Slashdot Mirror
2Ghz P4 Shown Off
mduell writes "Intel showed off their newest, fastest chip ever. The Pentium 4, running at 2 Ghz uses 400MHz Rambus Direct RAM(ugh). They also demo'd an Itanium server cluster running Linux with failover protection (what does this have to do with the chip?). Additionally, a 1Ghz P3-Xeon and a new 500Mhz mobile P3 that uses just 850 milliwatts when running most applications (5.5W max) were shown."
I am so sick of waiting for a dual-AMD mobo. It has literally been two years now since I heard the rumours about "oooh, there will be a dual K6-3 mobo by spring". Why, pray tell, does the friggin' CELERON have a dual CPU board, (Abit BP-6) but the Athlon and Duron and K6-2/3 are singular? Supposedly the Athlon has parallel logic built in. I know I'd be buying more AMD chips if they had good dual processor support.
Free music from Jack Merlot.
Why doesn't annyone recon this.
Only gamers need such power..but..there havent been worthwile new game concepts since the intro of q2 multiplayer , all q2 engine based games perform ok on current 800mhz cpu's wuth decent 3d accelerator.
- --[... The secret of the hanged man, the smile on his lips... ]-- -
If you read the article, you will notice that while they are shipping Xeon chips at 1 GHz now, they are still unsure as to a time frame on 1 GHz P-IIIs. And this despite the recent "announcement" of 1.13 GHz p-III's. How can you "announce" and "release" a product when you can't even buy the previous generation yet?
Despite quantity shipments of 1 GHz Athlons and Thunderbirds, there is no real way to get a 1 GHz P-III. That makes all of this just another set of smoke and mirrors - Intel takes a few high quality pre-production chips and cranks them up for a demo. Then they ship a very limited quantity of 1 GHz server chips - of course, server chips are better cooled and maintained, are much more expensive, and are ordered in much lower quantities.
So Intel has still failed to answer the real question at hand - can they actually ship a 1 GHz chip for the desktop? Can they capitalize on their market entrenchment, product quality, and technical expertise (all of which are vast, no matter your position) Or have they put too much junk in the trunk, spent too much time optimizing an overloaded, antiquated core, and lost too much technical drive to overcome the AMD challenge? Because right now, these "announcements" and "demos" sound like the last gasp of a dying dinosaur and not sound development from the once-undisputed king of the PC chip world.
Well, let me answer that for you: 'why not?' and 'you do.'
Progress and innovation (remember that word everyone?) is not made by producing more of the same crap but by always pushing boundaries. A chip of that speed almost certainly means new tech, and those, while initially expensive, will filter down to the common masses to we can all enjoy it.
So stop whining already!
"Hot lesbian witches! It's fucking genius!"
now toasters and ovens can have Web access, and don't even need the old inefficient heating elements...
How long will it take for M$ to bloat Windows so that you need one of these to run it?
I don't need a 2Ghz chip for anything I can think of. Why are people going to shell out large amounts of cash for these?
They need faster ships
...so we can have faster chips...
Hey - whaddya know - it rhymes!
(The mad poet strikes again! bwuahahahaha!)
(Spudley Strikes Again!)
To which I simply reply "dumbass!!!". But what I truly mean with this is that this will probably drive Intel's stock price up simply because of the people I described above being stupid enough to buy into it.
Yhcrana
The voices in my head don't like you
Ya know, these things run Linux too. Think about it ;) *g*
How about something more strenuous, like a BSP compile job under Q3Radiant using -vis -light -extra -threads? I have the perfect level for that; it takes up almost all 8192 units (that's about 1/5 of a mile in real world standards), and right now, without a lightmap, it's 400K. It would take a hell of a long time to compile this baby, hehe.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
If I want I can order a Athlon 1 GHz today, but getting anything faster than a P3 800 is a real big problem.
:-)
It's time for Intel to get things sorted out, and start real mass production of the faster chips, since AMD is really winning at the moment (not that I regret that
Every expression is true, for a given value of 'true'
If a 20 Stage Pipeline was a good move is to be seen. But the design takes the long latencies coming with a pipeline stall into account and tries to battle it at every front. This are better Branch-Prediction, ALUs working at double CPU core frequency and the Trace-Cache. since this is the first chip implementing a Trace-Cache i'm very interrested how this new cache model will influence performance.
To see how the new chip perform we will have to wait for neutral benchmarks. Perhaps it will not beat the Athlon clock by clock, but it will start with 1.5 GHz und will scale well beyond 2 Ghz this will make it the performance leader for some time.
About the floating point performance. IMHO Intel stopped beating the old x86 stack based FPU model to death and is walking along the way of SSE2. With a good optimizing compiler this will be pretty competitive. We can only hope Intel helps to get gcc to a point where it can optimize for the SSE Instructions as well as the Intel compilers.
thomas
I wondered how long it would take the Intel engineers to work their way through the DEC purchases they made and start using that technology in other areas. Given that a 200MHz StrongARM processor maxes out power consumption way below 1W (I have a feeling the figure is around 700mW) the power consumption of the Pentium processors looks pretty silly. Still there is no easy way to go from a streamlined low power consumption RISC design like the StrongARM and plunk all that technology into the Pentium line which requires a whole lot more transistors.
What I do take issue with is this 850mW figure for a 500MHz PIII. Intel's low power consumption tricks up till now have involved idling the processor when there isn't much happening, and I strongly suspect that this 850mW figure has a lot of idling in its measurement time frame. That figure of 5.5W max looks far more likely to really reflect the power consumption of the low power PIII. That is not to say that having a processor having various power consumption modes is a bad thing - the Amulet project has a more interesting take on this one (variable asynchronous clock speeds) - but I do wish that Intel would be more 'honest' with its figures. As for the rest of the announcements, I just request that you don't hold your breath waiting for these to appear on the shelves.
Cheers,
Toby Haynes
Anything I post is strictly my own thoughts and doesn't necessarily have anything to do with the opinions of IBM.
Folks,
From what we know about the Intel Pentium 4, it appears that the CPU is not optimized for something like Windows 95/98/ME, let alone desktop versions of Linux! It's better-suited for things that use large data sets, things such as large image files, large CAD/CAM drawings, and large databases, something more in the Windows 2000 or Linux server edition category.
I think people who will use Windows 98 SE, Windows ME and Linux desktop distributions will be far better off using the Celeron, Pentium III, Athlon and Duron CPU's.
It'll be interesting to see what the "Mustang" variant of the Athlon with its larger on-die L2 cache will do; if it is just a standard "Thunderbird" CPU but with a bigger L2 on-die cache it could become a great CPU for server machines (and will probably have the same pricing as the Pentium 4).
Raymond in Mountain View, CA
So now I'll be able to cook and eat my dinner without moving from the desk. Just hope it doesn't cook my nuts too.
I remember watching the first press conference from Transmeta (where Linus and Dave Taylor played Quake) and the "benchmarks" they were spouting. The Slashdot response was pretty evenly divided between "I want one tomorrow" and "what was up with those bullshit benchmarks?".
-B
If I understood correctly from posts (I couldn't read it, netscape keeps requiring kill -9 when it gets there.) It uses a real fast core, and the same 133MHz bus. This would us a HUGE multiplier.
RDRAM, is not the only one for the P4 (see the earlier discussion on P4s). SDRAM still lives in Intel products. RDRAM may have a little performance gain over SDRAM, but is it worth the cost? Judgeing from Intel's inclusion of SDRAM, I would say no.
32-bit. My 486 from 1989 or so is 32-bit. Why??? 64-bit Alphas have been around since 486s. When is IA-64 really comming out?
Massive MHz (or in this case GHz), but how does it compare with a P3 or Athlon in perfomance per MHz? or for that matter, an Alpha.
Low watts. Great, but does it have the reduced clock speed when not plugged in to a wall outlet.
P3s Athlons, etc. have been RISC, but they translate the instructions in hardware, requiring lots of extra transistors, and making them run hotter.
When is this "due" out. I am guessing that it is only a test processor, and not a final (release) processor. So when is it due out, and will it be like the 800MHz+ P3's avalability or IA-64's, due in 1999, release.
As I couldn't see the whole article, feel free to correct me.
I like announcements like this. It means that faster processors are on the way, which will drop the price of the current crop. Within 6 mos I hope to see the price of P3@800 become low enough so that I can afford to replace the two P2@450 CPUs in this machine without breaking the bank. Perhaps there will be a BIOS update for my motherboard by then that will allow me to go a little faster. I'm gettting rather tired of sitting around waiting for MSVC6 and SQL Server to do their thing (although parallel builds would be nice under Winders).
Yeah I appreciate the Xeon is more than a souped up celeron, but it's also less than a scaled down ultrasparc.
I was more taking the piss of the way that these marketing people latch onto numbers like those and tout them as key sales points when in fact they are meaningless.
This new xeon after all has half the cache of a pentium ii....?
Okay, here is the deal. If Intel can pull of the manufacturing of this thing, the 2GHz chip will not be far off. Meanwhile, I doubt it is possible for the .18 micron Athlon to be pushed up to 2GHz. If Intel can maintain this huge clock speed they've got two major advantages.
1) Their parts perform about as well as a much lower clocked Athlon for most tasks. However, give it something really regular like 3D, and it totally blows the Athlon away. Intel has gotten wise to the fact that nobody really uses consumer chips for anything other than 3D. Even the most bloated of Office apps don't demand much more than a 500MHz chip. However, get into anything 3D or media related (stuff that is pretty regular, but very compute intensive) then procs 1GHz+ are required. By performing about the same for most tasks, and totally blowing Athlon away in media, Intel hopes to get back their market share. This also explains why Intel is targetting this chip only at consumers (no SMP, the rumblings about using SDRAM) because the chip really wouldn't be ideal in a server situation.
B) Intel has the clock-speed advantage in terms of marketing. Like it or not, a huge number of people by their CPU for the clock-speed. In the market, a 1.4GHz Athlon vs. a 2GHz P4 at the same price will be a no-brainer for most people.
A deep unwavering belief is a sure sign you're missing something...
Hello? Were you reading? This thing has a 400MHz dual channel RDRAM bus. That's 3.2GB/sec, which is about 4 times faster than the Athlon bus. (which runs 100MHz SDRAM)
A deep unwavering belief is a sure sign you're missing something...
The NVIDIA Vanta chips are probably best here. You can get a 8MB version for about $40 on pricewatch, or a 16MB version for about $50. Around $55-60 you can get a Matrox G200.
A deep unwavering belief is a sure sign you're missing something...
I find it quite ironic to show a 2Ghz CPU when you can't even supply our 1Ghz. Has anyone seen a 1Ghz CPU from Intel somewhere lately? They seem to have all vanished from the price lists.
2Ghz is just hype, one more attempt to show people they are ahead in the clock speed race which nobody still follows except them.
"If liberty means anything at all, it means the right to tell people what they do not want to hear"
As with most of the recent Intel announcements, surely the most important question is "when are speeds like this going to be available in quantity?"
Without a decent answer to that, all of these announcements look more like an attempt to create FUD to work against AMD than like genuine advancements in the field.
Why do things like this even get posted to /. This article is just filled with inflated marketing hype.
... peer-to-peer networking
The 1GHz Xeon chip offers 256KB of Level 2 cache and a 133MHz bus, he said.
Wow so it's got twice the cache, a little over twice the clockspeed and a slighly higher bus speed than my OC'd celeron300 that I bought for pennies over a year ago.
The future is
Obligatory napster reference (dont flame me I do know what they really mean)
A 1.5GHz Pentium 4 system was then tested against an 800MHz Pentium III system in video capture. The 1.5GHz Pentium 4 was able to capture more frames of video than the 800MHz Pentium III
Wow good test. Curious how they dont mention any figures or how the difference in bus speed might affect the video capture performance. I HIGHLY doubt that the 1.5ghz machine was over twice as fast.
"Pentium 4 will be the fastest desktop processor in the world"
When it ships maybe, but when it actually hits the streets AMD should already be there. Intel seems to ship things an awful long time before you can actually buy them. They need faster ships
Not to imply Intel had rigged this one, but a single demo of a single P4@2GHz doesn't mean we'll that chip at that speed for sale any time soon.
One easy way in which Intel could have boosted performance numbers: Fab a P4 at 0.13 microns.
0.13 micron fabbing in quantity won't be around for a while, but IBM will, for a price, give you small runs on their X-ray lithography rig at 0.12 and below (as of years ago; it may be even finer now).
Intel may also have an experimental 0.13 fab line for fine-tuning processes before launch. 0.13 should be available in quantity around Christmas or so if I'm getting Moore's law right.
With either of these approaches, Intel would have to do custom tweaking of design parameters for the target process, but it might be worth the effort if it provides a 2 GHz demo chip.
Or, their uber-pipelined chip really _may_ run that fast in an aggressive cooling rig. See elsewhere for the short/long pipeline debate.
Sorry people but Intel is pulling a fast one here. What they have done is allow some part of the CPU's core to run at 2x the rest of the chip. If they were to do this to the PIII and find one which would run at 1GHz, then couldn't they say they had a 2GHz CPU?
Everything I've read on the chip disagrees with this. True, the ALUs on a P4 are running at 2x clock, but Intel hasn't yet made the mistake of marketing the part based on ALU speed. So, in the 2GHz demo, the ALUs were actually running at 4GHz.The AnandTech Editorial posted here a few days ago covers this.
For a more pessimistic view, check out this journal paper.
--
{Intel announcement template}
Today DD/MM/YY (date of last announcement plus one month) we announce the introduction of the Pentium X(increment the last number by one)that runs at XXXXGhz(increase Ghz 15%). This processor is the most advanced one that exists! It be available to the general public at an affordable cost in 36 to 48 months (maybe). At that time we will have released the Pentium XX that runs at XXThz (terra), which will be the most advanced processor ever built.
{/announcement}
Technology is only a vehicle. People are the ones that drive it.
Not to imply Intel had rigged this one, but a single demo of a single P4@2GHz doesn't mean we'll that chip at that speed for sale any time soon.
The 2GHz part was a handpicked chip, cooled like hell, and is far from being available.
The 850 mW number is measured "the Intel way", and therefore some considerable spindoctoring is involved.
Of course one can buy into the Intel marketing, but I prefer to spare my enthusiasm until I see that stuff for real, in volume, and tested by independent and reliable publications.
AMD was first with the 1GHz processor. Intel retaliates with a 2GHz. I'm betting that AMD will announce a 3GHz next week.
Expect world temperature to sharply increase by next year.
Yes, you're right I should have looked more closely at the specs.
I've been hearing things about the P4 for some time now, the 20 stage pipeline being one of them. I wasn't aware that they were doing significant things to compensate for the problems such a long pipleine introduces.
The other thing I heard is that its floating point performance is really bad. Now it may run some new style SSE2 floating point instructions at a decent clip, but how is that not a mere attempt at locking developers into using only those (patented I'm sure) instructions? Carrot and stick.
Not that I'm an intel hater. If intel makes the better chip then that is the chip I'm going to buy. Same goes for AMD or any other vendor. But based on their corporate culture and behaviour, I wouldn't be suprised if this chip is either a dog, or broken in some way designed to "lock in" users or developers. It wouldn't be the first time.
Lee
Muslim community leaders warn of backlash from tomorrow morning's terrorist attack.
I think that they did it to show off cool stuff running on their chip, not necessarily because they didn't trust the reliability of their chips.
And, BTW, if you weren't speaking seriously, please disregard this entire post.
---------
---------------- It is not a good idea to have a coffee drinking contest.
While this chip will run a 2Ghz, just how many instructions is it executing during each of those clock cycles?
With a 20 stage pipeline... not as many as a P3 or Athlon.
Intel designed this chip for very high clock rates with the assumption that Mhz ratings sell chips and systems because joe public is too stupid to know what IPC means. Sadly they may be right. Long gone are the days when the average computer shopper even knew how to use his or her system, let alone what went on under the hood.
Also, have you heard about how abysmal the floating point performance on it is supposed to be?
Hello Cyrix!
Lee
Muslim community leaders warn of backlash from tomorrow morning's terrorist attack.
The internal core is faster
The Bus is the same
The Ram is Rambus
The HD is still not really faster
The chip is STILL 32 bit. (Even my game console does better)
The intel pentium chip is a 78 firebird that is falling apart
Damn Pentium chips.. but we keep buying the stupid things. I don't think they are ever going to get the message that the current CISC/32bit archetecture is old and dead. *sigh*
--------------------