Slashdot Mirror

I still don't see it being viable. So instead of one fast thread you'll have two slow ones that are bumping into each other all the time. It just doesn't make sense. The core2 and AMD64 don't have enough bubbles to warrant it.

Alpha may have considered it, but given that neither the Alpha or Athlon have it suggests they abandoned the idea.

Consider this benchmark. Notice how the P4 has the highest clock cycle counts across the board for all of the tests? That's because there are gaping wide holes in the pipeline where another thread could easily fit.

In the case of the core2 and AMD64, I often see an IPC approaching 2 (and higher) when doing bignum math. That's the type of scheduling you can't do when you are injecting another stream of opcodes.

Tom

Support? Sure => http://libtomcrypt.com/ltc113.html

Oh wait... you saw that already ... hmmm ...

Look sparkles!!! Believe what you want. Just keep in mind the "diversity" of tests they run, like "Content Creation" and "Multimedia" and ...

It's been QUITE a while since the last benchmark included things like thread creation, compile times, crypto operations, etc.... I wonder why that is ... Also keep in mind the scale. Often you'll see huge differences in the bars and it could amounts to a 1% difference in performance (which any stats class would tell you could be lost in the noise).

I'm not saying all benchmarks are shite, just that most aren't very thorough or scientific. They just run [or claim to] a suite of boxed tests and post the results. They don't actually crack out the standard deviation and other stats to tell you what the results actually mean.

And if you're "best box" is an X2 4800+ then shut your gob, that's a decent processor. What are you doing where you're really hurting for more performance anyways?

Tom

Obviously you haven't read anything about Core2. Core2 is, on average, 20% faster at the same clockspeed as AMD's Athlon64.

No it isn't. consider this....

The Core 2 Duo has caught up, and in certain applications beat the AMD64 design, but in terms of IPC it's not really better. Where the C2D shines is power efficiency. My E6600 (2.4GHz) is running happily at 3.42GHz without overvolting or other fooling around. At this speeds I can compile, and run any crypto test I throw at it through my daily chores faster than any AMDX2 on the market. For a hell of a lot less than what the AMDFX series cost.

I imagine the AMD 65nm part is suitable for overclocking and in 2007 we'll see 3+GHz high IPC parts be standard in retail.

Tom

Actually I have patches in Allegro and bug reports in GCC [that have been fixed or are being fixed]. I've just recently filed a lkml notice about the ICH8 in 2.6.18. That's how I contribute to these projects.

On my own, I have http://libtomcrypt.com/ which has three major projects that I support, even commercially [I don't get paid but I have companies asking for help for products that they sell... damn I need money...]. I certainly don't say "fuck you user, fix the fucking problem your damn god fucking self you lazy cunt!" I'm sure I'd be out of users fairly quickly.

Next question?

http://libtomcrypt.com/ltc113.html

crypto work done in 64-bit mode on the Core 2.

Tom

:-) crypto benches.

Seems core2 is closer to Opteron but not quite there.

Tom

yes I do use the extra registers. :-)

I don't get the pushback on this. Anything you compile that is non-trivial most likely uses the extra GPRs. So you know the 84 shared objects you have loaded to make your KDE or Gnome desktop work? Yeah, they can use them. Between me hitting a key and it showing up in mozilla are probably three or four libraries. The more efficient they are the better.

Tom

I prove my worth as a developer [*] by what I do not what I hack together at a conference.

Also I think America [and Canada] got over, at least for us techies, certainly not for marketting purposes [arrg] the whole "gee whiz bang we're working with computers". The "Hackers" era has long since died.

That isn't to say there isn't the culture around though. Codecon, shmoocon, toorcon, defcon, etc are all around and surviving (the latter being one of the biggest).

[*] There is a difference between developer and coder that often gets missed. A coder is analogous to a contract construction worker. They know how to hammer or weld or rivet but don't ask them to design the building. A software developer is analogous to the engineer. Seeing the bigger picture, taking in requirements, designing a solution, testing [proving] procedures, etc.

Tom