved with whichever benchmark. Most of these "benchmarks" were taken after a full defrag and/or scandisk of the computer hard-drive, and most are taken right after a fresh boot for the OS.

Now scuse me, but an average gamer isn't going to be "cleaning" up their system before they launch their fav game. Most aren't going to be hitting the reset button before they start. Most are going to be simply popping the disc in and/or just starting the game from the desktop.


How well does a processor and system fare after having been on for a week solid running all sorts of programs when it runs a game?

That's the kind of benchmark I'm interested in. Not the "perfect system state" benchmarks. It has been in myexperience having built over 90 different AMD and Intel based systems then working on them later, that even at lower speeds, AMD chips offer a superior "working" performance. Can I prove it? No. Never bothered taking measurements. But what I know is this. After 2 weeks of up-time, my 850 T-bird with 256megs RAM boots UT2k3 almost 3 times faster than a 1.5ghz PIII with 512megs of ram. Both with Sonique (a known RAM hog) running in the background.



Now, as to where AMD is going... I think they will follow the example of IBM and it's Gekko chip. It would not surprise me to see AMD start to offer in late 2004 / early 2005 a TRUE HYPERTHREADING version of the Opteron that runs two complete 32bit instructions at a time, or one full 64bit instruction.

That's the direction I think they will take, but as I'm not a programmer, I have no idea if such an approach would be practicle for General Purpose X86 / X86-64 solutions.

HT = Poor man's Dual CPU system...

HT doesnt even come close to duals cpu systems when it come to performance. Plus if u do a little price shopping around, u can build equaly performing dual Athlon or even smocking P3 system for the price of one of those Intel HT cpus.