Intel Prescott Released

daemonslayer writes "The nondisclosure agreement on Intel's long awaited new Pentium 4, codenamed Prescott, has just been lifted. So can it beat its predecessor, the Northwood? Find out at Anandtech, Tom's Hardware, or any of the other thousand review sites." Or HotHardware, PC Magazine, XBitLabs, or HardOCP. Basically, looks like it's faster, but still not the fastest in all areas. Tide goes in, tide goes out.

Comments

Increased cache latency.

[eddy]

The most interesting characteristic of these new P4's is IMHO:"

"On the other hand, Prescott is looking at some massive increases in latency, the access latency for the Level 1 cache has quadrupled, and the Level 2 cache accesses are approximately 50% slower." -- Lost Circuits

Intel better ramp up that clock and/or have everyone optimizing for SSE3 if they want to dominate the benchmarks.

Suggested mod-limit: 3, Interesting

Re:Increased cache latency.

[Anonymous Coward]

On the other hand, Prescott is looking at some massive increases in latency, the access latency for the Level 1 cache has quadrupled, and the Level 2 cache accesses are approximately 50% slower." -- Lost Circuits

Wot this is though is a reflection of the chip going to go to much higher speeds, I read that it will be up over 4.5ghz eventually. So even with those problems u r going to end up with a better chip over all

Re:Increased cache latency.

[no longer myself]

u r going to end up with a better chip over all

I'm no expert in this area at all, but by increasing the number of steps in the pipeline it may be a preliminary move to bring the Paladium/Longhorn/DRM into the chip level. A better chip is sure to come, but it's going to be a chip with some erie strings attached.

Re:Increased cache latency.

[JamesP]

I'm no expert in this area at all, but by increasing the number of steps in the pipeline it may be a preliminary move to bring the Paladium/Longhorn/DRM into the chip level.

Sorry, but you aimed the duck and shot the chicken.

This has nothing to do with Palladium, rest assured...

Re:Increased cache latency.

[Anarke_Incarnate]

The issues they are having, especially with heat on this new process, are going to make ramping the clock speed difficult.

Re:Increased cache latency.

[Hoser McMoose]

Getting the heat out of the chip is not such a big problem as getting the power TO the chip in the first place. That's why Intel won't be making too many speed grades of the Prescott with the current socket type. Instead they will change to the the LGA775 socket that has (a lot) more power and grounding pins.

Re:Increased cache latency.

[Anarke_Incarnate]

yes, but ask yourself this, when you decrease the surface area that you wish to remove heat from, and increase the total heat output, the heat density is much much higher. Heat causes resistance, which in turn will cause more problems with integrity. It is going to be tricky just to keep that sucker stable at 4GHz unless they make some big changes.

Re:Increased cache latency.

[ameoba]

I'm no expert in this area at all, but by increasing the number of steps in the pipeline it may be a preliminary move to bring the Paladium/Longhorn/DRM into the chip level. A better chip is sure to come, but it's going to be a chip with some erie strings attached

Uh... no.

If you'd actually bothered to read up on CPU architecture & the press releases, Paladium is actually slated to -reduce- the number of pipeline stages by a factor of 3 and Longhorn plans on moving several of the early steps into software. I've seen some prototypes of Palladium systems running pre-release versions of Longhorn and we're looking at only 5-6 steps in the pipeline.

Re:Increased cache latency.

[Junks Jerzey]

Intel better ramp up that clock and/or have everyone optimizing for SSE3 if they want to dominate the benchmarks.

In reality, x86 benchmarks have become all but meaningless. They're all within a short distance of each other, and each chip is faster for some things, slower than others. There hasn't been a real breakway technology CPU in a long time.

Re:Increased cache latency.

[jeanicinq]

The reviews I read all seem to be geared with a point of view that the processor, eteched "intel confidential," would be the only that hits the market. That may be so to get a Prescott to work with the boards that are already on the market. The review articles seem to forgot to mention, but do hint, that the Prescott is designed for a new platform, the 775 socket. That would explain the latency.

The first versions of the Prescott do not look like a flagship model. It looks like it is geared for the developers that want to start to take advantage of the Prescott features now at an affordable rate. The developers get to use the same motherboard, memory, and chipsets as the pentium 4. Obviously, that is set to change with the 775 socket. Not enough support for the 775 socket is on the market. I have seen support for the PCI Express, and that is being quickly outdated. The cost to leap onto development of the applications for the Prescott and newer platform is less expensive with the pentium 4 motherboard compatibility.

With a socket 775, we could expect the pipeline to have a greater bandwidth in parallelism than what the current socket 478 can acheive. That would potentially increase the number of logical processors for hyperthreaded technology. The SSE3 instructions "monitor" and "mwait" hint that greater number of logical processors are possible.

Readable review

[Ed Avis]

Could someone suggest a review site that doesn't split every article across 20 web pages?

Re:Readable review

[after]

Just read the summary, it outlines the new in the most basic form so that even you can understand it.

With several months delay, Intel is now starting to transition from a 130-nm to its 90-nm production process. But the new processor crop, marked with an E after the clock speed number, will only ship up to 3.2 GHz for the time being with the 3.4 GHz version to follow later. The Pentium 4 Extreme Edition, meanwhile, is also available at 3.4 GHz, and takes the top position instead, flanked by the regular 3.4 GHz Pentium 4 version.

They are going 90nm now. First they will ship 3.2 GHz, and 3.4 later. The Extreme is already 3.4

Its not that hard, you know. Dont bitch about it.

Re:Readable review

[Emil Brink]

You can click once more at Anand's to get a Printer-version, which is really nice. It's how I usually read their content. Can't direct-link you (I think), since it's a JavaScript thing. Scroll down, and click on "Print this Article".

Re:Readable review

[yora]

www.anandtech.com. Select "Print this article" and you will have the article in a single page.

Re:Readable review

[ValourX]

Yeah, I'll have a very good review up later this week on The Jem Report, and I'll be testing in FreeBSD and GNU/Linux as well as Windows. I'll have some very interesting (possibly /.-worthy) things to write about if all of my testing goes as planned.

And... most importantly, I will be using retail product, not pre-release samples like all of the reviews posted above. That means you get the real review, not prototype results.

-Jem

Re:Readable review

[Rufus211]

both xbit's and anand's have nice Printable versions that are all on one page.

Re:Readable review

[eigerface]

Don't know whether you were looking for a "Funny" Mod or not, but.....

I only tried the first review, but generally, if you look for the "Format For Print" link (or something similar), all of the pages of the article appear in a one page window.

Cheers.

Re:Readable review

[Anonymous Coward]

Can't wait for your very good review with very interesting things. Meanwhile, a bit full of yourself, aren't you? Like, enough to ward off some potential visitors? Mellow down, man...

(Sheesh -- was I trolled?)

Re:Readable review

[ValourX]

Hmmm.... no, you weren't trolled, but you are trolling.

I have confidence in my reviews. If you have a problem with that, too bad. Maybe that's why you're a nameless coward on /. and not writing your own reviews?

-Jem

Re:Readable review

[LoganEkz]

Sure you can direct link to Anand with no advertisements / bagillion pages for a printable article ;)

Just replace the
/category/showdoc.html?i=article_number with
/printarticle.html?i=article_number
in any AnandTech article.

Thoughts.

[DarkHelmet]

In case you don't want to read the article, here you go from Anand:

If you're looking for nothing more than a purchasing decision let's put it simply: if you're not an overclocker, do not buy any Prescott where there is an equivalently clocked Northwood available. This means that the 2.80E, 3.00E, 3.20E are all off-limits, you will end up with a CPU that is no faster than a Northwood and in most cases slower.

I figured as much before the NDA was lifted. After all, with a 31 stage pipeline, the Prescott was bound to be clock for clock slower than it's previous incarnations.

This only makes me wonder. If a 4ghz Prescott is going to be much like a 3ghz Northwood, is AMD going to adjust its PR Rating to the new cores that Intel has? This will only end up confusing things, as a newly rated A64-3400 will be faster than a "Higher numbered" intel version.

Great... Just what we need. More PR confusion.

Re:Thoughts.

[The One KEA]

My biggest beef with Prescott is that Intel rather foolishly lengthened the pipleine and monkeyed with the core design without making the subsequent changes needed to increase clock speed. AMD had it right all along - efficient IPC and low clock speed.

This situation is shaping up, in my eyes, to be a repeat of the release of the Willamette P4 - an inefficient IPC coupled with a low clock speed nearly killed the P4 before Intel could increase the clock speed. The same thing is happening here - another inefficient IPC design with a clock speed equal to the current Northwoods, with subsequent losses in performance. And like another poster here said, the A64 3400+ still beats the Prescott in a number of benchmarks, or ties evenly with it. Despite Anand's statements about how higher clockspeeds increase the efficiency of the Prescott core, I still think that this processor is an expensive upgrade that doesn't do very much.

If Intel can't get the clock speed up on Prescott, I have a feeling that it's going to tank until the LGA775 packaging is finally brought out, which is going to mean more business for AMD and a lot of eggs on Intel's face.

Re:Thoughts.

[Epistax]

The problem is the number of pipeline stages grows as the product is implemented. By the time it's obvious thtat the problem can't be avoided, it's too late to fix it at the low level it needs to be.

I'd really like to address this question fully but I'm currently working on an Intel processor. It hurts me not to type what I want but I know better. ;-)

Re:Thoughts.

[JollyFinn]

Prescott was faster at some benchamarks, and prescott is cheaper to manufacture, presscot is currently PACKAGING limited in clock speed. [Those damn little pins cannot handle enough current to feed the beast, especially with REDUCED voltage.] The performance, as overall the 3.4Ghz prescott should be slightly better buy than 3.4Ghz northwood as it scales a LOT better in clock speed. Basicly if cannot move enough power to the die you cannot clock it high, so thats the new packaging that will unleash the power of prescott... Also its versio 1.0 ... The L1 cache numbers just look TERRIBLE, worser latency than 64Kb cache would of been, they probably fix that and performance jumps up again, or they really designed it with extreme clock speed, and perhaps a SMALL part just failed to get the high clock speed. The long pipeline isn't that bad because they improved branch predictor. Something tells me they made it run VERY high clock speeds and scale performance well with clock speed. BUT there was a little problem, and they couldn't get the high clock speed from day 1.

PS. I'll personally think getting A64 is better than these just price/performance point of view, intel is market leader and prices its products higher than equally performing underdogs products.

Re:Thoughts.

[ameoba]

You make it sound like Intel is stupid & trying to screw over the consumer with this product. I'm by no means an Intel fanboy, but you're kinda off base.

If you look at the benches, the Prescott cores generally aren't too much slower than Northwoods and the areas where they lag the most are the ones that SSE3 looks like it should alleviate.

The Prescott delivers respectable performance and will end up costing less at the same clock speeds than Northwood. We're not looking at an event like the original P4 launch where the new chip was not only slower but also more expensive & required hardware upgrades to use.

The Prescott is not being marketed as an upgrade to Northwood systems; while I'm sure Intel would love you to replace your 3.2C with a 3.2E, they're not suggesting it be done. They're just introducing it now so they can ramp up production before the Northwood gets phased out.

As far as the 775 socket goes, Athlon64 is also kinda waiting for a new packaging to reach its full capabilities; the A64 & A64FX lines are going to be moving to a unified socket that'll give the A64 access to dual-channel memory.

Re:Thoughts.

[Slack3r78]

If a 4ghz Prescott is going to be much like a 3ghz Northwood, is AMD going to adjust its PR Rating to the new cores that Intel has?

My guess is, no. Despite what may seems intuitive, the AMD PR rating is not scaled against the P4. The baseline measurement for AMD's PR system is a 1GHz Duron. IE: My Athlon 2600+ is supposed to perform 2.6 times faster than a 1GHz Duron would. Hope that clears things up for you some.

Re:Thoughts.

[philthedrill]

Anand mentioned this, but the decision to add pipeline stages, whether it be 1 or 11, wasn't done on a whim. In fact, just to go through with the design, verification, and implementation processes would take over a year (probably closer to two).

Even when you add one pipe stage, you have to worry about creating new hazards/bugs and making sure that the chip is functionally correct. In fact, I'm surprised that Prescott is even considered a P4 core. To me it seems like a radical design change.

My guess is that Intel's goal was greater absolute performance. In the end, they'll get better performance w/ Prescott than Northwood because they'll be able to hit higher frequencies. The performance that Northwood is achieving right now would probably be much more difficult to hit on a Pentium 3 core. And if the extra frequency will help sell the product and impress the naive consumer, then it's extra marketing firepower for them.

The problem with increasing the number of pipe stages is that it becomes more difficult to maintain your IPC... every branch misprediction, cache miss, TLB miss, etc. hurts more, and no amount of out of order execution can keep the pipe filled. We're seeing a trend towards SMT (like the Power5) and chip multiprocessing (Power5 also) just to make better use of our transistor budget.

What I think is really hurting the initial performance is the increased cache latency, at least for the short term. The side effects are probably mitigated by the trace cache. But they waited until 90 nm to release Prescott, and that's probably because of the extra power consumption and area of the extra pipeline registers, but there's also an increase in leakage, so it's a tradeoff. They shouldn't have any problem ramping up their clock frequency. Maybe this wasn't the right way to go, but one thing is for sure - they didn't take the easy route.

Re:Thoughts.

[Junks Jerzey]

AMD had it right all along - efficient IPC and low clock speed.

Of course you don't see any AMD chips that are *clearly* head and shoulders above Intel in benchmarks, though. Maybe a few percent faster here, a few percent slower there. It's all just noise.

Re:Thoughts.

[bhtooefr]

Actually, I've heard that the AMD performance ratings are against the Duron 1.0GHz, meaning that THAT CPU performs approximately like a Northwood.

Re:Thoughts.

[bhtooefr]

In fact, I'm surprised that Prescott is even considered a P4 core. To me it seems like a radical design change.

I do agree that the Prescott should be a Pentium 5, seeing as it's a redesign, and it has a new revision of SSE (Intel DID get away with Katmai being a Pentium III, after all - I think Katmai should have been the P2B, but...)

The performance that Northwood is achieving right now would probably be much more difficult to hit on a Pentium 3 core.

Do you mean a Tualatin core, or the P6 core that is the basis of the Pentium Pro, Pentium II, Pentium III, and Pentium M (and Celeron counterparts)? If it's P6, then you're wrong - a 2.0GHz Banias (no such CPU exists, but I don't know if Dothan enhances the core or just lets it hit higher clock speeds) could murder a 3.2 (maybe even 3.4)GHz Northwood in heat, cost, power consumption, performance, etc. Yes, I'm waxing poetic about the Pentium M, but it's a damn good CPU, and my next rig will either be an A64 desktop or laptop, or a P-M desktop (PowerLeap's developing an adaptor) or laptop.

Re:Thoughts.

[Flywheel]

The PR is not rated against the performance of the Intel processors.

Re:Thoughts.

[philthedrill]

Do you mean a Tualatin core, or the P6 core that is the basis of the Pentium Pro, Pentium II, Pentium III, and Pentium M (and Celeron counterparts)

Well, when I was writing this, I was only thinking of the PIII/Tualatin. You're right about the Pentium M... but it does have some major design changes. For example, up to the PIII, the core uses micro-ops (uops). According to this, the Pentium M uses macro-ops. This says so too, even though they don't specifically call it a macro-op. I tend to think of the Pentium M as somewhat "loose" implementation of the P6 core.

Re:Thoughts.

[Witsu]

That may be the official case, but I think AMD is planning on consumers comparing the PR rating to an equivalent P4. Notice the geek oriented 64FX has a model number that looks nothing like Mhz, but all the cheaper A64s do. PR ratings are obviously designed to trick the average consumer, to whom clock speed is everything into thinking they are getting a faster chip. Prescott will just mean AMD's model number system will continue to work longer, unlike with the AthlonXP, which often performed alot lower than it's PR number suggested

Re:Thoughts.

[brucmack]

That's the thing though. In general, it isn't clock-for-clock slower than Northwood, giving generally equivalent performance at 3.2 GHz.

This is what amazes me about the chip... Intel has done a really good job improving branch prediction and other parts of the core to keep the pipeline filled as much as possible. Being able to increase the pipeline length by over 50% and not have a large decrease in IPC is just amazing.

And when you consider that a 3.2 GHz processor runs up to at least 3.8 on air cooling now, you just know this thing is gonna ramp up to 5 GHz next year.

Re:Thoughts.

[Anonymous Coward]

Can you give any details on the Powerleap part?

This is the first time I hear about a solution like this...I presume it would fit into a ("Pentium4") Socket478?

Re:Thoughts.

[bhtooefr]

I e-mailed PowerLeap, asking if they would consider making an adaptor from Socket 479 (used by the Pentium M) to Socket 478. I recieved a reply stating that they were already working on this upgrade, and to expect something in Q1 2004 (this was in November or December). Once I get a chance, I'll post the e-mail to my site (which reminds me, I need to get that back up).

Re:Thoughts.

[mrm677]

My biggest beef with Prescott is that Intel rather foolishly lengthened the pipleine and monkeyed with the core design without making the subsequent changes needed to increase clock speed. AMD had it right all along - efficient IPC and low clock speed.

You don't read slashdot do you? This post links to some very recent research that claims lengthening the pipeline is actually the right thing to do.

Another armchair architect.

Re:Thoughts.

[Deliveranc3]

As someone who owns a considerable amount of AMD stock let me just say phew. These processors are crap, Intel screwed the pooch again.

The chances of them now bringing a competitive product to market before the 64 bit revolution hits are approximating 0.

Second the AMD performance ratings, as has been said before they are compared to an overclocked thunderbird (Old athlon). So a 2100+ is a little faster than a Thunderbird would be at 2.1ghz.

Of course they are changing all that with the new Athlon 64's (Above 3400+) but I'll keep you updated when I know what they are doing. Until that time just think of a very very fast thunderbird. It's easier if you remember that XP's still run on socket A.

Re:Thoughts.

[Anonymous Coward]

No... just cheaper and earlier to market...

Re:Thoughts.

[Hoser McMoose]

... the areas where they lag the most are the ones that SSE3 looks like it should alleviate.

I don't anticipate that SSE3 will have much of an effect on performance, certainly not like SSE2 does. It's really just filling in a few holes, instructions that probably should have been included in SSE2 but weren't for whatever reason. Some odd special-case scenarios might see a big boost, but for the most part I would throw out a guess of 0-5% max for most programs, with the majority falling closer to the 0% side of things.

The Prescott delivers respectable performance and will end up costing less at the same clock speeds than Northwood.

This, of course, is the kicker. If/when the Prescott is cheaper than the Northwood, then it starts to make sense.

We're not looking at an event like the original P4 launch where the new chip was not only slower but also more expensive & required hardware upgrades to use

I would say that nearly everyone who would want a Prescott IS looking at hardware upgrades. Only the latest and greatest P4 motherboards will support the chip, and if you've already got a 2.8C GHz P4 or faster, why would you want to upgrade now? Even most boards made 6-8 months ago are unlikely to support Prescott chips, particularly those > 3.0GHz which require new voltage regulators.

Will this work?

[after]

Can I overclock it to 5 HGz ?

Re:Will this work?

[PowerBert]

Once you find the right lady I'm sure you can have as many HuGz as you like ;-)

Ohhh Ice..... hhuhhh plheahhd mhoooo tohunfhgg!!

Re:Will this work?

[ameoba]

Generally, when you get a new chip like this, the first ones aren't going to OC well at all; production is still in the early stages & they're lucky to get chips that make the higher speed bins. It's generally when the processor gets more established & good processors are marked down to lower grades than they're good for that you find good OCers.

Re:Will this work?

[Idarubicin]

Can I overclock it to 5 HGz

No. Overclocking leads to read errors from the keyboard buffer.

Slower!!!

[PhrozenF]

Why don't you go check benchmarks before you say faster....Most sites say it is slower than the earlier Pentium 4 because of the increased number of stages in the pipeline. And obviously, it's beaten blue by the AMD A64 3400+ in more than half of the real-world benchmarks.

Sure, the increase in cache helps, but the increase in pipeline stages really kills intensive non-repetitive computing tasks...

and oh...i think I got first post!

Re:Slower!!!

[sbennett]

It's marginally slower now, but the margin is less than when the original P4 came out, and for much the same reasons. But then again, the original P4s were designed to run above 2GHz, so they were slow at 1.4GHz. I'd suggest that Prescott is probably designed to be running in excess of 4GHz, so it is slower now than the Northwoods. That'll change once they start ramping up the clock speeds, and the effects of a longer pipeline become less significant.

Re:Slower!!!

[maraist]

That'll change once they start ramping up the clock speeds, and the effects of a longer pipeline become less significant.

I disagree.. I think that the performance enhancements were due to factors other than the lenghtened pipe and fast clock.. The clock merely compensates (currently badly) for the added [wasteful] buffers, longer latency, and deminishing marginal return on a single enhanced variable (clock-speed).

Intel needs to create a market for it's higher priced CPUs.. So by having a nominal performance chip, they can increase the other variables (cache performance being a big one), and thus charge an enormous premium.

I believe that they could go a long way to enhance the performance of their existing P4 archtecture, but they need more marketing power.. They don't want to waste time/money advertising Pentium 5. Additionaly, the "extreme-edition" moniker on a similarly clocked CPU is going to be a hard-sell. Thus they will make the most money on clock-enhancements.

AMD has the potential to capitalize on this by getting a higher benchmark rating, virtually for free, so I don't really see this as a big win.

The only issue is that it's cheaper to design a CPU with more stages than to optimize a lower-stage-count to get more Instr/sec. So AMD might not be in a position to get a truely faster cpu out any time soon, and relabeling their existing CPU's won't go over very well.

from the amd information minister...

[Anonymous Coward]

Our INTEL says Opteron is faster =)

Re:from the amd information minister...

[OmniVector]

that doesn't mean much to the average user, who is going to be buying the consumer level chip (like an amd64). opterons and their motherboards currently still fetch a premium and aren't in the average consumer's price range. though at least opterons are more consumer-oriented than xeons in price.

Re:from the amd information minister...

[evanbd]

In case you weren't aware, the opteron and the amd64 are the same chip. The only differences are certification and memory bus width, but my understanding is that amd will be selling 128 bit wide athlon64 parts. Some people just like to refer to the 64 bit parts as opterons for clarity.

Re:from the amd information minister...

[ameoba]

not quite :

Athlon64 - the consumer level chip. limited to single-chan. memory

Athlon64 FX - Opteron targeted at single-proc workstations.

Opteron - A64 FX with SMP capabilities.

Re:from the amd information minister...

[radixvir]

opterons and their motherboards currently still fetch a premium and aren't in the average consumer's price range

as opposed to the price of extreme edition cpus?

Re:from the amd information minister...

[Anonymous Coward]

The new 939 pin A64s will have dual channel memory just like the FXs.

Re:from the amd information minister...

[Anonymous Coward]

Or even more accurately:

K8 core, 1 RAM channel, 1 HT link: "Athlon 64"
K8 core, 2 RAM channels, 1 HT link: "Athlon FX-5x"
K8 core, 2 RAM channels, 1 HT link: "Opteron 1xx"
K8 core, 2 RAM channels, 2 HT links: "Opteron 2xx"
K8 core, 2 RAM channels, 3 HT links: "Opteron 8xx"

(Note: I'm not up to date on L2 cache sizes; it's either 512kB or 1MB but seems to vary within the FX series, which is confusing. Also, note how the FX is a "rebranded" Opty 1xx.)

Aww man

[Rosco P. Coltrane]

Basically, looks like it's faster, but still not the fastest in all areas

Hemos, damn you, I was about to read the articles and you spoiled the end ...

Re:Aww man

[MikeCapone]

Actually, he was being an unreliable narrator...

2 Jags Prescott

[reality-bytes]

Its only people from the UK who will know what I'm talking about but every time I hear 'prescott', an overweight, drink-laden scruffy politician with a McDonalds voucher in his pocket springs to mind.

This also conjours up an impression of the Intel Prescotts being in ineffecient, environmentally-unfriendly and handling code in an annoying accent.

Why oh why couldn't Intel have called them something else, like Intel Bloody Powerful chips?

Re:2 Jags Prescott

[Anonymous Coward]

not the best name in the world, I must admit.

Assault story:
http://news.bbc.co.uk/vote2001/hi/english/ newsid_1 335000/1335107.stm

Re:2 Jags Prescott

[Rosco P. Coltrane]

Its only people from the UK who will know what I'm talking about but every time I hear 'prescott', an overweight, drink-laden scruffy politician with a McDonalds voucher in his pocket springs to mind.

Intel markets it as the TedKennedy core in the US.

Re:2 Jags Prescott

[BabyDave]

But on the other hand, it could also suggest that the new processor packs a punch.

Re:2 Jags Prescott

[reality-bytes]

After careful freeze-frame examination of that incident me and my mates came to the conclusion that he "hits like a girl" and "We could take him any time".

Interestingly, we suspect that one of my mates is currently being beaten by his wife to be. ;)

Re:2 Jags Prescott

[pklong]

Yes, but it does include the M4 bus line, designed specially so Prescot isn't held up when the bus gets busy.

Re:2 Jags Prescott

[kfg]

Interesting. As a Yank the first thing that springs to my mind is a nifty little classic British motorsports event.

Not that I've ever noted a personal tendency to represent the average American mind you.

KFG

Re:2 Jags Prescott

[Anonymous Coward]

..and not forgetting punching a member of the public. I wouldn't buy a processor named after something that looks like Jabba the Hut.

Re:2 Jags Prescott

[sepluv]

Well my initial thought when I saw the /. headline was: it may not be environmentally-friendly (but it if you watned that you'd go with Transmeta (Linus's old employer) CPU's), but it will go as fast as two jags.

BTW, the town were the punch-up incident happened (that the BBC article calls a "tourist resort") is just round the corner from me, and is not the kind of place that it is unheard of for punch-ups to happen in. It has been voted worst town in Britain several times in a row. Also, my mates who live there call the town, "the gateway to hell".

Also, did anyone hear the Ogg Vorbis recording of the Linux Expo 2003 discussion (that came with the Linux Format DVD)? -- the quality was crap but I heard them taking the piss out of 2jags.

Re:2 Jags Prescott

[pklong]

For anyone wondering who this Prescot Person is, here is a picture of him when he's not in a Jag.

Re:2 Jags Prescott

[reality-bytes]

I feel sorry for the bike....

Re: 2 Jags Prescott

[jdtanner]

I have visions of error boxes with a Northern twang!

"Eeeeeeh, by eck lad there's trouble at' mill...you've got a chuffin' floating point exception!" ;-)

Re:2 Jags Prescott

[sam0ht]

Yup, I was thinking the same. It's the 'fat, ugly, stupid, brute-force' association that really gives a bad impression of the chip to me, as if it's a dumb design that will only be able to compete by throwing gigahertz at it.

Re:2 Jags Prescott

[lotho brandybuck]

Prescott Bush, anyone? Naming your processor after the Grandpappy patriarch of America's current ruling (gag) dynasty? if there's no conspiricy there I'll munch my all-you-minie-um.

Prescott will be like the P4

[gumbysworld]

Prescott will be like the P4. It will be slow in the begining as they milk every mhz stepping they can but will slow start to shine when they pump up the MHZ.

Its a shame but that is how it goes and went with the P4 it need more speed to be able to show it true worth.

It would be nice if they said screw it and just released it a 4.0

Re:Prescott will be like the P4

[ameoba]

First off, Prescott -is- a P4.

Now, I'll assume you're talking about the initial P4 launch (Willamette). This is nowhere near as embarrasing as that was. The Prescott is only a few percent slower than the Northwood at the same speeds and works in (most) existing motherboards. The Willamette, OTOH, when it was launched, required new motherboards & RAM and was getting beaten horribly at benchmarks by chips with a significantly lower clock speeds.

While upgrading from a Northwood rig to a Prescott rig would be silly unless you're making a significant clock speed jump too, the Northwood looks respectable. If you consider that projected prices on them are somewhat lower than Northwood, it all balances out.

Re:Prescott will be like the P4

[Anti_Climax]

...but will slow[ly] start to shine when they pump up the MHZ...

Actually, they call that glowing...

More Reviews

[RedSynapse]

Tech-Report Prescott Review
accelenation Prescott Review
Ace's Hardware Prescott Review
Gamers Depot Prescott Review
HardTecs4U
Hexus
K-Hardware Prescott Review,
Legit Reviews Prescott Review
LostCircuits
MBReview Prescott Review
VR-Zone
X-bit labs Prescott Review
XtremeSystems Prescott Review
Extreme-tech Prescott Review

Re:More Reviews

[swv3752]

You could at least throw in LinuxHardware.org

Hemos has a bit of the CmdrTaco

[Sixty4Bit]

He typed in the name of the site and the URL and got the name wrong. Genius!

[H]ard|OCP is the correct spelling! But even if you wanted to leave off the goofy characters it would still be HardOcp...

This is very excusable... if Hemos was up late watching the superbowl. Keep up the good work guys. You are still the best geek site around!

Ironically

[shayera]

When viewing the article at Anandtech, I'm pestered with ads for the AMD64 Opteron..

Hardcop?

[marol]

Really...

Will software catch up?

[erick99]

Intel marches steadily on with new chips and planned obsolescence for the old chips. I tend to feel that software is lagging in terms of taking advantage of more powerful and faster processors. I suppose some programs such as PhotoShop can take advantage of faster chips when rendering large and complex files. Still, I think the processors are, by and large, way ahead of software.

Happy Trails,

Erick

Re:Will software catch up?

[Reivec]

I don't like the mentality that just because CPUs are faster that means software should bloat itself up more and more so that you HAVE to have the fastest chips just to run your OS. What crack are you smoking man? The fact that software isn't keeping up is a GOOD thing IMHO. I mean think about it, it means that we can do everything we need to do and have plenty of room to spare. This leads to greater multitasking as you can do many many things at once and not get bogged down. It means you games will run smoother, even if you have programs running in the background.

I personally do NOT want to see software makers TRYING to push CPU limits just for the sake of pushing them. We will just end up with a bunch of useless "features" that no one wants or uses and it will be just like all the cell phones out there :-P.

Normal people will continue to buy middle of the line stuff and will continue to say "gee, this is much faster than my old machine" which is how it should be. If they got new software with that new machine that used up all the cycles then it wouldn't feel much faster and there wouldn't be much point in buying it now would there? And there are plenty of people that will always have the "It can't be too fast" mentality. ie. gamers, multimedia people (images and movies), scientific computing, rendering farms, and I am sure lots more. I personally think things are on track.

Re:Will software catch up?

[crzybkr4]

Actually, I think software should take advantage of this. Perfect example -- my Transmeta-based laptop hardly ever runs at more than 300 Mhz as reported by Longrun. Occasionally it spikes up at full speed at 933 Mhz. Even playing a DVD or an MP3 requires no more than 300-500 Mhz. Heck, even with other programs running in the background!

My point is that unless the software takes advantage of the extra speed, the extra speed is meaningless in most situations. Most applicaitons won't feel any different between a 500 Mhz machine and a 2 Ghz machine... Users don't write their instant messages any faster.

Games should and do take advantage of faster processors.

Handwriting recognition, voice recognition, gesture recognition, biometrics, etc. should all take advantage of the faster procesors. Otherwise, software is in fact lagging...

Re:Will software catch up?

[Junks Jerzey]

Intel marches steadily on with new chips and planned obsolescence for the old chips. I tend to feel that software is lagging in terms of taking advantage of more powerful and faster processors.

Faster processors have enabled more dynamic higher-level languages like Python, Erlang, and Smalltalk to shine. This results in more robust software, software that's also quicker to implement.

(In reality, the proper approach would be to design a CPU to run Python or Erlang or Smalltalk directly. Ericsson prototyped such a chip for Erlang, and when running at 10MHz it was outrunning the Erlang emulator on a 500MHz UltraSparc by a factor of 30. And Erlang is one of the fastest interpreted languages out there. Not to mention the power consumption was something like 1% of the UltraSparc.)

Re:Will software catch up?

[Anonymous Coward]

and Sund did the same with Java and Javachips, but I am not sure if they actually managed to implement any java stuff in silicon or they abandon the idea

Re:Will software catch up?

[killfixx]

Duh...it's the opposite...the lengthened pipeline decreases the dependency for software bloat...because now the CPU does the work...

Re:Will software catch up?

[Anonymous Coward]

because now the CPU does the work

Do you even know what a pipeline is?

Re:Will software catch up?

[fitten]

...and they had CPUs that ran native LISP as well, while interesting, they weren't practical.

Re:Will software catch up?

[TheLink]

If a CPU is too fast for you just add compression and/or encryption and more data.

There are tons of other areas you can think of where the CPUs just aren't that fast.

With a fast enough CPU enough people may decide that performance would be acceptable in most cases to implement SSH servers in a reasonably sane+popular language just to avoid buffer overflows in what should be security software.

Re:Will software catch up?

[TheLink]

If Erlang changes you'd have to change the hardware. This might be acceptable for stuff that never gets upgraded or changed, but otherwise, nah.

There are many good reasons for the abstraction layers between a programmer's chosen programming language and the CPU.

If Language X goes out of fashion you can still use something else.

Plus, with the abstraction layers it is easier to have programs originally written in Language X AND Language Y AND Language Z on the same CPU, all executing at a decent speed.

Unless you're going to write (or worse - rewrite) everything in Erlang, that 10MHz Erlang chip isn't going to be very useful in practice.

Re:Will software catch up?

[Anonymous Coward]

Kernel 2.6 took more than one minute to compile.

Computers are not fast enough yet.

Re:Will software catch up?

[John Courtland]

You could also just upgrade the microcode, in which case you would have a more future-proof design. (In thoery anyhow, Capitalism doesn't enjoy future proof designs).

Re:Athlon Performance Ratings

[naztafari]

"is AMD going to adjust its PR Rating to the new cores that Intel has?"

I don't think so. AMD Athlon PRs are not measured against Intel Chips.

AFAIK, the AMD Athlon PR numbers are the newer CPUs' (Athlon XPs, 64s) ratings against the older Athlon Thunderbirds which were the last ones that were labeled and sold in MHz/GHz.

So roughly, an Athlon XP2600+ would be akin to an Athlon Thunderbird that was theoretically made to run at 2.6GHz

Remember, a 1.33GHz Athlon Thunderbird stacked up pretty well against a 1.7GHz P4 back then, and only lost out on SSE optimizations.

What market is it for?

[MountainMan101]

You have to ask the question!

With the Athlon 64, IA64 and G5 vying for the 64bit market, and Athlon offering native supports of 32-bit binaries. Why would anyone want a new series of Pentium 4E?

Is Intel feeling that Athlon may be about to make leaps and bounds in the small business/desktop market?

Re:What market is it for?

[Anonymous Coward]

Because there's *no* reason to buy a 64-bit chip unless you're running a server or development. Windows currently is not 64-bit and that being the market dominance why would you release a chip that 95 percent of the home users can't use till possibly 2006?

It's like trying to sell the orcs of mordor solar powered cars. They won't work till the return of the king. :)

Re:What market is it for?

[Anonymous Coward]

It may be helpful to think about what market it is clearly not intended for. By not being 64 bit, you would not be able to work with very large amounts of RAM (say 8GB), which would make it totally inappropropriate for high-end digital video editing, servers with extremely large session caches, or scientific/engineering applications. That rules out anything interesting, so it must be meant for Windows.

Re:What market is it for?

[pi radians]

"It's like trying to sell the orcs of mordor solar powered cars."

ARGH! I just hate car analogies when discussing processors! ;)

Re:What market is it for?

[Hoser McMoose]

Uhh, Microsoft is planning on releasing WinXP for AMD64 processors in Q3 of this year (to coincide with the release of SP2 for WinXP IA32). Windows is already 64-bit, and has been for a while, but only for IA64 (Itanium).

MS will also bring out a 64-bit/AMD64 version of Win2003 Server at about the same time. Hardly 2006.

Page info

[after]

Actually, you can link it. Its still a large ammount of information to take in though. Good midnight read, I guess :|

Re:Page info

[Emil Brink]

Indeed you can. That's weird, because I actually am geek enough to read the JS-link, and to guess that it should be possible to paste together, but when I tried something that my memory tells me was identical to what you've posted, it failed. I guess my memory sucks, then. Good job!

British Politician

[anaplasmosis]

Intel did realise that "Prescott" is a much lampooned British politician, didn't they? Official bio here; http://www.number-10.gov.uk/output/Page1376.asp Highly entertaining game here; http://www.urban75.com/Punch/prescott.html

Just suck it up Intel, GIve us what we want!!

[OlivierB]

Is it just me or does anybody have the felling Intel is completly lead by Marketing GHZ frills? Looks to me like they didn't make the most efficient chip, they just designed a straight shooter for 4/5 GHZ. We all know their current P4 Extreme are real Power hogs and not all that efficient. Thus my question, why can't they focus on delivering a 9nm version of the Pentium M? With it's low consumption and heat they could have surely clocked this big boy in the 3.2GHz area and taken care of AMD. All these benchmarks won't make a difference as Mom & Pop will go to COMPUSA and be this computer is Faster-than-3Ghx-because-it's-a4-Ghz. Time to get some AMD stock

Re:Just suck it up Intel, GIve us what we want!!

[Muad'Dave]

...delivering a 9nm version...

9nm??!! The quantum effects would be crippling! Oh, you mean 90nm. Whew! :-)

Re:Just suck it up Intel, GIve us what we want!!

[bhtooefr]

They don't want to admit that P6 will last a VERY long time. In their opinion, P6 should have died the day the first Williamette was released, and possibly earlier. P6, however, was the only option for a mobile CPU. I don't think that Banias OR Dothan could have hit 3.2GHz without insane cooling, but they could easily hit a "4000+" rating on either the AMD Duron 1.0-based scale or a Northwood or Prescott-based scale.

Re:Just suck it up Intel, GIve us what we want!!

[Mr. Piddle]

We all know their current P4 Extreme are real Power hogs and not all that efficient.

Reflecting on the recent SUV craze in the USA, this really isn't hard to understand. However, markets do change--just look at the newer generations of station wagons labeled "crossover" SUVs. People are realizing that they never really wanted a 10,000lb SUV all along, and we're moving back to the early 80's super-practical family mover.

One thing that hasn't penetrated in the computer markets is that 100W CPUs really can cost tens of dollars per year in extra power consumption relative to more efficient CPUs. Now that computers are getting well under $1000, those tens of dollars might not seem so trivial any longer. I predict we'll see the current no-holds-barred CPUs move aside for more reasonable ones by the end of the decade as the computer markets mature.

Re:Just suck it up Intel, GIve us what we want!!

[shotfeel]

Just MO, but I see the biggest force pushing down power consumption being the good-old portable. People don't want laptop computers that are half the speed of a much cheaper desktop box (at least I don't).

So it's kinda like the celeron?

[josh glaser]

Not technically, of course, but in it's market position and the like? About time...it's good to have a low-end "mainstream" processor, but the celeron is just plain old. The celeron should be officially replaced, or better yet, discontinued without a replacement. But anyways, about Prescott: Well, another new version of the P4 is nice, but, if it's an economy processor, shouldn't they call it, like, Pentium E (a la Pentium M)? Simply calling it Penitum 4 and then slapping E on the megahertz rating is just going to confuse people. Of course, I could be wrong. Maybe intel wanted to make another ultra high-end extreme processor, and failed miserably at it. :)

Re:So it's kinda like the celeron?

[Slack3r78]

The "slap a letter on the end" to distinguish chip revisions approach has worked fairly well for Intel so far. P4 2.4C, for example, would indicate a HyperThreaded P4. While I personally prefer AMD chips, this is one thing I do feel Intel has gotten right - the only real way to distinguish which core revision you're getting with AMD is to look at the chip in the package, so you had to be able to distinguish the chips visually. Not a problem for me as I spend a lot of time working with them, but it's a problem for those that aren't that familiar with the chips - especially for chips like the XP 2000+ that went through at least three different cores.

Re:So it's kinda like the celeron?

[Anonymous Coward]

No, it's not at all like a Celeron. The "E" isn't for economy, it's just their way of differentiating the processor. And it's the older chips getting the E anyhow.

The slow initial speeds aren't a failure. The performance stucks now because the chip is designed to scale to very high speeds (as has been mentioned about 1,000 times already). When it hits 4+ GHz you'll see more competative numbers.

Re:So it's kinda like the celeron?

[TheLink]

Well over here you can tell the difference between the athlons - the price lists usually tell you whether they're Barton (or previously- Thoroughbreds, Palomino, Thunderbird). And they are sold as such.

The people who don't care can stick to the nnnn+ XP numbers, and I doubt it'll really make a diff to them.

Re:So it's kinda like the celeron?

[Slack3r78]

Only if a reseller is clearing stock. For example, when the Barton 2600's where released, they were actually slightly cheaper than the Tbred-B 2600's at the time (basing this on NewEgg's pricing). Of course New Egg makes a point of noting which core the CPU they're selling you is, but not all resellers do that. My point was that Intel's code letters make it absolutely clear which chip revision you're getting, leaving little room for confusion.

Re:So it's kinda like the celeron?

[TheLink]

http://www.lowyat.net/

Price lists linked from left sidebar.

BTW: RM3.8=USD1

Prescott?

[Anonymous Coward]

As i come from the UK, it makes me happy that Intel's fattest chip is called Prescott.

Linux

[millahtime]

Has anyone done a test of the AMD64 running a 64 bit OS vs P4 running a 32 bit OS? Say Linux. To see the difference when they full power of the chip is taken advantage of. Especially with rendering. It may be a little like comparing Apples and Oranges but comparing these 2 chips can be that way. And to throw in a 32 bit OS on the AMD64 chip isn't really fair to it's power. It's not really using the full potential of the chip.

Re:Linux

[Anonymous Coward]

I only upgraded recently so Ill have to wait for the Intel Rik Walla

Re:Linux

[Anonymous Coward]

Infoworld just did a great review of RHEL AS 3 using Vanilla 2.4 and 2.6 kernels on Opteron, Xeon and Itanium systems. Since RHEL has native x86, x86-64 and Itanium ports that should be pretty fair. I'll give you a hint. Opteron burned 'em all

Re:Linux

[TR0GD0RtheBURNiNAT0R]

And to throw in a 32 bit OS on the AMD64 chip isn't really fair to it's power. It's not really using the full potential of the chip.

Perhaps not. Unfortunatly, for the 45 billion of us stuck with Windows, 32 bits is all we can work with. With that in mind, I say that it still is fair game (for now, at least) to compare AMD64 chips with Intel chips in a 32 OS.

Though it would be nice to see Intel procs get smoked even more... =)

I think...

[cozziewozzie]

...he means this hard cop.

Origin of the name Prescott

[andy666]

Prescott is the assistant to the navigator in Moby Dick, and he is known for being excellent at making rapid computations.

Re:Origin of the name Prescott

[stevel]

Intel uses "geographic" names (towns, mountains, etc.) for all its code names, since they can't be copyrighted. Prescott is a town in Oregon (a state from which Intel draws many of its code names), so I think that the connection to Moby Dick is just a coincidence.

Re:Origin of the name Prescott

[stevel]

.. copyrighted or trademarked, I should have said...

Re:Ironically (OT)

[ErrorBase]

Use the folowing Regexps ...
1) /.*adtech\..*/
2) /[^o]ad[\._svi]*[/\.]/
3) /b[ae]nner[s\.\/]*/

And things should be a lot clearer ...

The Tech Report

[muyuubyou]

16 pages, but they're properly labeled and you can always jump to the conclusion.

Linkie

Their marketing prolicies are really sucks.....

[deconvolution]

It seems that both AMD/Intel are trying to double everything in a single CPU, such as L1, L2, even DDR to DDRII :). I just wonder why they have lack product lines of dual-CPU systems for main consumers? Their SMP solutions are either very high-end not going to face main consumers, or fading to the market because of slow enough (both of them always have unresonable prices).
For most people, the most areas Pentium4/Althon XP take advantages are 3D applications, data servers, and some scitisfic applications. However, a SMP system with two main stream processors also can achieve the simpilar(just slower a little bit) scores. Those applications always can be implemented through parallel approaches. (I believe it already have done this during the designing time....).
For example, for SMP solutions, I have to choose between Operton and Althon MP, but in actually I want a dual-althonXP with the double prices. I think that such system is what many other people really want to buy insteading of investing massive money on new processor/cooling system for better performance.... I believe there is not a big technical matter for this just trying to force us follow their single processor upgrading ways....
"Why spend billions, when you can spend millions?"

Re:Their marketing prolicies are really sucks.....

[Slack3r78]

I think you're pretty much off base here. The Athlon MP is an XP with SMP support, much as the Athlon FX is essentially a 1xx series Opteron. If you want SMP, you're going to have to spend more as the extra transistors needed to handle SMP don't come free. As it stands, an Opteron 240 runs about $215 and a board for it's going to cost your anywhere from $200 on up.

I think the biggest point that you're missing is the fact that increasing the number of processors in a system does not linearly increase the processing power of that system. For most people, it's far more cost efficient to simply opt for a faster single processor as a second CPU wouldn't be utilized properly most of the time anyway. All that said, if you absolutely have to have a consumer level SMP system, I know of at least one company that will happily sell it to you - but again, you have to be willing to spend the money if you want SMP.

Re:Their marketing prolicies are really sucks.....

[deconvolution]

The Athlon MP is an XP with SMP support... The fact is that Althon XP is Althon MP disabled SMP support. Some reports said the dual-XP had the same (even better) performance compared to Althon MP.
I think the biggest point that you're missing is the fact .... Yes, the second CPU wouldn't be utilized properly most of the time, but the single P4 3G is the same, which cannot say always 100% utilized for those applications you mentioned!

Re:Their marketing prolicies are really sucks.....

[bhtooefr]

Actually, the extra transistors on the MP are already there on the XP, it's just a bridge on the XP isn't filled, so it doesn't identify as an MP.

Also, the extra transistors are on the pre-Tualatin Celerons. Ever heard of the ABIT BP6? A dual-S370 board from when S370 was only used by Celerons.

Re:Their marketing prolicies are really sucks.....

[philthedrill]

I think what you're getting at is, "Why build/sell a fast uniprocessor when you can use cheap processors and connect them into multiprocessor systems for high performance?"

There are some issues with this. Parallel programming is hard, and not every application supports it. In fact, most programs aren't optimized for MP systems, because worrying about that stuff is a pain in the ass (synchronization, locks, etc.). So on an application level, it doesn't really benefit most consumers. The best way to sum it up is with Amdahl's Law, which states that your speedup for an application is limited by the fraction of your code that is parallelizable. In short, you'll get the greatest benefit by working on your single threaded performance.

Second, and you probably already know this, but adding processors does not linearly increase your performance (e.g. adding a second processor does not double your performance). Sometimes it is linear (or even superlinear), but usually, it's not. There's overhead with cache coherence involved, and some contention with memory accesses, and (back to Amdahl's Law), you're limited by how much of the code is actually parallelizable (which is likely to be much less than 100%).

But one benefit of MP systems is that the OS can load balance computationally intensive apps on separate processors. The real question is, do the majority of users out there really hammer their systems like that? I think the answer is no, and those that do are willing to pay the premium.

So after all this, if you're still adamant about buying a multiprocessor system, consider the tradeoffs. On one hand, the Athlon MP only goes up to 2P. It's also a point-to-point SMP, which can mean a bit higher latency on coherence activity than a shared bus. The Opteron is glueless, so the latency of coherence activity should be quite low. But, the Opteron is not SMP, it's NUMA, which often requires some optimization for memory placement. But it is the future for AMD, it scales well, software support is getting better by the day, and it's 64-bit (if that matters to you).

Re:Their marketing prolicies are really sucks.....

[chez69]

Do you know what NUMA is? NUMA systems are where you have seperate processor/memory boards that look like one machine. a 2 way Opteron machine is NOT a numa system.

Re:Their marketing prolicies are really sucks.....

[deconvolution]

Parallel programming is hard, and not every application supports it.
======= Yes, but most of high-end applications support it since they usually are deveolped from a multi-processing environment. The other single thearding programs usually are not processor intensive.

you're limited by how much of the code is actually parallelizable
======= The acutal performances do not depend how many percentages of the whole code being parallelised but related to how much _core loop code_ is parallelised. Regarding my little experience for a 3D application, 95% processor time was spent on 5% core loop code and the others for 95% code....

Re:Their marketing prolicies are really sucks.....

[shotfeel]

There are some issues with this. Parallel programming is hard, and not every application supports it.

and

The real question is, do the majority of users out there really hammer their systems like that? I think the answer is no

I have to say I disagree. IMO many people, including home users, are running more than one application at a time -that means the apps don't have to have multiple, parallelizable threads to take advantage of multiple processors. Maybe its the people I hang around with, but more and more of them are working more with multimedia files -wether it be listening to music while surfing the web or capturing/transcoding the latest home movie while playing the latest Yahoo game they're hooked on. They can stress even the fastest systems.

Second, and you probably already know this, but adding processors does not linearly increase your performance

and

those that do are willing to pay the premium.

If you can buy two 2 GHz processors for the price of one 3.4 GHz processor, you may not paying a premium at all, and two 3 GHz processors get you performance that's simply not available otherwise.

I agree there are a number of people who really only do one thing at a time with their computer and don't stress it too much, but I think the number of people who could benefit from multiple processors is waaay underestimated.

Re:Athlon Performance Ratings

[Dan Ost]

The way it works is that an XP2600+ is 2.6 times faster than a 1GHz Duron,
a 3000+ is 3 times faster than a 1GHz Duron, etc.

This is according to "PC Hardware in a Nutshell" 3rd edition (O'Reilly).

Can anyone back this up with a reference from AMD?

Editor sk1llz

[PhraudulentOne]

Hey there editors, nudge nudge, wink wink, its HardOCP, not HardCOP.

my next server will be

[Anonymous Coward]

a dual Athlong, since it's better, and cheaper. I guess Intel is going down the path of no return. Oh well. The same could be said of Microsoft's LongHorn. Here's to hoping MS digs in deeper with LongHorn and then realize a year after it's release that it blows.

Still 32 Bit...

[axis-techno-geek]

That is just so 90's ;)

It's the early Pentium 4 all over again.

[MtViewGuy]

Does anyone here remember the early Pentium 4 1.4 and 1.7 GHz chips built on the Socket 423 form factor? With only 256 KB of on-die L2 cache and few programs (at that time) that could take full advantage of the Pentium 4's SSE2 multimedia extensions, small wonder why the CPU was much-disliked originally. It wasn't until Intel came out with the newer Pentium 4's with the 512 KB L2 cache and software that fully took advantage of SSE2 extensions that the CPU finally took off in popularity.

I don't see the Prescott-core CPU's become popular until software catches up with supporting all the functions of the CPU; we may see that with Windows XP Service Pack 2 and later builds of the Linux 2.6.x kernel.

Re:It's the early Pentium 4 all over again.

[ameoba]

My first response was to say that the OS isn't really going to care about the new instructions, but then I looked to see what was actually in SSE3...

Hyperthreading specific instructions?

Anyone got any real info on them; from what I've read so far, the specifics of what SSE3 can do are kinda fuzzy...

Re:It's the early Pentium 4 all over again.

[ph4rmb0y]

It wasn't until Intel came out with the newer Pentium 4's with the 512 KB L2 cache and software that fully took advantage of SSE2 extensions that the CPU finally took off in popularity.

Well, from where I sat, the P4 didn't take off in popularity until Intel discontinued the P3 ...

Re:It's the early Pentium 4 all over again.

[Luminous Coward]

Search for Prescott New Instructions Software Developer's Guide on Intel's site.

Re:It's the early Pentium 4 all over again.

[kavachameleon]

I have one of those chips, you insensitive CLOD! Actually, it suits me just fine, and I don't have money for an upgrade. I also am running with RDRAM....

However...

[91degrees]

The "Fat Labour Politician" processor is likely to be delayed.

Hard O - C - P

[Anonymous Coward]

Uh, Hemos, that's HardOCP, not HardCop. Hard Cop sounds like a pr0n title!

So... if you half the cache speed...

[Ayanami Rei]

you're only back where you need to be at 3.0x2 = 6.0 GHz. Nope, it looks like they've screwed up that part. L1 latency is a killer.

speed vs design

[john_uy]

it is quite interesting that the new prescott core has a lower latency for the caches and higher stage pipeline.

but given the fact that a big percentage of decrease in latency from existing northwood cores and big increase in pipeline does not reduce the speed *significantly.* it can still compete with the current northwood with a small drop in performance on a clock per clock basis.

given these things, i think the cpu may be designed quite well given its current performance with numerous internal slow downs. i'm sure in their next core, they will be able to reduce the latency by significant amounts and increase the branch prediction system thereby causing their future cpus to perform better than current iteration.

i believe the current purpose of prescott is to do a couple of things. first, refine their 90nm processing of the cpu. they will be able to iron out manufacturing bugs (like yields.) they will also be able to improve in the design of the cpu (to put minor revisions to improve the manufacturing or even performance.) they will be able to earn more (since 90nm should product more yields for them.)

probably, i believe that in around 1 year's time, just like their transition from williamete to northwood, their cpu will be much faster. they should be able to solve the latency of their cache. they may already adopt a very good branch prediction unit that will reduce the effects of a very long pipeline.

also, this year will be a transition year of technologies. so pretty much everything you buy not will almost be worthless by next year. the cpu packaging will be changed to lga. slots in the computer will feature pci express. i/o will be standardized with usb. storage devices will be sata. intel will be prepping up for speed wars next year. (i think this usually happends every other year where there is a speed war and there is a slow increase in speeds by both sides.)

Re:speed vs design

[the melon]

Read the articles again. The cache latency of the trace cache, data cache and L2 cache all went up, not down. Infact all of them almost doubled.

Re:speed vs design

[Anonymous Coward]

Also, the execution engine is no longer double-clocked. So there are three strikes against it right off the bat: higher latency, 11-step longer pipeline, slower execution engine. The fact that it is equaling the performance (a little better in 1/3 of benchmarks, a little worse in 1/3 of applications)of Willamette, is very profound.

It is clarly a very efficient design. The high power consumption despite the improved transisters hints that more of the chip is being used at once (the improved hyperthreading). The multi-tasking benchmarks seem to indicate this. Look for this procesor to ramp in speed very quickly. The architecture change will enable this ramp. It's much the same as the Pentium III to IV change. However, last time, Intel did not understand that the public would associate a name change, not with an architecture change, but with performance improvements (future headroom does no good for the consumer). This time, Intel kept the hype down, and did not go the rate of calling it the Pentium V. I suspect they are saving the Pentium V name for the new package and 64-bit extensions, probably with chipsets designed around DDR II and PCI Express.

Re:speed vs design

[Loki_1929]

Who's feeding you this stuff?

"given these things, i think the cpu may be designed quite well given its current performance with numerous internal slow downs."

So.... new CPUs are supposed to run slower than those they are replacing? I see...

"i'm sure in their next core,"

Which won't be just a core; it'll be a whole new CPU running the Tejas core, and will probably be marketed as a Pentium 5 or thereabouts. It's due out some time early next year.

" they will be able to reduce the latency by significant amounts"

What latency? Where? How much of a latency redction are you expecting to see that you refer to it as "significant"? Do you have information sources with which to back this up, or is this something you gathered from an in-depth conversation with Ms Cleo at $4.99/min?

"increase the branch prediction system"

What does that mean?! I assume you mean that they're going to improve branch prediction. Well now, this helps to a certain extent with predictable branchs such as those seen in simple loops. Currently, branch prediction units are rating somewhere around 90 - 95%, depending on which Intel/AMD tech docs you're reading. How much of an improvement do you hope to accomplish above and beyond this?

"thereby causing their future cpus to perform better than current iteration."

So Prescott performs slower than Northwood, and this is Good(tm), and the next "core" is going to perform faster than Prescott, and this is also Good(tm). Seems like Intel just can't lose - make a chip that's slow; good - make a chip that's fast; good. Amazing.

" first, refine their 90nm processing of the cpu. they will be able to iron out manufacturing bugs"

You're kind of supposed to do that before you start shipping the CPUs.

"probably, i believe that in around 1 year's time, just like their transition from williamete to northwood, their cpu will be much faster."

Again, your conversations with Ms Cleo are really just a waste of your money, as she's feeding you bad information - possibly from evil spirits. In a year from now, We'll have Tejas-core CPUs, probably called Pentium 5s, shipping to the masses. They'll probably have 64-bit instructions at the ready, if not enabled, and they'll probably also be dual-core capable. In a year from now, if they're still trying to hammer out problems with Prescott, they may as well start reselling Opteron CPUs.

" they may already adopt a very good branch prediction unit that will reduce the effects of a very long pipeline."

They already have an excellent branch prediction unit. Unfortunately for your comment, it has little to do with the long pipeline. The effect of the longer pipeline is to allow for scaling to higher clock frequencies. By lowering the IPC and extending the pipeline, they can ramp up in frequency, whereas Northwood was approaching a frequency ceiling. Higher clock frequencies offset longer pipelines, just as better branch prediction somewhat offsets high latency from cache misses in certain situations.

"so pretty much everything you buy not will almost be worthless by next year."

Completely incorrect. What you buy now will always be outdated in a year's time - this is true of any given time in the tech industry. However, computers being bought now will continue to function in an increasingly limited fashion for the foreseable future. I still see PCs in use from the early 1990s. They don't run the latest and greatest software, but they often do all that which is required of them by their respective users.

" the cpu packaging will be changed to lga."

For Intel CPUs. Sockets for AMD CPUs will remain the same (with the exception of an addition of 939-pin) for the forseable future. Even socket A will live into 2005, according to AMD's brass.

"slots in the computer will feature pci express."

Re:speed vs design

[john_uy]

sorry about that. freudan slip i guess? i was thinking of increased latency times and not lower.

Re:speed vs design

[john_uy]

So.... new CPUs are supposed to run slower than those they are replacing? I see...

i am not looking at the marketing aspect. i am just analyzing it from a technical perspective.

What latency? Where? How much of a latency redction are you expecting to see that you refer to it as "significant"? Do you have information sources with which to back this up, or is this something you gathered from an in-depth conversation with Ms Cleo at $4.99/min?

its just like what happened with the transition from willamette and northwood. the cache was improved (aside from increasing it) and latency decreased.

What does that mean?! I assume you mean that they're going to improve branch prediction. Well now, this helps to a certain extent with predictable branchs such as those seen in simple loops. Currently, branch prediction units are rating somewhere around 90 - 95%, depending on which Intel/AMD tech docs you're reading. How much of an improvement do you hope to accomplish above and beyond this?

from the willamette to the northwood, intel improved the branch prediction of the cpu. (sorry that i typed the wrong term of increase instead of improve.)

Again, your conversations with Ms Cleo are really just a waste of your money, as she's feeding you bad information - possibly from evil spirits. In a year from now, We'll have Tejas-core CPUs, probably called Pentium 5s, shipping to the masses. They'll probably have 64-bit instructions at the ready, if not enabled, and they'll probably also be dual-core capable. In a year from now, if they're still trying to hammer out problems with Prescott, they may as well start reselling Opteron CPUs.

does it ever occur to you that by releasing the prescott now, they will be able to refine their design. (as the stepping increases, they are able to improve the yields of higher performing cpus.) so by the time tejas arrives, it will be much easier for them.

They already have an excellent branch prediction unit. Unfortunately for your comment, it has little to do with the long pipeline. The effect of the longer pipeline is to allow for scaling to higher clock frequencies. By lowering the IPC and extending the pipeline, they can ramp up in frequency, whereas Northwood was approaching a frequency ceiling. Higher clock frequencies offset longer pipelines, just as better branch prediction somewhat offsets high latency from cache misses in certain situations.

aside from the frequency, improved branch prediction will, of course, lessen misprediction thereby reducing the stalls in the very long 31 stage pipeline. if intel did not improve on their branch predictions and increased the pipeline then you will see slow downs since you may see a lot of mispredictions. one way of improving the prediction system is increase the (memory or cache, i forget) that will store the history of commands so the chip will avoid misses in the future.

Completely incorrect. What you buy now will always be outdated in a year's time - this is true of any given time in the tech industry. However, computers being bought now will continue to function in an increasingly limited fashion for the foreseable future. I still see PCs in use from the early 1990s. They don't run the latest and greatest software, but they often do all that which is required of them by their respective users.

you may have a working computer but the resale value is no good. i am implying that big architectural changes to the computer system will start this year - not just the cpu. from the memory (ddr2), i/o (pci express), storage (sata.) almost all the internal workings of the computer in terms of interface at least will start to change by this year. it is no longer very useful if you will buy i.e. lots of hard drive that have pata while there will be sata (and of course pata will still be there but they will start reducing the number of available ports in favor to sata.) and maybe in two years time, if you may purchase a 100gb hard dri

Re:speed vs design

[TR0GD0RtheBURNiNAT0R]

this year will be a transition year of technologies. so pretty much everything you buy not will almost be worthless by next year.

So, remind me how this is different from any other year?

The funny thing is...

[Maelstrom696969]

...that of course, Intel just keeps ramping up processor speeds in order to improve performance. They don't seem to know any other way to do it.

When AMD shrunk the pipeline and started waking people up to the fact that there's other ways to skin a processor than just upping the clock, that was bloody revolutionary. But yet, not only did Intel not abandon the narrow-minded "higher clock=faster" mentality with Prescott, they have actually fed it by making it so that they HAVE to make higher clock speeds in order for Prescott to perform at peak optimization and efficiency! Sheesh.

Trust me, I'm of course not saying that higher clock speeds aren't better - just that AMD has proven in the past that there's other things that can be done besides that, and it's just a matter of time before they come up with something else to beat Intel down with, IMHO.

---A witty .sig proves nothing.

Re:The funny thing is...

[angle_slam]

Intel already knows that processor speeds don't necessarily mean performance. See, e.g., Centrino.

SSE2 vs. SSE3

[Kjella]

SSE2 *had* to be used in many apps to get decent performance, since the FP performance sucked. SSE3 on the other hand are very few instructions - and if you look at the DivX 5.1.1 benchmarks that supposedly is using the new LDDQU instruction, it can't even beat the equivalent Northwood.

The rest also sounded like "special interest" functions, much like many of AMDs 64 bit extensions. Great for specific uses like scientific calculations and cryptography, not that much for general computing. Cache is already quite high at 1mb, if you look at die size any higher would seriously add to Intel's costs.

The only real promise there was the ability to ramp up speed, though that's not a bad one. But then again, there's no doubt AMD has things to match. One being the switch to x86-64 with more registrys, two being the move to 90nm process, third getting dual channel DDR support on their AMD 64s (the FX I consider a special interest processor for the time being) and so on.

Bottom line? Looks like the CPU market will remain highly competitive for quite some time to come, for the good of all us consumers. But I didn't see any figures on Prescotts power consumption, anyone see it? I imagine a 5GHz Prescott would start to drain a *lot* of power (= cost, noise, heat)...

Kjella

Re:SSE2 vs. SSE3

[EMN13]

There were some figures on power consumption that suggested Prescott runs quite a bit warmer than Northwood and the same clockspeed.

That's not really surprising either: A feature-size shrink (such as the move now from 130nm to 90nm) lowers the power per transistor, however, it increases the power per unit area (IIRC proportionally to the sqrt of the shrink).

Of course - a different processor layout could completely counteract these effects, and a change to 31 stages instead of 21 stages would qualify as such I suppose :-).

http://www.aceshardware.com/read.jsp?id=60000317
does list some power data, and rates a 3.2 GHz prescott as having a 103W _typical_ power consumption...

Conclusion the prescott has a power consumption problem; hopefully later revisions will fix this...

Less efficient?

[Anonymous Coward]

Where do you get this notion that Prescott is less efficient? Because it has a deeper pipeline? There is one rule for processor design - If it increases performance, put it in. There are hard limits to what you can do at a certain clock rate as the speed of light only moves so fast.
For a really efficient processor, you can get an asychronous one which will run exactly as fast as possible given its voltage and temperature. Intel has more research in this area then AMD does, having designed a fully asychronous scheduler.

Re:Athlon Performance Ratings

[naztafari]

They got it the other way around, but that's more or less correct.

A 1GHz Duron is roughly the same speed as a 1GHz Athlon Thunderbird since the 1GHz Duron was the then newer "Morgan" cores (which came out with the Athlon XPs). Benchmarks then showed that 1GHz Morgan Durons ran more or less at the same speed as the older 1GHz Thunderbird Athlons.

The increase in performance mostly came from the additional SSE instructions present in the 1GHz Morgan core Durons. Yep, especially since Quake III makes heavy use of SSE Instruction sets and used to be Intel-centric

I don't want to bother slogging through AMD Tech specs, but the PR numbering system I rattled of was from a rough memory of reading AMD tech specs over naming conventions with their Performance Ratings.

I'm waiting for Quantum

[theatre_freak]

Prescott? I'm still holding on with my 486/66 waiting for Quantum Computing.

Re:I'm waiting for Quantum

[TR0GD0RtheBURNiNAT0R]

I hear ya, brother!

...oh, wait you said quantum computing, not money... My mistake.

Freuden slip on Hemos' part.

[Anonymous Coward]

Well, we know what to get him for his birthday.

Informative?

[Wesley Felter]

What do trusted computing features and operating systems have to do with the pipeline length? I can't tell whether these posters are clueless or trolling.

Prescott

[HogynCymraeg]

So can it beat its predecessor, the Northwood?

With a swift left jab, a Prescott can beat anything. Even egg throwers from Wales.

HardCop?!?

[Anonymous Coward]

I think you meant [H]ard|OCP

I don't even want to speculate on what kind of site HardCop might be...

What about all the Trusted Platform Stuff?

[mh_cryptonomicon]

A worry about Prescott is that it's supporting all the Trusted Platform felgercarb that's been in the news recently. Cryptonomicon.Net has a few links to related web sites and opinions in the article: Intel Debuts Prescott.

Power dissipation: 89-103 Watts

[wirelessbuzzers]

89-103 Watts max power dissipation

Ick. That's gonna hurt.

Re:Power dissipation: 89-103 Watts

[Junks Jerzey]

89-103 Watts max power dissipation

Ick. That's gonna hurt.

And this is the big issue with the Prescott and x86 architecture in general. Sure, Moore's law, blah, blah, blah. But diminishing returns kick in hard if power consumption goes up at the same rate, and we're seeing some scary numbers now. It already looks like we'll be past 150 watts by the end of the year. How long can this continue?

Re:Power dissipation: 89-103 Watts

[TheLink]

IMO the x86 stuff plays a very minor role if at all in terms of power dissipation issues.

If you look at the various chips in spec.org (Sun SPARC, POWER, PRIMEPOWER, Alpha ) many of them dissipate about the same amount of power given the same performance level.

Design and fab methods seem to play a bigger part nowadays. e.g. SOI or not SOI. More cache or less, cache 100% active all the time or not, etc.

The highest performance PowerPC chips are lower power than the x86 equivalents, but the other chip vendors don't do as well. Look at the Itanium 2 (Madison)- 107W @1.5GHz (thermal design power).

If you talk lower performance chips, then there are the Transmeta chips, the various low performance x86 chips, or the Pentium-Ms. These are pretty power efficient.

x86 or not seems rather irrelevant to power consumption.

Re:Power dissipation: 89-103 Watts

[Junks Jerzey]

x86 or not seems rather irrelevant to power consumption.

The difference is that the x86 line represents the processors used by businesses and individuals in desktop machines. If the the power requirement doubles, then this is significant. It used to be that the monitor was the biggest source of power, but this is no longer true now that LCD displays are becoming standard.

Re:Power dissipation: 89-103 Watts

[TheLink]

My point was the x86 architecture isn't the real cause of high power consumption.

You said: "And this is the big issue with the Prescott and x86 architecture in general."

Does anyone know

[Ann Coulter]

when and where the technical resources for SSE3 will be released?

Re: NUMA

[philthedrill]

Separate processor-memory boards isn't what defines NUMA; it's the way the processor and memory are organized. Each processor has a local memory which has a lower latency than the remote memory access. It's for scalability, since SMP systems don't scale as well when it comes to bandwidth and memory access times.

Note to mods:

[Anonymous Coward]

If you have no clue about a subject... you are not qualified to moderate posts regarding that subject as informative.

A post can say "you're stupid, here's how it really works" and still be complete BS.

Re:Note to mods:

[Anonymous Coward]

I wasn't expecting to get modded informative... it was supposed to be funny.

Re:Athlon Performance Ratings

[Anonymous Coward]

> Can anyone back this up with a reference from AMD?

That's the problem -- AMD used to publish benchmarks showing the relationship between PR numbers and the 1Ghz chip, but now all they publish is a Mhz to PR formula.

It's safe to say that AMD's numbers are almost entirely driven by what Intel has on the market.

Huh?

[Loki_1929]

"Basically, looks like it's faster, but still not the fastest in all areas."

Just what definition are you using for the word "faster"? To my eyes, it's slower than the older Northwood core in the majority of real-world situations, clock-for-clock. If you're talking about absolute performance, then it's significantly slower than, say, the AthlonFX CPUs. Even the biased-as-hell airbags at Tom's didn't have much good to say about this CPU. That's not to say that it wont see strong performance gains as applications are recompiled to support the new features (SSE3 et al) of Prescott, but I don't have the slightest idea where from where your statment comes based upon the reviews and benchmarks published thus far.

You make it sound like it's a superb new chip that outperforms almost every other chip on almost every application. That's criminally wrong.

Re:Athlon Performance Ratings

[Loki_1929]

" The way it works is that an XP2600+ is 2.6 times faster than a 1GHz Duron, a 3000+ is 3 times faster than a 1GHz Duron, etc.
This is according to "PC Hardware in a Nutshell" 3rd edition (O'Reilly).
Can anyone back this up with a reference from AMD?"

Nope, but I can happily dispute it. According to this document from AMD's site, it's based on the performance that would come from a 2.6GHz Athlon processor. The specific core to which it's compared is not detailed, however it would only make sense from a marketing perspective to compare it to the last core used prior to PR implementation, which was the Thunderbird core. The Thunderbird core, as any other, will not necessarily scale in performance in a perfect linear fashion simply by increasing clock frequency. Thus, it is more correct to state that an AthlonXP 2600+ performs similarly to a Thunderbird-core Athlon at 2.6GHz, if such a creature were to exist. To say that it is 2.x times faster is to belie the reality of diminishing returns.

new egg cooker

[ahmetaa]

You can throw your old Athlon Thunderbird's, Becuase it takes only 2 minutes to make scramble egg by using prescott. http://www.aceshardware.com/read.jsp?id=60000317

varying multiplier?

[LuxFX]

Is a varying multipler new or has that been done in processors before? The multiplier is listed as "14-16"! In low-speed (3.2Ghz - 3.4Ghz) I see that is uses 16 multiplier. But in the AnandTech overclocking test they got it up towards 3.7Ghz and it was using a 14 multiplier. A shame, because in that test, the FSB was over 1Ghz! At a 16 multiplier would have resulted in speeds close to 4.2Ghz.

Can somebody fill me in on what's going on?

Re:varying multiplier?

[Anonymous Coward]

The varying multiplier is to emulate three different speeds of the processor.

Re:varying multiplier?

[LuxFX]

So the varying multiplier is only on the engineering sample? That makes sense.... Thanks!

Benchmarks?

[ProtonMotiveForce]

Intel isn't in the benchmark business. They try to make money.

Prescott will be cheaper, they can make more of them, and it will scale to higher speeds (good for marketing). That's what's important, not benchmarks, at least to Intel, and to Intel's stockholders.

WTF naming convention?

[Anonymous Coward]

Northwood? google on Operation Northwoods ... a spook program to spoof terrorist attacks on the U.S.

Prescott? As in Prescott Bush?

176.gcc

[Anonymous Coward]

Anyone else notice the almost 18% decrease in the gcc misprediction rate on the 176.gcc portion of SPEC CPU2000?. Has anyone done any real-world gcc benchmarks under Linux?

Re: NUMA

[chez69]

thanks for correcting me.

pricing

[Anonymous Coward]

I havn't searched for any pricing on the new P4, but does anybody think that Intel is actually going to reduce prices on it? Instead of seeing increased core production as a way to lower prices for us, they are going to see it as a chance to increase their profit margin by charging us the same amount of money.

thats the way I see it, at least.

I think the wrong question is being asked

[PierceLabs]

The thing that I'd like to know is what motivated intel to increase the length of the pipeline. Intel generally doesn't go off half-cocked when they change the core. There is usually a reason they do what they do irregardless of the implementation which may perform poorly.

sick and tired

[johnnorthwood]

I'm very sick and tired of changing my name each time Intel decides that they need a new processor...

SSE3 info [Re:Does anyone know]

[Anonymous Coward]

The SSE3 moniker is new; it was Prescott New Instructions until very recently. Check out the Prescott New Instructions Software Developer's Guide.

Re:SSE3 info [Re:Does anyone know]

[Anonymous Coward]

http://www.intel.com/cd/ids/developer/asmo-na/eng/ microprocessors/ia32/pentium4/optimization/43988.h tm

Re: NUMA

[nicolasf]

Non Uniform Memory Access