Slashdot Mirror
What 1.7Ghz Is Like
Beanie writes: "Today Intel announced their 1.7GHz Pentium 4. It's crazy to think about the fact that just one year ago we were breaking the 1GHz barrier and now we're almost up to 2GHz. AnandTech has a review of the Pentium 4 1.7GHz and they compare it to the recently released AMD Athlon 1.33GHz." And Otis_INF writes "Tweakers.net had the oppertunity to run some benchmarks on a system with an Intel Foster CPU on board, placed on an early i860 based board. The complete sneak preview (in english) is here. It smokes the P4 in some benchmarks."
The i960 was out a fair number of years before the i860. The i860 was a floating point monster for it's day (and the first superscaler I ever used). Oki tried to sell a line of workstations based on it. In fact I think they sold well to NTT, but they didn't sell well in the USA. The only embedded use I know of was in SGI's Reality Engine as geometry engines.
I had two in my office for about a year.
What a stupid name to use for their new motherboard chipset. They had previously used the i860 name for a series of CPUs (which are not compatible with the x86 series). The i860 was widely used for embedded applications. It's successor, the i960, is still available for Intelligent I/O (I2O) usage. Searching for i860 on Intel's web site is really going to be confusing.
Software sucks. Open Source sucks less.
I think the problem is that even the current kernel for Linux (2.4.3) and the upcoming Windows XP are not going to take advantage of the longer pipelines and SSE2 instructions on the Pentium 4 out of the box just yet.
It's only a few high-end applications and high-end games that will use the power of the Pentium 4, and even those are very uncommon nowadays.
I think people are going to realize that when AMD goes to the Palomino Athlon core, there still will be no advantage to going to Pentium 4.
Raymond in Mountain View, CA
There's Moore's Law as originally stated, and then there's Moore's Law, the marketing strategy of planned obsolesce.
--
Business. Numbers. Money. People. Computer World.
I'm not being zealous for any particular chipset or processor. Trying to attack my argument on grounds of being "typical and transparent" is ludicrous. How about some factual basis why my argument isn't sound. You'd be hard pressed to come up with a decent argument due to your apparently infantile reasoning ability. The K6-3 was a major blunder for AMD because they lost money on its production. They had a low production volume and the on die L2 cache only added to the cost in silicon. Three cache layers the fastest of which runs at the speed of the memory clock? Thats not what I call a great idea. The K6-3's performance was NOT on par with that of a P3 in any sort of test you could run. They also timed the release of the K6-3 badly compared to the release of the Athlon. The K6-3 was a competitor for the P2 and a poor one at that which was too little too late. The Athlon however was released concurrently with the P3 to act as a direct competitor for it. My friend's Athlon 500 tops my P3 500 (Katmai) in 3DMark by a couple points with really similar hardware which made the Athlon a nice choice for people interested in the most bang for their buck. I think the P4 is almost as much of a mistake as the K6-3 was, it costs alot and performance per clock isnt on par with competing chips. Your comment is all I expect anymore from the ten year olds frequenting slashdot nowadays.
I'm a loner Dottie, a Rebel.
On a related note, AMD is testing a new silicon that is said to help with the heat issues, which will help their own 1.7 GHz chip.
Jethro
Quidquid latine dictum sit, altum viditur.
Windows NT Historical Timeline
July 1989 - The first bits of NT run for the first time on a system built by the NT team using the Intel i860 processor.
January 2, 1990 - Bill Gates brings together NT's top designers to discuss the importance of running NT on Intel's 386+ processors and to choose a new RISC processor other than the Intel i860.
cpeterso
> The only ones that are applicable for me are a) constant computing (folding@home) and b) gaming; both areas where the P4 excels.
Actually, several of the review sites are showing that the AMD beats the Intel in about half the gaming tests.
And that at about half the price, both for processor and for memory, based on what's currently listed on pricewatch.com.
I'm also curious about folding@home. Does it run through a lot of memory? The review sites left me with the impression that the P4 gets its advantage -- when it does get one -- from the bandwidth performance of its horribly expensive RDRAM.
If folding doesn't use a lot of memory, then the Athlon might actually win. (Either way, it probably wins at performance/price by a large margin.)
If you happen to have access to both kinds of machine, I'm sure lots of people here would be interested in a folding benchmark, though.
--
Sheesh, evil *and* a jerk. -- Jade
I have a G4 466 (OSX), a Dual CPU Intel 800, a Origin 3000 class server (Work of course)
I would say the only machines that don't perform in relation to the mhz rating is the SGI Origin 3000 and the Sony Viao. The SGI is clearly much much much faster than my PIII w/equiv total equiv mhz (~1600)... and the Sony Viao laptop is much slower than my Apple G4 466.
But, my Athlon 700 kicks the crap out of my Sun Blade workstation at just about everything I give it. My Dual PIII workstation absolutely blows away my G4 Workstation running OSX at just about everything I do. (Fair enough, my Dual 800 is SCSI, my G4 is not)... and funny enough, my G4 seems to be just a hair slower than my 500 mhz Sun blade workstation.
Of course there are no numbers or hard benchmarks. these are just my personal observations on equipment that my company and myself owns (it's my company)
But again, without this myth alive how would I be able to justify my 2-3x pricetag in buying commercial Risk hardware over Intel?
Anyway, thats my take..
geez.. and what a rant.
--------------------
Would you like a Python based alternative to PHP/ASP/JSP?
Actually, there were working computers at the time. Stratus, a vendor of ultra-expensive fault-tolerant servers, has used m68k, i860, and PA-RISC processors in successive versions of their servers. The old Stratus XARs were RISC-based, and I believe that they were out in that timeframe. (It was really funny to have a machine running an i860 chip next to another machine with a i960-based network card.)
Of course, these machines would've been completely unsuitable for development of NT, but they were there.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
The tweakers.net article had it wrong. The Pentium 4 uses the i850 chipset, according to links I followed starting at Intel's front page, then to the Pentium 4 page, and finally to the chipset page.
The i860 and its successors were interesting chips. As I mentioned in a reply to a post underneath your original post, at the last company I worked for, we had old servers running an i860 as the CPU sitting next to modern servers with the much, much faster i960 chip running their network cards.
Of course, it was even funnier that my roommate at the time was still using a i386 PC with 8 Megs of RAM. Every day he worked with a network card that had much more processing horsepower and RAM than his PC! I used to tease him about that all the time.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
The 1GHz of a year ago was a PIII which was comparable to an AMD Athlon. Today's "almost 2GHz" P4 comes in SLOWER than a 1.33GHz Athlon in most benchmarks, so it's hardly double the speed.
The P4 may be a good chip one day, but it sure isn't there yet.
I've talked to someone who ran the Windows XP beta, and he says that it's deathly slow. Soon enough, desktop PCs will need more speed, thanks to Microsoft.
Of course, I won't be running XP, but with KDE/GNOME, I can enjoy the same "benefits" (speed-wise).
Window Maker.
------
Just to clarify: the instruction set hasn't changed for the Pentium 4. Its still the same tired old x86 CISC ISA from 1980, with a few new added instructions (SSE2) that are of little use to most applications (and users).
Spyky
Did anyone else do a doubletake on this? Or did I just miss a motherboard chipset in the 8xx series that happened to have the same number as Intel's old 32-bit mil-spec CPU line?
You may have missed it because it's not out yet--the i860 (codename "Colusa") is the chipset for the upcoming 1-2 proc P4 Xeon (codename "Foster"; actually, in an effort to be hugely confusing, the official name is now simply "Xeon") systems which should be released in a couple months or so. FYI, the 4+ proc chipset will be designated the i870.
And yes, it is a bit odd that Intel is recycling this rather inauspicious brand number. I suppose not many in the industry have a long memory.
(For more info on the first i860, Paul DeMone had an interesting article at RWT comparing its ambitious but flawed design to Itanium and its potential pitfalls.)
There used to be an error page showing some cool cars (so you had something to look at when they fixed the servers), it used to show when things really got fscked, don't know if it's still in use.
---
How soon until manufacturers need to start looking at via effects, trace-to-trace coupling, board materials, trace loss, etc more closely? Things are starting to get pretty fast. There is a point when standard PCB manufacturing techniques go out the window.
But, you are dead on with the practical point, who cares? As it currently stands, the P4 is a complete waste of money. For the PC, nobody needs a 1.7Ghz chip. For a server, that clock speed would be handy, but only with about 1Mb of onboard cache.
Someone you trust is one of us.
The real story today is not the P4, but the prices. Intel is slashing prices big time, ahead of their .13 micro manufacturing process which wont be operational until the end of this year. Basically, they are starting a price war with AMD, and it looks like it will be vicious. Why? PC Manufactures can read, and the verdict on the P4's real performance frankly no good. The P4 has a long way to go before it can be considered an improvement. Of course, consumers are idiots and they buy CPU's based on clock speed alone. However, the PC market is hosed right now. By the time the PC market recovers, AMD will be there with its next gen chips. This price war is something that Intel can afford. I wonder if AMD can afford it? AMD's manufacturing costs have always been more competative than Intels. However, a 50% price reduction has to sting, and AMD wont have .13 micron technology by the end of this year.
Someone you trust is one of us.
As blair1q commented, I also did a doubletake on the name. The i860 was a really kinky chip that did some things very fast, though it appeared to be too weird for most compilers to do a good job of letting C language tell you which part of the processor to run your stuff on.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
The 64 bit arena is where the real fortunes will be made and lost. Hope AMD doesn't misplay their hand there.
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
The irrational part of my mind cries, "Look! 1.7 GHz! That's almost 2 GHz! Blazing speed! Raw, brute, merciless POWER! More!! MORE!!! I'm still not satisfied!!!"
I can hold off the irrational only so long. When that magicical hertz hits two gigs, the irrational is going to sneak up behind my rational mind with an icepick.
"Everything louder than everything else!" -- Meatloaf
And let me guess - you didn't even bother to read the article, did you? Because this issue is specifically addressed. Here's the relevant section:
During our tests the Pentium 4 1.7GHz always operated at 1.7GHz and did not fall victim to any clock throttling because of heat. You shouldn't worry about the Pentium 4 dropping its clock speed because of heat unless you are running the processor without a heatsink/fan installed.
Actually, to be completely anal, it states the number of transistors on an integrated circuit will double approximately every 18 months.
"Prefiero morir de pie que vivir siempre arrodillado!"
Freudian struggle? You're more right than you could possibly realize :) The original posters quote was from Tom Lehrer's "Smut:"
More, more, I'm still not satisfied!
Stories of tortures
Used by debauchers
Lurid, licentious and vile
Make me smile.
Novels that pander
To my taste for candor
Give me a pleasure sublime.
Let's face it I love slime!
1.7GHz is probably not much different than 1GHz, or even 500MHz to the average user. The only real difference the average user will notice is when they get the bill. And of course there are bragging rights.
As I, as well as the linked articles, pointed out, initial optimizations using SSE2 are not promising. In fact, even applications that have received a good deal of hand-tweaking to use all of the latest instructions run slower than the same app, sans SSE2 of course, on the Athlon 1.33GHz. And the gap would widen even more if, instead of wasting their time with SSE2, they just hand-optimized the program to run faster on the Athlon.
First off, I read a good portion of the reviews that I found linked from Blue's News:
Source Magazine
Target PC
Hardware Unlimited
Tech Report
Gamer's Depot
What's the upshot? That even with each processor's "ideal" system (DDR on the Athlon, RAMBUS on the P4)-- well, the P4 kicks ass at Quake 3: Team Arena. I mean, it's really really good at Quake 3. So good, in fact, that-- well, you won't be running anything else, I hope?
Because in almost every other app, the cheaper Athlon 1.2 equals or outperforms the P4. That even includes apps such as POVRay that did some early optimizations for the P4's extended instructions. I recommend reading the Tech Report's overview if you're interested in that; they have more details on exactly which instructions were used, and the current state of Intel's compilers for the chip.
Keep in mind, of course, that the compilers are still a bit beta-ish-- sometimes they actually make the programs run slower. But they never appeared to actually make it faster than an Athlon 1.2.
Debate what you will about future extensibility, and so on-- but unless you're going to be playing a whole lot of Quake, if you're looking for a new system you should grab one of those cheap Athlon CPU/Motherboard combos selling for $300 at Fry's.
Why are we going faster? Especially Intel? They need to concentrate on making a better quality chip, not a faster one.
Remember that company that made inexpensive chips? They thought about making a new chip with a better FPU (among other features). It didn't need to be faster than the competition, just a better quality chip. They called it the Athlon.
What's funny is that Intel thought they could compete with this better quality chip by making a faster chip, which they released to early, and had to recall.
Guess some people don't learn their lesson...
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
New /. poll:
How long until the new 1.7GHz gets recalled
- 1 month
- 1 week
- 3 days
- When cowboy neal gets one
- its already recalled.
I still find it humorous that they compare the 1.7 Intel to the 1.33 AMD...
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
I will tell you what 1.7 Ghz is like... since this came from Intel it is most likely like a 1.3 Ghz Thunderbird... =)
True, but people have been applying it to raw speed for a couple decades now, and the law has applied equally well in that manner. (I'm feeling too lazy to look it up at the moment, but I seem to recall that the notorious "Jargon File" updated their definition of Moore's Law to reflect the alternate aplication of it.)
CPU's have been close to doubling every year and a half for quite some time now, so it should not be shocking to anybody that we have gone from 1 to 1.7 in a year. That was my point.
Information wants to be anthropomorphized.
Since that time, a trend has emerged of chip speeds increasing at a rate of about 2x every year and a half. Since it has held up (for far longer that Moore could have expected, and applied in ways that he didn't mean for it to be) for so long, we jokingly refer to it as a physical Law of nature, named for the guy who first said it.
Some people would cite this as an example of a "self fulfilling prophesy" (everybody agrees that "Moore's Law" is a reasonable expectation, so that's what everybody shoots for to keep up with the competition). Whether that's true or not, I'll let others speculate about. "What if" debates give me a headache.
Information wants to be anthropomorphized.
OS X is a multitasking operating system. The old Mac OS was not. That is why the old Mac OS was able to be so fast and responsive, even on really old hard ware. Drag the mouse, and that's all your CPU was thinking about: dragging the mouse. Open an application, and you were forced to twiddle your thumbs while the application launches, but 100% of that machine was thinking about launching your application.
For years, Apple users have been screaming that they wanted preemptive multi-tasking, even though almost all Macs are used as single-client machines. Well, now you got it. Apple runs Apache web server nearly as fast as any other UNIX, but guess what, there's a price for all that power: doing several things at once takes more effort than doing one thing at a time.
Personally, I don't give a shit if IE or Omniweb takes an extra 10 seconds to launch, because now I can do other things while it is launching, instead of having my system be effectively dead to me while it loads an app from the hard drive.
But if reizing windows quickly is more important to you than having lots of background processes while you work, then boot to OS 9.1 and stay there. You won't be able to run UNIX apps, and your box won't be much good as a server, but at least you can launch your web browser quickly, which seems to be what really matters to you. I'm not just brushing you off here, I am serious. It really sounds like OS X is not for you at this stage in its development.
Information wants to be anthropomorphized.
Mmmm... burgers.
Information wants to be anthropomorphized.
Was progress at the speed of Moore's Law always crazy, or did it just become do today?
Information wants to be anthropomorphized.
1.7GHz? That's really pushing the speed limit, I wonder how often it has to cut down to 850MHz because it gets too hot.
Education is a better safeguard of liberty than a standing army.
Edward Everett (1794 - 1865)
moore's law is not tied to clock speed, but rather transistor count.
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
you don't find it the least bit convenient that there were enough 486's with defective FPUs?
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
well, quite bluntly, intel has fucked us over time and time again... let's go for a walk down 'intel memory lane' for a second...
what's the difference between a 486DX processor and 487 co-processor? the pin arrangement! that's right, intel realized it would make more money by selling a 486DX as a '487 math coprocessor' even though that by installing it, you disable your 486SX. (effectively making your '487 math coprocessor' the main processor)
what's the difference between a 486DX and a 486SX? intel intentionally fucked up the FPU on the 486SX! yeah, on the original 486SX's, there WAS an FPU, but it was disabled. thanks intel!
difference between the older celerons and their pentium 2 brothers? nothing! well, except for the fact that intel broke the l2 cache on them so they could sell value chips. the cache was THERE, you just couldn't use it. if you've ever opened a PII cartridge, you'll notice that it's a socket 370 chip on a slocket. surprise...
the pentium series bugs. F00F!
intel has driven me out of my mind. i'm an amd convert till they start fucking it up.
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
"I hope I don't make a mistake and manage to remain a virgin." - Britney Spears
There is no such thing as clock limiting circuitry.
The chip derives its clock from a frequency provided by the motherboard. There is no way it can "know" how fast it is running unless there is some sort of frequency generator in the processor itself, which is fraught with many problems, and will probably never happen.
-
I've had enough abrasive sigs. Kittens are cute and fuzzy.
Well, since you seem to know so much about it, and how I am wrong, why don't you enlighten us?
If you can offer nothing but critisism without explaining what indeed is wrong with my facts, then you are just trolling.
-
I've had enough abrasive sigs. Kittens are cute and fuzzy.
of id vs ego (and perhaps a dash of superego as well)
Where the id says "More! Grunt! More!"
The ego says "I am satisfied. All's right with the world."
And the Superego says "Consume. Spend. Buy. Just do it!"
You know what? Mac people have had to rationalize for far too long. We've had to settle for 400Mhz!
Argh! When will someone come to quench our thirst for raw power? Motorola? IBM? Apple?
Do not bother to fight the irrational/id. The best you can do is placate it and compromise; promise it a *dual* 2Ghz system!
Geek dating!
GPL Deconstructed
Yeah, it's just crazy to think that it's almost like every year to eighteen months processor power doubles. Someone should theorise a rule about it or something!
You knew it would happen: Kyle at [H]ard|OCP got his hands on one of these and overclocked it. And just look at the results.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
Fortunately, most college computer labs are filled with virgins that can be used for just this purpose :)
And of course, RISC chips like the G3, G4, etc. do far more per clock cycle than their Intel/AMD counterparts. At this point Intel's sort of saying "LOOK HOW FAST WE CAN MAKE THAT QUARTZ JIGGLE! OH YEAH!" Who cares, look at the practical side I guess.
MHz isn't really a milestone anymore because it's not very significant anymore.
--
--hongpong.com
I think the /. crowd is most pissed off by the fact that Intel's technical decisions are made partly with a view to marketing. The "Internet architecture" BS a couple years back Intel's own have engineers bitched there is no way to optimize a CPU for Internet - net bandwith will always be the limiting factor, not the CPU. But we still saw all those idiotic "Intel gives you power for the Internet" commercials. Now we've got a chip that appears to be grossly overpriced for its performance, released now only because of "MHz marketing". I think the P4 has an good architecture which scales like crazy, but why was it released before it reached 2 GHz and that scaleablity was an asset and not a liability? The UID for CPUs and didn't help Intel's image either - that pissed off the general public too.
That said, assuming the P4 continues to play catch-up well, I'll buy one if it offers more value for my money.
You said it! And it'll be hysteria-inducing when we go from almost 2 GHz to 4 in the next eighteen to twenty-four months!
Yowza! I'll believe it when I see it!
Obliteracy: Words with explosions
In response to this, Intel and AMD (and vendors of other computer components) have slashed their prices. In the short term, this will result in a boost to sales. In the long term, it's a disaster waiting to happen. Many users and businesses that would have waited to upgrade at higher prices will upgrade now due to the rock-bottom prices. As a result, the industry will see a lot of low profit sales in the near future. But what happens in a year or two? Will there be some compelling application that is going to compel the average user or business to splurge on a 4ghz system? I doubt it.
In the worst-case scenario, this might lead to the failure of either AMD or Intel. Look at what the price wars did to the hard drive industry already. Where is Micropolis, Quantum, or Conner? Western Digital, once the preeminent IDE drive manufacturer, had losses of $7 to $10 million for its most recent quater. In almost any case, it will be likely to result in higher prices and less competition. And that's probably not good for any of us.
Did anyone else do a doubletake on this? Or did I just miss a motherboard chipset in the 8xx series that happened to have the same number as Intel's old 32-bit mil-spec CPU line?
--Blair
In a recent press conference, Intel stated: "Not only is the new Pentium 4 a technological breakthrough in terms of processing performance, but users can cook 4 hamburgers in under 10 minutes on it's new larger-sized heatsink"
Slashdot: Open Source, Closed Minds.