Sign in with
Sign up | Sign in
Your question

Hammer vs Prescott?

Last response: in CPUs
Share
June 4, 2002 6:50:46 AM

I'm just wondering what AMD is planning to do to the Hammer when Prescott comes out. I've read in places that Prescott will have 800mhz FSB, 0.09 micron, 1-3 megs of cache and hyper-threading and possibly a whole lot more. If this is true, this will be one heck of a CPU. What future revisions to the Clawhammer will AMD be introducing in 2003 aside from shifting to 0.09 micron. Perhaps a Hammer with two cpu's in one chip. I know this may be looking a little too far ahead considering Hammer hasn't been released yet, but I can't help but look to the future.

More about : hammer prescott

June 4, 2002 6:52:07 AM

The hammer 2 will have ion blasters deflector shields and a flux capacitor!!!

Point?

Wait till its out or near out before wondering how amd will counter the deadly "threat".

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
June 4, 2002 11:31:25 AM

"I've read in places that Prescott will have 800mhz FSB, 0.09 micron, 1-3 megs of cache and hyper-threading and possibly a whole lot more"

If i remember correctly, there were many rumors before the first athlon came out saying it had 512kb to 1mb cache and many other things which it did not have. The moral of this story is dont believe anything until its official.

I need a 1.5 Ghz Athlon + 512mb ddr ram to write emails......honestly
Related resources
Can't find your answer ? Ask !
June 4, 2002 2:13:18 PM

Quote:
The moral of this story is dont believe anything until its official.

<i>Especially</i> when it comes to Intel processors. Since the dawn of the P4, Intel has had this nasty habbit of not actually releasing a product as good as the hype.

Williamette had a stripped out FPU, a downgrade in FSB, and a ho-hum cache.

Northwood fixed the cache, but <i>didn't</i> fix the FPU. Further, the FSB in the Northwood was only recently fixed.

Prescott? Well, we're <i>all</i> waiting for Intel to <i>finally</i> address the P4's FPU with sincerity. Jumping the FSB from 133/533 to 200/800 (skipping both 150/600 and 166/666 completely) seems highly unlikely. (Personally, I expect that if anything, Prescott will launch at the same FSB as NW and <i>eventually</i> ramp to a 166/666 FSB <i>after</i> its release, much like the NW did vs Willamette.) The 1MB of cache ... <i>maybe</i>. More than likely, we'll see Prescott itself have 512KB and Prescott Xeons have 1 - 3 MB. Hyperthreading? Ha! Until MS can actually get their -explitive never entered- straight, Intel would be fools to make <i>all</i> of their Prescott CPUs Hyperthreaded by default. So if Intel <i>did</i> release Hyperthreading in Prescott for mainstream use, it would probably be by an <i>optional</i> setting in BIOS and/or jumper/DIP on the mobo at most. At most, Xeons would probably be the only Prescotts to actually have Hyperthreading enabled by default, and even then the world will still whine and moan until MS actually get's a Hyperthreaded CPU's detection straight.
And then there is the Yamhill extensions that <i>may</i> be in Prescott. Personally, I expect those extensions to be in the core. However, if people at home can actually enable those extensions (instead of Intel releasing a 'new' chip with them enabled) is completely unlikely.

So yes, I expect that at <i>least</i> part of the Prescott's core to be 'wasted' from most SOHO user's standpoint. Further, I expect Intel to drastically fail to meet the expected Prescott hype. At most, Prescott for SOHO use will barely exceed the performance of a ClawHammer and Prescott Xeons will barely outperform Opterons.

Intel just isn't interested in pushing the performance of their cores to their maximum potential. All they want is to play tag with AMD (to avoid being classified as a monopoly) while retaining their part of the market share. They don't seem to give a flying-fig about actually fixing the flaws in their cores to produce an execution monster. So in all likelyhood (at least by my calculations) Prescott will just be a continuation of this theme of mediocrity.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.
June 4, 2002 2:42:48 PM

What does soho mean? I've seen the acronym thrown around a bit but don't know what it is...

MStakem
June 4, 2002 2:49:30 PM

i think it means small office/home office, may be wrong though.

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
June 4, 2002 4:07:11 PM

Quote:
i think it means small office/home office, may be wrong though.

Yep. In other words, SOHO is a quick way of saying "those of us who don't have conveniently huge budgets to purchase in bulk at discounted prices". Heh heh. Basically, the typical PC user.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.
June 4, 2002 4:17:54 PM

Oh yeah? Well not only can you cook eggs on Itanium 2, but feel free to throw on a few steaks. Hell, throw on the whole cow.

<font color=blue>Hi mom!</font color=blue>
June 4, 2002 11:22:39 PM

Haven't I already talked about Prescott in detail on some other topic?

Actually, I read somewhere that the current CONFIRMED specs were:

4Ghz
800 mhz FSB

I think I read it at ace's or anand, I forget. Anyways, it said that these were the preliminary specs released by Intel. Plus it said there would be a bunch of micro enhancements. I'm pretty sure Intel won't lie about fsb/clock frequency. As for big cache, yamhill, hyperthreading etc, these are all just rumours (for now).
June 4, 2002 11:40:08 PM

And my nonexistant fish lives, breathes, eats, drinks, and can do 600 Mflops.

<font color=black>Need Money!! Accepting Donations to help better my future. Thanks!</font color=black>
June 5, 2002 2:29:14 AM

Ya know Pheonix, this is one of the few posts I entirely agree with you!

However not to make Intel look bad, but it is true that we should not expect so much off the Prescott, even if we needed it to be. What I do feel though, is that they know of AMD's CH, and if it proves to be a powerful machine, they HAVE to tweak more.

Now for Dark_Archonis, Anand said in their interview will Intel's 'cronies', that Prescott will begin with 533 possibly, and 667MHZ, due to the possible future using DDR333 Dual Channel. That is if the RDRAM story IS true. They will go for 800MHZ with DDR400 DC then.

Personally, Prescott must have an added FPU, or else I will never see the P4 in its entirety, and I may never give 100% respect without that. I simply cannot explain why I think that way, but to me, without this needed component, the P4 still drags, hiding that at hyper clock speeds. I wonder about the Prescott pricing though, ayayay.

As for Yamhill, I somehow doubt it coming on Prescott. I thought Yamhill was supposed to be a different core chip in itself. But I definitly would be surprised if they omit using x86-64. They know the consumers will use it, as AMD pushes it, and graphic card companies already apply for producing cards for it. It may be early to use it, but Intel needs to follow AMD on this one, or else it may not bode well.

--
Meow
June 5, 2002 3:36:17 AM

"Personally, I expect that if anything, Prescott will launch at the same FSB as NW and eventually ramp to a 166/666 FSB after its release, much like the NW did vs Willamette."

NW comes in the "A" 400FSB, or "B" 533FSB.

"And then there is the Yamhill extensions that may be in Prescott. Personally, I expect those extensions to be in the core. However, if people at home can actually enable those extensions (instead of Intel releasing a 'new' chip with them enabled) is completely unlikely."

Intel will make that decision at the time near release. If the Hammer is successful, and people want x86-64, you will see the yamhill unlocked. That's from the rumours on the net.....Why would they hold it back? That doesnt make any sense.

"All they want is to play tag with AMD (to avoid being classified as a monopoly) while retaining their part of the market share."

I agree w/ that.....I think Intel is guilty of dangling the carrot, and are capable of pulling away. While AMD is really givin us all they got.....

I sold my sig for $50.
June 5, 2002 5:14:23 AM

I agree that if Intel wanted to they could crush AMD. They could go into a price war on a scale much larger than we are seeing now. But it would mean taking massive profit loss (or actual losses) for a couple of years. Something im not sure they are willing to do cuz the stockholders would get pissed.
As far as Hammer vs Prescott. AMD expects to "keep up". But until its released, noone can say for sure.

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
June 5, 2002 2:40:28 PM

Quote:
Haven't I already talked about Prescott in detail on some other topic?

Beats me. I sure as heck don't have the time to read <i>every</i> post in <i>every</i> topic to either confirm or deny that.

Quote:
Actually, I read somewhere that the current CONFIRMED specs were:

4Ghz
800 mhz FSB

I think I read it at ace's or anand, I forget. Anyways, it said that these were the preliminary specs released by Intel.

I trust that about as far as I can throw it. Until a product is actually shipping, specs are about as trustworthy as a lawyer.

Quote:
Plus it said there would be a bunch of micro enhancements.

There damn well better be! If Intel doesn't make a few (such as adding in that missing FSB power and such as actually designing the chip to feed <i>all</i> of it's processing units in a single clock cycle) then the P4 core will continue to be a mockery of the original specs.

Quote:
I'm pretty sure Intel won't lie about fsb/clock frequency.

I'm pretty sure Intel isn't going to release a P4 with an <b>800</b>MHz FSB unless they have a memory solution available that can actually meet the bandwidth demands that such a chip would have.

Quote:
As for big cache, yamhill, hyperthreading etc, these are all just rumours (for now).

Until the chips hit the shelves at the stores, they're <i>all</i> just rumors.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.
June 5, 2002 2:57:14 PM

what are you talking about. you havent clearly understood how an USS Enterprise works.


<i>if <b>you know</b> <font color=white>you don't know<font color=black>, the way could be more easy ...
June 5, 2002 2:59:18 PM

Quote:
Ya know Pheonix, this is one of the few posts I entirely agree with you!

It can happen. Heh heh. As I've said in the past, I'm not pro-Intel any more than I'm pro-AMD. Just with so many AMD fans running around, I ended up having to debate for Intel's side just to make it fair. In reality, I'm just not really impressed by <i>either</i> company yet. They're both <i>good</i>, but neither has impressed me as being <b>great</b>. So from time to time, I'll knock Intel too. ;) 

Quote:
However not to make Intel look bad, but it is true that we should not expect so much off the Prescott, even if we needed it to be. What I do feel though, is that they know of AMD's CH, and if it proves to be a powerful machine, they HAVE to tweak more.

I'm sure Prescott will be tweaked nicely. I'm not sure <i>how</i> nicely, but the Northwood still left a <b>lot</b> of room for improvement with very little tweaking in the architecture. So I have little doubt that Intel will fix enough of the quirks to make Prescott perform slightly better than anything AMD can throw at them.

Quote:
Now for Dark_Archonis, Anand said in their interview will Intel's 'cronies', that Prescott will begin with 533 possibly, and 667MHZ, due to the possible future using DDR333 Dual Channel. That is if the RDRAM story IS true. They will go for 800MHZ with DDR400 DC then.

Exactly. Start with 533 and then later offer something better. This way current memory technologies are still supported and Intel can move on up once new memory technologies come out that will support the increase in bandwidth needed for the next FSB ramping in the Prescotts. After all, what is the point of releasing a high FSB chip if you can't feed it data any faster yet? It <i>might</i> slightly improve performance in asynchronous-clocked implementations (such as DDR) because then the CPU will sometimes not have to wait as long for data to come in. It would still be a pretty sorry performance improvement though until the memory itself improves.

Quote:
Personally, Prescott must have an added FPU, or else I will never see the P4 in its entirety, and I may never give 100% respect without that. I simply cannot explain why I think that way, but to me, without this needed component, the P4 still drags, hiding that at hyper clock speeds. I wonder about the Prescott pricing though, ayayay.

An added FPU at the least. The respect thing is easy to understand. The original specs called for it. Without it, pure FP performance sucks wet donkey. I wouldn't say that the P4 is 'hiding that at hyper clock speeds'. If anything, the P4 is hiding it with a necessity for SSE2 optimizations. The hyper clock speeds are just Intel's way of improving performance. Whether you increase the number of Instructions Per Clock performed or you increase the Clock rate, both increase overall performance. AMD chose to make their chips with a better IPC. Intel chose to make their chips with a higher Clock rate. It doesn't make either method any better or worse if they both work.

Quote:
As for Yamhill, I somehow doubt it coming on Prescott. I thought Yamhill was supposed to be a different core chip in itself. But I definitly would be surprised if they omit using x86-64. They know the consumers will use it, as AMD pushes it, and graphic card companies already apply for producing cards for it. It may be early to use it, but Intel needs to follow AMD on this one, or else it may not bode well.

Oh, I bet Yamhill extensions <i>will</i> be in the Prescott core. I just doubt that they'll actually be enabled at initial Prescott release. Later, just like a FSB increase, Intel might release an x86-64 flavor of the Prescott with Yamhill enabled.

I think when Intel finally releases x86-64 CPUs depends more upon a combination of how well Itanium 2 sells and how well Opterons move into the 64-bit server market than on how much SOHO users want it or graphics card companies want it.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.
June 5, 2002 3:09:23 PM

Quote:
NW comes in the "A" 400FSB, or "B" 533FSB.

Exactly. And Prescott will probably start with a 533MHz FSB to pick up exactly at the point that Northwood is at. It <i>might</i> even fit into the same sockets. (What a concept <i>that</i> would be.) So you could just plug your new Prescott into existing mobos that support NWB.

If Intel was smart, that would be what they would aim for anyway. However, I don't exactly trust Intel to do the smartest thing.

Quote:
Intel will make that decision at the time near release. If the Hammer is successful, and people want x86-64, you will see the yamhill unlocked. That's from the rumours on the net.....Why would they hold it back? That doesnt make any sense.

Because Intel will no doubt view a Yamhill-enabled Prescott as a threat to sales of McKinley and future Itanium-based products. And because Intel is hardly scared into action by rumors. What will enable or disable Yamhill will probably be how badly Opteron crunches Intel's share of the 64-bit server market. If their precious Itanium sales drop too low, they'll do whatever it takes to stay in the market. Short of that though, I doubt we'll see Yamhill for years.

Quote:
I agree w/ that.....I think Intel is guilty of dangling the carrot, and are capable of pulling away. While AMD is really givin us all they got.....

Yeah. The evidence is all there. It's pretty sad, really. But then, if AMD can increase production, then they can start advertising more and increase sales. Once they start making a real profit and take enough market share, maybe AMD will finally no longer be struggling to keep up with what Intel just has to yawn to achieve. Then the <i>real</i> performance battles will begin.

However, you have to admit that the conspiracy theory that AMD is actually being privately run by Intel just to keep the monopoly noose from falling around Intel's neck does have some plausability. In which case, AMD might never really push Intel's R&D.


Tech support said take a screen shot.
Putting it down with my .22 was the humane thing to do.
June 5, 2002 5:42:05 PM

Quote:
<i>Eden says:</i>
I thought Yamhill was supposed to be a different core chip in itself.

I agree. It's of course not impossible for it not to be, but I've always thought of Yamhill as a completely separate core.

Well, actually I've always thought of Yamhill as a street in Portland, but that's beside the point.

Quote:
<i>Eden says:</i>
They know the consumers will use it

Nobody knows that. It could die a lonley old man with no kids. I doubt it, but who knows.

Quote:
<i>Eden says:</i>
but Intel needs to follow AMD on this one, or else it may not bode well.

Not necessarily. If they can get future IA64 chips to run x86/IA32 code well, they don't really need to worry about x86-64. They might go that direction anyway, but it wouldn't be an absolut necessity in that case.

Quote:
<i>LED says:</i>
If the Hammer is successful, and people want x86-64, you will see the yamhill unlocked.

Absolutely. And I think that's part of what Eden was saying, too. But Intel knows to listen to consumers. Why else would we have the i845?

Quote:
<i>texas_techie says:</i>
They could go into a price war on a scale much larger than we are seeing now.

Sort of like what lots of people accused Micron of doing. Drive prices into the ground and then buy out their smaller competitors.

Quote:
<i>texas_techie says:</i>
As far as Hammer vs Prescott.

And don't forget it's also Hammer vs. Itanium, which should be interesting. I'd like to see an 8-way Opteron against an 8-way Itanium 2 in 64-bit. Itanium is heavily FPU, and that's always been Athlon's strongest point. However, Hammer will most likely have better memory usage, which'll help in ALU operations. Should be interesting.

Quote:
<i>texas_techie says:</i>
AMD expects to "keep up". But until its released, noone can say for sure.

But do you expect either company to say "Sorry, we have nothing that can match our competitors"? I suspect both companies know more about the other than they're willing to admit, and Clawhammer and Prescott will be a more interesting battle than we've seen in a while.

Quote:
<i>slvr_phoenix says:</i>
It might even fit into the same sockets.


And be like an XP on my KT7a-RAID. It fits, but it don't work :tongue:

<font color=blue>Hi mom!</font color=blue>
June 6, 2002 11:23:50 PM

These specs were up at Anand about 2 months ago, and anyways, These were supposed specs, seems that they weren't exactly true. I'm still fairly doubtful that they'll release it at 533 FSB. If the clawhammer performs as well as AMD is hyping, then Intel will need more on the P4 then to just raise it to 4Ghz, and shrink it. I'm also doubtful that they'll add any new major core refinements like an extra FPU. It would be easier for them to add cache, throw in some micro enhancements, and probably add Yamhill and HT to the side. Adding an FPU would require them to rework the core, and I think they're going to save their "big guns" for the P5 (IMHO).

Now, you're probably thinking I'm pro Intel all the way and that I hate AMD. That's not true. In reality, I prefer Intel <i>slightly</i> over AMD. Now, there are many reasons why I like Intel and their hardware (CPU's mostly) but I won't bother getting into that now.
June 7, 2002 12:00:32 AM

Well well, it looks like a big debate is on. As a side note, what is the point of switching the sockets? Is it only so that people have to buy new mobos or does it increase performance or something?

My firewall tastes like burning. :eek: 
June 7, 2002 1:11:53 AM

this thread is most amusing... remind me of the "imagine your dream computer" thread were people had things like
100Thz processor with 4gig cache, 10Thz quadpumped fsb with quad channel QDR ram and a 50 gazillion rpm HDD LOL


<font color=blue>Pants Down! Turn Around! Bend Over! You're about to Experience Telstra broadband! :lol: 
June 7, 2002 1:25:41 AM

They can't just run 800MHZ FSB without the RAM standard, if RDRAM can't get there.
First RDRAM is far from getting to PC1600 yet, to match 800MHZ FSB IIRC, to be synchronous. ATM, DDR400 Dual Channel is the only way to achieve these high synchronous FSBs, and thus explains why Intel cannot just move that fast there. A smooth transition is the best way.

--
Meow
June 7, 2002 1:40:53 AM

Sockets are switched mainly to change the voltage on processors. I believe the 423 --> 478 switch was because of electrical regulation problems, though I don't remember the specifics offhand. I suppose another reason would be to make sure people aren't putting a 2.5GHz P4 on a board with EDO RAM and then saying the P4 sucks (just an example, of course).

Of course, the socket is not the only thing that determines board compatability, please reference the KT7a that will not accept Athlon XPs or above.
(Yes, I'm still pissed about that.)

<font color=blue>Hi mom!</font color=blue>
July 28, 2002 12:35:39 PM

Intel's Prescott is to have a 667MHz bus, Intel will see wheather to turn on Yamhill. Why would intel NOT add an Extra FPU, Why would they have to rework the core? Maybe it is already part of the design. Also if Intel is making double pumped 32 bit ALU's (As opposed to today's 16 bit) why wouldn't Intel just make the FPU double pumped and add two of them?
July 28, 2002 1:36:40 PM

You just had to bring this back huh? Flammage to commence again I think. :wink:

:lol:  Finally, I get a capitalized title!! WOOT!! :lol: 
July 28, 2002 7:47:43 PM

I don't think you realize the tremendous amount of die space more FPU's would take let alone the work to accomplish the trace paths on the CPU as well as expand the scheduling buffer's ports.
Besides, more FPU's probably won't help. What needs to be done is an improvement over FP instructions like FXCH.
Either that, or Intel needs to release a scalar FP extension. SSE/SSE2 is great and all but there are some things that just can't use SIMD.
July 29, 2002 1:28:47 AM

Off die though, otherwise, in .25um, it would have been a HUGE core.

"When there's a will, there's a way."
July 29, 2002 2:22:30 AM

almost official
1 mb L2 maybe more for nocona.
166 FSB for springdale chipset DC DDR 333 after maybe a move to 800FSB for DDR 2 at q1 2004.The same happen for xeon with delay and the same for itanium with big delay

Maybe include

32 bit alu due they have 1 running in lab for 6 month.
larger L1


false rumor or does i dont believe
twin pump FPU
twin pump L1
X87-64 bit


There also rumor on IA-64/X87-32 Core name tejas if true not before 0.065 micron.

The day i meet a goth queen that tell me Intel suck.I turn in a lemming to fill is need in hardware.
!