Prescott->Tejas...? This is absurd!

Wait a minute there... I was kind of thinking...

Intel moved Tejas up for 2nd half 2004, right? One year or so after Prescott. That's sooner than expected. I just got it: this is too fast! Prescott is designed to fade out at 5.5Ghz or so, and Tejas scales all the way up to 9.5Ghz or so, right? Isn't that absurd?

If Prescott scales to 5.5Ghz from 3.4Ghz launch in 12 or so months, that means it'll go up 200Mhz every month! Intel would have to launch a speed-bumped CPU every two months or so (frequency bumps increase to 400Mhz/bump!)! What is going on there? I don't get it. This means that, by early 2004, we'll have to be well into 4Ghz processors... (hey, I'm not complaining or anything - isn't that one hell of a roadmap?) Plus, Tejas might be 65nm... Strange, my brain can't even cope with that... Maybe AMD is on to something? Maybe Intel has 64-bit support - Yamhill - not on Prescott, but on Tejas? And they are rushing it?...

<font color=red><b>M</b></font color=red>ephistopheles

Will Intel alienate some customers buy introducing faster CPUs too soon?

__________________________________________________
<b><font color=red>Three great virtues of a programmer are: laziness, impatience, and hubris.</font color=red><b>

Just because a chip <i>can</i> scale to a higher speed doesn't mean that it <i>will</i>. Intel may very well just release the newer chip at lower speeds. The speed ramp could stay exactly on schedule in spite of any cores coming out sooner.

"<i>Yeah, if you treat them like equals, it'll only encourage them to think they <b>ARE</b> your equals.</i>" - Thief from <A HREF="http://www.nuklearpower.com/daily.php?date=030603" target="_new">8-Bit Theater</A>

Exactly.

And when you think about it, really, Intel has been doing this for a while now anyway. I mean look at the Northwood core. It had extra room that Intel hadn't planned on utilizing until AMD delayed the A64 so much that they didn't have much choice. So Intel changed their roadmaps and released the Northwood Cs.

So it really wouldn't surprise me at all that Intel is setting themselves up to have a lot of unused headroom available in their cores. It allows them to both be prepared for a quick speed jump should AMD suddenly become competetive again, and it also allows them to delay releasing a newer core with a higher IPC should AMD's performance become so uncompetitive that Intel fears gaining too much market share.

Intel is definately playing the conservative game now. AMD caught them with their pants down once, but I doubt that we'll ever see anything like that again. Intel seems to be prepared for anything lately.

"<i>Yeah, if you treat them like equals, it'll only encourage them to think they <b>ARE</b> your equals.</i>" - Thief from <A HREF="http://www.nuklearpower.com/daily.php?date=030603" target="_new">8-Bit Theater</A>

Maybe so, but that doesn't detract from the possibility of Intel leaving a lot of headroom free and doing speed bumps only as often as needed.

Anyway, I think that the increase of each bump is reasonable. First of all, you use multipliers, and as the FSB increases, each multiplier bump increases too. Also, there would be no point in increasing a 5Ghz Processor by 66Mhz to 5.06Ghz, 5.13Ghz, and so on...

<font color=red><b>M</b></font color=red>ephistopheles

Just like you said, it would be absurd to raise the clock by so little, as the current 200MHZ. Think about it, 5.2, 5.4, 5.6?
Not only this, but the real reason behind it, is the proportional scaling. From 4 to 4.2 you only add a max of 5% of performance with a 5% clock raise (assuming the FSB is perfectly holding it, and no other performance increases will appear until Tejas). Further on, from 5 to 5.2 you add only 4%.
It is only LOGICAL that they add 400-500MHZ per step starting from 5GHZ, and 300-400MHZ by 4GHZ.

So basically it is normal, really. AMD had to raise the per-step clock rise since the 2GHZ, because adding 66MHZ as usual, save for the core upgrade small clock rises, is gonna do jack for competition.

--
<A HREF="http://www.lochel.com/THGC/html/news.html" target="_new"><font color=red><b>Join the Tom's Hardware Guide Community Photo album, send us your pics!</font color=red></b></A>

Yes, and it should not be a surprise that an Athlon, if it ran at 3.5GHZ, can still compete against a 5GHZ P4. It's only proportions here.

Of course apps that like raw clocks will enjoy the P4 tons more, but that's the fault of crap programmers who don't know how to program for IPC parallelism.

--
<A HREF="http://www.lochel.com/THGC/html/news.html" target="_new"><font color=red><b>Join the Tom's Hardware Guide Community Photo album, send us your pics!</font color=red></b></A>

Don't forget Tejas goes to 65nm with 1.2 FSB. Maybe with the early release of Tejas we can see normal pricing of chips like $500 or less on release date instead of $637. since they will be releasing a bunch of new Cpu.

No I totally am cool with that. I do prefer a programmer's input above all on this.

I do however have a small thing against some of them, after experiencing some of the crappiest game programming seen to date, in games like Commanche 4 and Splinter Cell.
I realize graphics programming is not the same, but you get the analogy.

--
<A HREF="http://www.lochel.com/THGC/html/news.html" target="_new"><font color=purple><b>The official Tom's Hardware Guide Forums Photo Album, click here to contribute!</font color=purple></b></A>