SSE2 Hammer

[Corona999]

So with the inclusion of SSE2 into the Hammer line, will this take away any advantage the P4 has over AMD's products. With a level playing field like this it'll be interesting to see how Intel competes with AMD. By the time Hammer is released I think Intel will probably introduce hyperthreading into the desktop P4 line. I heard a while back that including SSE2 into the Barton line is a possibility too. This will probably be in a future revision of Barton, similar to the change from the Thunderbird core to the Palomino core. I'm looking forward to the 3rd and 4th quarters of this year.

 

[FatBurger]

SSE2 doesn't give that big a performance gain in average, everyday use. AMD might also implement SSE2 worse than Intel did. Or they might implement it better. We'll have to see.

<font color=blue>Hi mom!</font color=blue>

 

[Kemche]

Or they might implement it the same way.

"640K ought to be enough for anybody." - Bill Gates.

 

[Kemche]

By the time CH release, Intel will be much further in Preformance lead then AMD. From the looks of it T-bred will not be a good processor and we would have to wait till barton to get some more preformance per clock for XP processor so by the time Barton releases Intel will have a ton of pure Mhz lead and also some preformance lead.

AMD will not be able to match Intel's P4 preformance until CH is released in Q4.

KG

"640K ought to be enough for anybody." - Bill Gates.

 

[eden]

Yes but it certainly helps P4, so that WOULD represent a disadvantage to Intel IF CH has the promised performance, which then added to SSE2, would devalorize Intel's tactic to use SSE2 as shield for the low FPU.

--
Intel had 2.2GHZ Willies left...I bought one, it powered my entire neighborhood!

 

[FatBurger]

Of course, but Corona99 seems to be saying it's going to be the thing that makes the difference between the P4 and Hammer (which I disagree with), and/or that it's an unexpected thing (which I also disagree with).

It'll help, of course. But it's not a new thing, and it's not the sole item that will make the difference in this battle.

<font color=blue>Hi mom!</font color=blue>

 

[mr_gobbledegook]

Especially in 64bit mode since the Hammer will have twice the number of SSE units than the P4. SSE optimised x86-64 code should offer great performance.

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple>

 

[texas_techie]

there is some doubt that Barton will even be released. I guess AMD may want to concentrate on making Hammers. And Barton would only be around 5 months or so until Hammer came out. So there is no point ( at least now)
I gotta admit. I am confused as to where AMDs roadmap is going. No t-bred yet, Barton is a big question mark, And im curious what AMDs next value processor will be.. Whatever it is, it will have a hard time competing with a p4 celeron.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.

 

[zengeos]

By the time CH release, Intel will be much further in Preformance lead then AMD.

Not necessarily. However, it DOES seem that Intel is blowing it's speed ramp early to get some sort of definitive lead over AMD in performance. Where Intel may be doing themselves a disservice is that the Northwood can only go so high, speed-wise, before it peaks (I've read it will probably have problems ramping beyond 3.2 or 3.4ghz)

Now, assuming that Thoroughbred is a poor performer and doesn't ramp well, this will put AMD in a difficult position performance-wise for what...4 months? (Assuming Claw debuts in October) Of course, we can't say for sure since Barton may come out prior to Claw...perhaps in July or August. Barton should give Athlon some legs until Claw if it comes out sooner.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>

 

[FatBurger]

More than one person has gotten P4s to 4GHz. Granted, that's with LN2 and just stable enough for the WCPUID screenshot, but it would suggest that Intel could get to ~3.5GHz at least without needing another die shrink.

<font color=blue>Hi mom!</font color=blue>

 

[eden]

That's where your words say the total meaning of what Zengeos meant. If it's only stable with LN2 also, just to run WCPUID, IF it is only, then that would mean the real safe, threshold to avoid another P3 1.13GHZ fiasco, would be 3.4GHZ, the ClawHammer's friend.

I would like to direct the attention though to those who have been worried that PR3400 would not compete well to a 3.4GHZ, and tell you all to forget the numbers, and again, focus on overall. IF CH ramps easily in PR, therefore reaching PR 3700 in say, 150MHZ more, then I see no problem in it competing overall performance, it'll be a bit like Intel's high speed ramping, but this time it's in PR. Remember again, if the PR is Tbird based, indeed the high IPC of Hammer will contribute to a very sensitive PR towards MHZ ramping.

--
Intel had 2.2GHZ Willies left...I bought one, it powered my entire neighborhood!

 

[texas_techie]

Its hard to say where intel will be speed wise in October. But lets assume its around 3.0 gigs stock. Thats being generous. They are at 2.4 now, October is 5 months away. Intel would need 600mhz in 5 months.
Anyway lets assume they can do it. AMD could still compete with that number. Forget PR ratings, I know the rumor is a debut of 3400. Forget that rumor.
Hammer will debut with a lower PR, but performance wise will be ON PAR with a 3 gig P4. Expect a ramp up of 200 mhz really quick. So by Feb-2003, Hammer will likely surpass P4. But not by much. The key will be the performance of the board, not necessarily the CPU. Intel really raised the bar with the 533 FSB.


Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.

 

[Matisaro]

Whatever it is, it will have a hard time competing with a p4 celeron.

A willamette with 128k cache?


I dont even think that could beat a 1.3ghz duron which is out now, let alone a tbred or whatever amd has as its next budget chip, what makes you think amd will have an issue with the p4 celly?

The Cash Left In My Pocket,The BEST Benchmark

 

[eden]

But in the end, wouldn't that mean Hammer was rather useless, and that in the end we got a 5% boost in performance, as if it was just another meaningless AXP stepping?

--
Intel had 2.2GHZ Willies left...I bought one, it powered my entire neighborhood!

 

[FatBurger]

Intel is at 2.53 now, you're a week behind :tongue:

<font color=blue>Hi mom!</font color=blue>

 

[Matisaro]

A 5% boost in performance over what?

The debut hammer will compete with a 3ghz p4 b, according to texas, it can only go UP from there.

The Cash Left In My Pocket,The BEST Benchmark

 

[bront]

The Athlon XP will likely become the next value CPU, with the Athlon 64 being the high end single, and Opteron being for SMP systems.

This does mean that the T-bread will go from high end to value in less than 9 months, and Barton may never be anything but...

Performance wise, the current AXPs should trounce the C4 (Wilty), as you're taking an old willmette and crippling it.

I don't need an anti-virus program, I just boil my computer once a week.

 

[eden]

Well the way Texas puts it, it seems as overall performance, from the highest-end CPU at that time, which is likely to be the Northwood 3GHZ or whatever highest one, that in the end the boost going to CH, would be something disappointing, or only 5% more. I know it's 25% higher in IPC than ATHLONs, but what about overall performance agains the top-of-the-line. As texas puts it, it seems not worth it much anymore if it only does so little against top P4s.

--
Intel had 2.2GHZ Willies left...I bought one, it powered my entire neighborhood!

 

[Matisaro]

But your not listenting eden my friend, if the hammer is released faster than the fastest p4, it will have plenty of room to scale while the nw will be near its end(3.4ghz or so).

The Cash Left In My Pocket,The BEST Benchmark

 

[eden]

Yeah maybe now I get you. I just hope it ramps so fast too, with its very sensitive PR, to MHZ jumps, so that Prescott isn't a threat. But sooner or later, the PR per MHZ compared to P4s will no longer be strong, and AMD may not sound anymore using the high per-clock reasoning. This isn't technical, just from a POV of the PR's influence!
I hope that the 0.09m Hammer is the gate for AMD, to add even more features! That would compete Prescott easily, almost negating Prescott's advancements, back to a Northwood vs Hammer kind of world.

--
Intel had 2.2GHZ Willies left...I bought one, it powered my entire neighborhood!

 

[Guest]

Amd have been stole sse,sse2 from intel,so that really sucks.

 

[Guest]

I don't know, I see the Tbred in June, starting to push the current Pally core down into the duron level.

Barton released in August more than likely perhaps on the SOI or not. Believe someone has asked at one of the tech tours and got conflicting information. I should call the Austin office and find out for sure.

But with the barton release the Tbreds will also move to the duron level. now they can complete drop the .18 process.

That locks out the current p4 core used for the celeries.

With the Athlon64 release, AMD will want to keep a premium for a bit. to boost up profits. So I imagine for a few months, you will see prices comparable to P4's on the athlon64, and the Opteron will carry a even larger premium and that means Dual Processor support also.

The everyman Dual processor for a while wil be the barton.

So with the release of the athlon 64. The tbreds will be moved down into the LOW levels. Bartons will be in the mid range and keep the athlon XP nomiker and also have a MP version. Probably on SOI, (really depends on the availability of soi substrates)

I imagine that the Tbred core will be moved over to UMC, to all dresden all of the SOI production. As UMC comes up on SOI the Barton core will move over to it and Dresden will concentrate on the opterons and athlon 64's

By the time the mobile comes out (based on .09 technology) I see the barton would have moved to the Duron level.
I don't see the x86 core getting another die shrink. so the transistion to .09 will only occur on the x86-64 family and occure at dresden first. UMC will carry the .13 load until it is time to transistion it to .09 probably mid 2004 to late 2004 at that time the x86 core will probably no longer be produced. at least not my AMD.

El

 

[zengeos]

By the time the mobile comes out (based on .09 technology) I see the barton would have moved to the Duron level.
I don't see the x86 core getting another die shrink. so the transistion to .09 will only occur on the x86-64 family and occure at dresden first. UMC will carry the .13 load until it is time to transistion it to .09 probably mid 2004 to late 2004 at that time the x86 core will probably no longer be produced. at least not my AMD.

Yeah, After Barton there are no planned improvements to the 32 bit Athlon core. Chances are, as you suggest, sometime in late 2003 or early 2004, AMD will do another die shrink to .09 micron and perhaps add SOI to Athlon 32. BUT that's all just a guess since AMD has no indications of it on their roadmaps. Indeed, AMD's roadmap only goes out roughly 18 months. So, it will be interesting to see as Athlon 32 transitions to the entry level; price points around $50-$60 to around $120. Luckily, these are profitable pricepoints once AMD fleshes out their .13 micron process and increases the qualified candidates per wafer.

I'm hoping lower clocked Athlon 64's might even intro at around $150-$200Say.. Athlon 64's at PR 2800 or thereabouts. However, only time will tell (and AMD, of course) But it seems like a reasonable entry point for Hammer.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>

 

[Kelledin]

Amd [has stolen] sse,sse2 from intel, so that really sucks.

Actually, Intel gave AMD rights to use SSE/SSE2, presumably in return for royalties. This is a well-known fact; please don't re-hash a long-dead argument.

<i>If a server crashes in a server farm and no one pings it, does it still cost four figures to fix?

 

[zengeos]

Actually, AMD and Intel have technology sharing agreements. For instance, Intel can use hypertransport as well as X86-64 if they so choose.

What it amounts to is a means to ensure some standardization on the PC hardware/software front.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>