[Taristin]

All I will say is

AMD > Intel

[Flipside]

I use one of each, a 1.7Ghz (1.4 Overclocked) for games, and a 3Ghz Intel system for my Lightwave stuff. The Intel performs awesomely on Lightwave, but I seem to have a lot more problems with keeping it cool (though, in it's defence, it is running at over twice the speed of the AMD).
The AMD is a good, solid little workhorse, which just keeps on ticking, even when it had to run 24/7 as the Internet server.
In total, I'd say that I've had less problems with the AMD system, however, a lot of the problems I've had with the Intel system couldn't really be blamed on the processor itself, so I really wouldn't like to say which is better, as such.

As far as the lawsuit is concerned, I think this really boils down more to busniess practice than the performance of the chips themselves

[mikhael]

True.

For what its worth, my AMD 2600+ is probably the most amazing chip I've ever used. Not because it performs so well in an application or whatever. Its done something far more amazing, in my opinion.

It routinely handles temps that are ridiculous: it ran at 101C for over a month before the fan bearing froze up, causing the system to shut down. One of my buddies (an engineer at AMD in Austin, Texas) said that its probably should have blown out spectacularly ("that chip should be glowing red slag", in his words).

[redmenace]

In regaurds to HT and its performance, And take note I mentioned one specific situation and definatly not in general.
http://www.tomshardware.com/cpu/20050603/index.html

[redmenace]

More news
http://www.theinquirer.net/?article=24350

[aldo_14]

My understanding RE: HT is that it requires software optimisation.  Certainly it did for single-core; IIRC programmes which weren't optimised to load balance on HT chips would actually run slower than on a non-HT chip.

[redmenace]

More news
http://www.tomshardware.com/business/20050712/index.html
http://www.tomshardware.com/hardnews/20050712_171125.html

[mikhael]

Originally posted by aldo_14
My understanding RE: HT is that it requires software optimisation.  Certainly it did for single-core; IIRC programmes which weren't optimised to load balance on HT chips would actually run slower than on a non-HT chip.

No.

Software that is not SMP capable will run on one core of a HT CPU. No loss. Programs like Lightwave and Max, etc, which are SMP aware, will split threads to as many cores (be they on the same die or seperate dies) as they can find.

[IceFire]

I have only to add that I hope AMD stays in the chip making business.  My Athlon XP 2700+ is the best chip I ever bought.  Its damn fast...damn fast.  Makes me want to say it twice.

Even when its two and a half years old now, its one of the best systems.  I'm also partially biased because I built this machine myself.  But I'm just so very impressed.

In a year or two time, I think another AMD offering will be seriously considered.

[aldo_14]

Originally posted by mikhael


No.

Software that is not SMP capable will run on one core of a HT CPU. No loss. Programs like Lightwave and Max, etc, which are SMP aware, will split threads to as many cores (be they on the same die or seperate dies) as they can find.

Actually, the hyperthreading setup is across logical, not physical processing 'cores'. Resources are shared across both logical processors, so you need to balance - optimize - executional threads to prevent deadlock or conflict, same as for software threads.  Offhand, the dynamic scheduler used on the chip assigns tasks to aribtrary logical PUs; it's not a case of sending to a single logical PU, as the scheduler has no knowledge of which instruction belongs to which excecution thread (expressly that way; it's purpose it load balancing instructions across the logical PUs).

Within the Xeon specifically, the cache is - for example - shared between the logical threads.  This is a specific bad point; there's no intelligent cache control to prevent one (excution) thread monopolizing it.  Which of course can lead to thrashing and/or deadlock between threads.  Benchmarks on HT processors shown performance degredation with non-optimized programs, likely as a result of aforementioned cache thrashing during memory intensive sections.

 Hence, you need to optimise for HT processors to overcome the deficiencies within the architecture and take advantage of potential performance improvements.  (same as any architecture; difference is that Ht was operating within the general x86 architecture shared with multiple other chips)

[redmenace]

http://www.tomshardware.com/hardnews/20050901_200055.html

[WMCoolmon]

ATM I don't feel a need to go pointing fingers at people for AMD's market position. I'm more than happy with my AMD64 processor, hyperthreading is pretty much what drove me away from Intel, because under Windows 2000 it seemed to make the CPU appear as two separate ones with half the processor power.

[StratComm]

Originally posted by WMCoolmon
ATM I don't feel a need to go pointing fingers at people for AMD's market position. I'm more than happy with my AMD64 processor, hyperthreading is pretty much what drove me away from Intel, because under Windows 2000 it seemed to make the CPU appear as two separate ones with half the processor power.

Which is precisely why I have it turned off.  A largely useless feature, I'll admit, but harmless when disabled.

[karajorma]

So wait a sec AMD claim that Intel are artificially inflating CPU prices and instead of selling their processors at a reduced, more realistic price and making a killing that way they instead simply sell their own processors at the same artificially inflated price and then ***** about it in an anti-trust case?



Someone really needs to explain simple economics to these guys.