Too hot to handle: Intel aims to launch the P4 3.6 GHz in mid-2003. For this exclusive report, Tom’s Hardware Guide tested CPUs of the future, bringing you benchmarks for P4s in the 3.6 GHz, 3.33 GHz and 3.06 GHz variations. The article also tests against the future AMD AthlonXP 3000+ and 3400+. Update: More information on Intel’s plans and technogies for the future.
Intel arguable makes great products, but there recent line of pumped up P4 show little to do improvements over the previous “slower” processor.
Is this just another example of that?
Just nitpicking. While the single processor ClawHammer would be called Athlon, it wouldn’t be Athlon XP (many rumuors said it would be Athlon 64, who knows?).
> show little to do improvements over the revious “slower” processor.
What do you mean?
Very interesting article. Thanks for posting it:-)
Referring to Page 4 “Comparison Of All P4 CPUs” . . . for the first time I see the term Northwood “A” core . . . well, I have BeOS running on P4 at 2.26 GHZ . . . so that must be the core I’m using.
Applying P4 patch seemed to make no difference for this core. BeOS runs great but does not have the processor “pegged” (i.e. when you click on “About BeOS” it can’t name the processor or tell you its running at 2.26 GHz).
Besides, the article merely mentions ClawHammer, not “tests against the future AMD AthlonXP 3000+ and 3400+.”
Well, enough nitpicking for today
suddenly my Athlon XP 2100+ sounds like something from the biblical times! :-S
Unfortunately this shows that ultimately, bigger pipelines, even with crappy branch prediction, allow an inferior processor to brute force its way to the top with higher clock speeds.
It’s apparent, however, that this requires enormous amounts of memory bandwidth in order to be effective, no doubt one of the reasons why they’re moving to a 667MHz QDR FSB while AMD seems content with 266MHz DDR.
> allow an inferior processor
I do not account the Pentium4 as an inferior processor. AS LONG as you have compiled your apps with SSE2, the P4 flies and has absolutely no match. Check the benchmarks for apps that were compiled for the P4 and you will see how fast it can be… http://www.tomshardware.com/cpu/02q3/020909/p4_3600-13.html
The same was for the PEntium Pro. It took years until people start optimizing their apps for i686 or for ppro. But when they started doing this, the CPU could really show its real teeth.
I wonder if macs will still be sitting at 1/1.25ghz by the time this is released:)
I bet these P4’s can brute-force their way through pretty much anything.
OS X on Intel, it’s time for Apple to “Think Different”.
Anyone else with me?
Ask me if I care, sorry I have to get this of my chest.
What am I going to use a 3.6Ghz P4 for? I only do trivial stuff like writing code/documents, surfing and playing some music. I suspect most people do stuff along these lines.
If we go to the extremes we play the occasional game, or encode video (perhaps the only area where a fast cpu really shines).
I must admit I am tired of the “Most frames per second” battle between Ati and Nvidia, and the most herz battle between AMD and Intel.
What would be much more interesting would low-foot print machines, low power consumption, fanless designs. All that stuff..
Just my humble opinion.
A cluster of dozen of these monsters can crack DES (unix passwd file) in a matter of hours!! And to think that in 1997 they needed 14,000 computers.
Other than that, the winter is comming to Israel and soon It will get cold. lucky for me, My roomate has a new AMD that will keep us both worm for the winter.
I agree sort of, but what is great is that all those people out there that want these things push the specs for ‘base’ processors up at the same time the prices go down. I mean it is amazing what you can get for $800 these days. If AMD and Intel weren’t pushing the envelope it would never happen.
Okay, we can agree on pushing the envelope.
But this incredibly rapid turnover of cpu’s only cause all other components to become outdated so rapidly that there is no keeping up.
A cpu upgrade these days normally means that you also have to upgrade memory and motherboard. Why don’t they just integrate all that on a single PCB (a little like Via Eden, should be very cheap), so that when the time comes you just yank out the mainboard a pop in a new one.
goop: Other than that, the winter is comming to Israel and soon It will get cold. lucky for me, My roomate has a new AMD that will keep us both worm for the winter.
Since when does Israel not have winters? It has winters, just that it doesn’t snow. It rains. The tempurature isn’t that low. Winter nontheless. That’s med weather. Best weather on earth. (Besides, I have a Duron here, not so hot. I also used Athlons and Athlon XPs, not hot.)
And Naish, no one is expecting you to keep up. Just buy every 3-5 years, you would be fairly up to date. There is no reason to upgrade whenever Intel and AMD releases something.
Ugh…I’m still waiting for my new XP 2000 (Got broken stuff, had to return it)
Why shouldn’t we upgrade to newer faster CPU’s? It is really needed for graphics and many people play games. And it’s fun (except for said incident)
if you are happy with what your current pc can do, then dont upgrade it. dont complain that technology evolves faster than you can figure a use out of it. plenty of people out there who can. if you want slow upgrades, purchase a mac. for me, normally a MB and memory lasts for around 3 cpu upgrades.
>A cpu upgrade these days normally means that you also have
>to upgrade memory and motherboard. Why don’t they just
>integrate all that on a single PCB (a little like Via
>Eden, should be very cheap), so that when the time comes
>you just yank out the mainboard a pop in a new one.
Because some people on some occasions can just simply swap the processor for an upgrade. Perhaps because with certain applications a memory upgrade can do more to increase performance than a processor upgrade. Finally, just because I buy Intel CPUs doesn’t mean I buy Intel motherboards, nor would I want to.
Unfortunately this shows that ultimately, bigger pipelines, even with crappy branch prediction, allow an inferior processor to brute force its way to the top with higher clock speeds.
That “inferior” processor was designed specifically to reach these high clock speeds, but you act as if Intel just got lazy and decided to remove the good stuff. Processors are not better just because they have a higher average IPC count; what matters is how they perform, not which design you happen to favor.
Incidentally, I favor AMD processors because they tend to have a lower price/performance ratio. Intel will probably continue to chip away at this lead until Hammer comes out. I have no idea why anyone would pay through the nose for a low clock speed 64-bit CPU; people seem to think that 64-bit CPU’s are naturally twice as fast as 32-bit CPU’s.
OS X on Intel, it’s time for Apple to “Think Different”. Anyone else with me?
How about, “Think Differently“? 🙂
OS X on Intel, it’s time for Apple to “Think Different”.
Anyone else with me? x86 Macs? Nope never Going to Happen, Now lets NERVER Mention this on Osnews Again! And soon the G5 (or that Power 4 Chip)will soon be out and blow everything else out of the water!
I’m glad AMD continued to push the envelope even after developers started optimizing for the early pentiums. The Intel philosophy has always been; dumb consumers + high mhz rating + premium price for their brand name = excellent profits. Since the k6 AMD has been a better value, and a better upgrade path. They don’t frequently change their chipsets to prevent drop in upgrades and they don’t partner with memory vendors trying to play the lawsuit game.
In related news, anyone checking out the XScale fiasco in the PocketPC world these days?
Athlon 1Ghz:
16:03:44 up 3 days, 39 min, 6 users, load average: 0.12, 0.04, 0.01
PIII 800:
15:57:12 up 7 days, 22:22, 3 users, load average: 0.00, 0.00, 0.00
useful
OS X on Intel, it’s time for Apple to “Think Different”.
Anyone else with me? x86 Macs? Nope never Going to Happen, Now lets NERVER Mention this on Osnews Again! And soon the G5 (or that Power 4 Chip)will soon be out and blow everything else out of the water!
>>>>>>>
Um, dude, lighten up. Besides, even Power4 can’t help Apple now. Take away the multiple cores and giant cache and gobs of bandwidth that the real Power4 has (which the desktop version will have to remove) and the Power4-desktop will be lunchmeat for these souped up x86 machines. Even a real Power4 is only 50% faster in floating point as a 2.5 GHz P4. The point of these chips isn’t individual speed, but rather speed in a multi-way machine. (Just as the point of the P4 isn’t speed via IPC, but speed via clockrate). Thus, putting it in a desktop config castrates it significantly. Now, a 2.4 GHz G5 might be another story, but I doubt we’ll see it before x86 chips reach 5 GHz+.
I mean, I think these super fast processors are great. You could drop a wok on them and stir fry a good meal, but who needs that for day to day work? The main resource I use on my machine is memory, but even with databases running, browsers running, code editing, mp3’s playing.. I don’t peg the CPU. I play the occasional game, but those are fine. For 99% of the users whose most complex game is The Sims, what is this for?
It seems like there is a push for “the fastest”, but that accounts to probably 0.5% of their sales.. Everybody else wants a reasonably cheap, reliable computer that does every day things. Get the computer smaller.. cooler.. more reliable.. and it will start to matter. Put in faster and hotter CPUS, drive up power consumption and heat generation and cooling requirements, and how does it really help the average user?
(And this shouldn’t be taken as.. Why do we need progress? But really.. compare this to hard disks. *Right now*, what does a user need 160GB on their desktop for? In 5 years, perhaps everybody will want full motion video.. but right now, what does this sort of thing get me? I guess the ability to store all of the MS security patches..)
The bigger market of intel is the windows environment… and there, the majority of the programs are closed source and thrus can’t be compiled…
So, it will take at least one year after the CPU reachs the streets to the common user to see the diference… as it happen with pentium pro (as you said…)…
Cheers…
I do. Try compiling heavily templated C++ code in gcc. You’ll immediately want to spring for the fastest proc availabe. Other uses include encoding MP3s and videos, 3D rendering, gaming, just running KDE, etc. A lot of these things (especially the compiling is busy-wait stuff. You can’t really go and do something else for the five or so minutes while a compile finishes. And you need the results of the compile to continue your work (debugging, etc). Thus, until these actions complete faster than human response times, its not fast enough!
*Right now*, what does a user need 160GB on their desktop for?
>>>
Dude. I’ve got a 30GB and I’m hurting. Game installs these days are multiple gigabytes, then you have your music collections (encoded at high bitrates, for optimal quality), your videos (a webcam is a great toy, btw!), and your Linux distro ISOs (okay, maybe that’s just me) and you can blow a hundred gig easy.
Heck, I’m pretty happy with my 1995 Sun SPARCstation 5/85. Well, except for the 8bpp framebuffer.
It’s fine for surfing the web, coding, listening to music, GimP-ing images, and so forth. You just have to be careful not to use overly bloated apps. Although processors (and systems) are a heck of a lot faster than they were five or ten years ago, the vast majority of the public has computers that *feel* just as slow as they did back then.
(And this shouldn’t be taken as.. Why do we need progress? But really.. compare this to hard disks. *Right now*, what does a user need 160GB on their desktop for? In 5 years, perhaps everybody will want full motion video.. but right now, what does this sort of thing get me? I guess the ability to store all of the MS security patches..)
My wife is a pretty average user. She ripping dialogues stored on tape into .aiff files and loading them into a database for ease of access. Uncompressed uses up a ton of space — and going .mp3 loses some quality for only a 50% space savings. A few years ago she was playing with uncompressed video and the home machine couldn’t handle it. If she had an IDE raid 7200 drives with an extra few hundred gigs she would have been able to work at home.
Anyway on the main topic of the pentium 4 I hope they start adding some more cache to these chips or you are just going to get 3 billion no-ops per second.
http://news.com.com/2100-1001-957194.html
For speed freaks, Otellini also demonstrated a 4.5GHz Pentium 4 processor onstage. He was able to rev the chip to 4.7GHz before the machine konked out and offered up a blue screen.
Keeping data safe
In the second half of 2003, Intel will introduce LaGrande, a security technology that prevents hackers or viruses from obtaining or corrupting data in a PC.
“This will minimize the ability of people to steal your credit card number or break into your hard drive to snoop,” Otellini said. “Protecting users’ data, protecting users’ identity, protecting transactions are all on the list of things we want to do.”
LaGrande places a secure wrapper around selected hard-drive data, as well as around the keyboard, the display and the interconeects inside the computer, said an Intel representative. Currently, data that gets sent to commerce sites is encrypted while traveling between a PC and a server. But once it’s back on a hard drive, it reverts to its original form, making it valuable if it can be stolen.
Conceptually, LaGrande is similar to IBM’s RapidRestore, a feature on IBM notebooks that lets users store applications and data behind a secure partition on the hard drive. The technology, though, will have other functions. It is possible to use it in conjunction with digital rights management programs, such as Microsoft’s Palladium, to prevent piracy, which in turn could help promote legal entertainment downloads.
Otellini said users will be able to turn LaGrande off. “It will be opt in,” he said.
LaGrande will make its initial appearance in Prescott, Intel’s next generation of desktop chips, coming in the second half of this year.
Although LaGrande sounds useful, Intel could have a difficult time getting support for it, especially from Microsoft.
“Just like Microsoft eventually integrated signal processing into Windows, they will eventually integrate security processing, and they will do it on their own time,” said Peter Glaskowsky, editor in chief of the Microprocessor Report, an industry newsletter. “Intel needs to work with these things with Microsoft from the beginning.”
Glaskowsky compared the effort to Native Signal Processing, a chip technology that Intel tried to popularize in the early ’90s. Microsoft was strongly opposed to the effort and cajoled other developers not to support it. Eventually, Intel dropped it.
“I call this native security processing,” said Glaskowsky.
I do not account the Pentium4 as an inferior processor. AS LONG as you have compiled your apps with SSE2, the P4 flies and has absolutely no match.
Think about this, a single mispredicted instruction in the Pentium 4 results in a minimum of 17 wasted cycles (thanks to its ungodly 20-stage pipeline)
I’m unable to find any difinitive statistics on the Pentium 4’s branch prediction accuracy (even in the whitepaper) Intel claims it to be 30% more accurate than the Pentium III, which had a branch prediction accuracy of 90%. This puts the Pentium 4’s branch predictor at about 94% accuracy.
The way the numbers work out, you can deduct the percentage of branch instructions in programs you’re executing from the Pentium 4’s raw speed, and that will give you its effective speed. Think about it this way, let’s assume that 17 cycles are wasted per misprediction. Now, of course your milage on this will vary depending on how many branch instructions are in the code you are executing. Checking a program I think to have relatively heavy branching (a parser) I find approximately 40% of the opcodes are branch instructions. This is a bit heavier than you’d probably see in day to day programs, but it serves to demonstrate the point.
So, in said program, in 250 instructions, about 100 of them will be branch instructions. Assuming 94% accuracy, that means that 6 of these 100 branch instructions will be mispredicted. In the worst case, with no overlap, this means that a total of 102 cycles will be wasted executing 250 instructions.
In other words, branch heavy code kills the Pentium 4. This is why Intel is constantly showing you wonderful benchmarks of its SSE2 performance in hopes that the horrible branch prediction penalty paid by the depth of the pipelines will go unnoticed.
That is a bit misleading. First an Athlon would have paid a price of 50 cycles or so on the same code. Now, the P4 gets 150 cycles of work done vs 200 for the Athlon. It needs 30 percent more clocks per second to make up for that which is generally the case. On code that is less branch heavy, it still has those extra clocks.
King Mac Of Apple Land: x86 Macs? Nope never Going to Happen, Now lets NERVER Mention this on Osnews Again! And soon the G5 (or that Power 4 Chip)will soon be out and blow everything else out of the water!
For how long? Motorola nor IBM has the economic scales to give them enough incentives to push the envelope. Plus, long term speaking, the low cost of x86 can really benefit Apple’s profit margins.
What really pisses me off is a bunch of close minded Mac users that can’t accept that Apple killed PPC, and can’t accept that x86 is the best choice, economically, in the long term. What do they rebutt logic with? “It would never happen”.
tapelgig: In related news, anyone checking out the XScale fiasco in the PocketPC world these days?
Everyone is waiting for PocketPC to be optimized for it. XScale wouldn’t give you extra performance if you are a plain ARM app. Meanwhile, a lot of ISVs already optimized their apps for XScale.
But heck, it doesn’t matter. XScale is already worth it, without performance increase: longer batery life, less heat emission…. 🙂
elemur: The main resource I use on my machine is memory, but even with databases running, browsers running, code editing, mp3’s playing..
You would see increase in speeds for code compilation, dude. (My 1GHz duron had a 10% increase in GCC speeds after being upgraded from a 800MHz duron).
elemur: Put in faster and hotter CPUS, drive up power consumption and heat generation and cooling requirements, and how does it really help the average user?
Only P4 increased their power consumption by a small amount. As for heat, I’m using Athlons for this case. Athlon XP (Palamino) was cooler than Athlon (Thunderbird). Athlon XP (Thoroughbred) was cooler than Athlon XP (Palamino). And for sure, Athlon XP (Barton) would be cooler than Thoroughbred, and ClawHammer cooler than Barton.
Private Citizen: For speed freaks, Otellini also demonstrated a 4.5GHz Pentium 4 processor onstage. He was able to rev the chip to 4.7GHz before the machine konked out and offered up a blue screen.
It is Windows fault, pre-SP3 Win2k and pre-SP1 WinXP, Windows can’t accept more tha 3GHz. Besides, there is an reason why it is still a prototype and not released yet.
SpyWarePlus.
No, hackers “can’t” get your credit card. Every “honest” company out there has direct access to even your hard drive.
Very very nice 🙂
well, at least they accept it.
LaGrande = “TheBigOne” if read in spanish…
Now, the Billion Question :
1984 = second half of 2003?
(read: Orwell)
Intel has joined the party and decided to kill the PC industry.
The idea of a PC will go on, though it will come from China. It may be built with VIA chips, with Chinese chips, with AMD chips. Or whatever is available without FritzInside.
Security kills the economy. This is well documented in history. Freedom and open systems are large economies. Closed and secure systems are small, stagnant, and stifled. There is no innovation and there is no technology economy.
Intel is just being stupid. Why did they even spend a dime on Linux if they are going to turn around and make a damn Fritz chip?
All this goes to show, the future of the technology world will not be created in America.
It seems like there is a push for “the fastest”, but that accounts to probably 0.5% of their sales.. Everybody else wants a reasonably cheap, reliable computer that does every day things. Get the computer smaller.. cooler.. more reliable.. and it will start to matter. Put in faster and hotter CPUS, drive up power consumption and heat generation and cooling requirements, and how does it really help the average user?
Because this continual speed king battle drives prices of lower-end processors down. Ditto that for 3D accelerators, memory modules, hard drives, etc.
Intel has joined the party and decided to kill the PC industry.
Yes, and firewalls spelled the death of networks. Imagine them building products that would allow *you* to protect *your* own data. The %&#*@&% pigs!!!!!
LOL
Why does this sort of thing happen? I mean, there is a point where people will say, “This is stupid. Every time I buy a computer, there is a ‘better’ one for the price I paid. I’ll just wait a while before I buy again.” Then the market crashes. If consumers can’t tell there’s a difference, they will behave as such, period. Same thing happened in 1984 (year, not book) in the video game industry. I just wonder when it will be?
Anyway, I am waiting on FMD-ROM and Serial ATA, and maybe something like QDR SDRAM, or something. I don’t know…
booc0mtaco: I mean, there is a point where people will say, “This is stupid. Every time I buy a computer, there is a ‘better’ one for the price I paid. I’ll just wait a while before I buy again.”
These people would always be like that, and was like that before the price/performance wars between AMD and Intel.
The economy for PCs is not flat because people don’t want to spend money to buy something expensive, and the next day get retrench. It is recession.
But I do believe the industry would break down one day: just how cheaper could it possibly go?
Intel has joined the party and decided to kill the PC industry.
Yes, and firewalls spelled the death of networks. Imagine them building products that would allow *you* to protect *your* own data. The %&#*@&% pigs!!!!!
LOL
A firewall doesn’t run your computer. It is not your operating system. You can get an open source firewall. I don’t think you’ll be able to get an open source Intel FritzInside processor. You will never know what the processor is doing.
With FritzInside being mostly used for DRM, it is naive to think that “you” will be able to work with “your” data. There is nothing that Intel or Microsoft has put out that champions the rights of invididuals to their own data.
With Intel’s FritzInside processor, the following laws will be able to be enforced on your machine (from The Associated Press):
Government may monitor religious and political institutions without suspecting criminal activity to assist terror investigation.
Government may prosecute librarians or keepers of any other records if they tell anyone that the government subpoenaed information related to a terror investigation.
Government may search and seize Americans’ papers and effects without probable cause to assist terror investigation.
So, tell me again why is it good to have a government sponsored FritzInside processor?
Once a government gets too paranoid, it breaks, trust me. I bet Bush would go in the way of his father. The government is paranoid, I guess…
A firewall doesn’t run your computer. It is not your operating system. You can get an open source firewall. I don’t think you’ll be able to get an open source Intel FritzInside processor. You will never know what the processor is doing.
A P4 microprocessor doesn’t run a network hub. It is not your network. You can disable LaGrande and get an open source operating system. I don’t think you’ll be able to get an open source hub. You will never know what the microprocessor inside the hub is doing.
How long do you want to continue trading pointless comments like this?
With FritzInside being mostly used for DRM
There is no such thing as FritzInside.
it is naive to think that “you” will be able to work with “your” data.
It is naive to run around without proof decrying the death of personal liberty. It is also incredibly stupid.
With Intel’s FritzInside processor, the following laws will be able to be enforced on your machine (from The Associated Press):
The AP has a history of publishing outrageous lies about both Bush Administrations.
So, tell me again why is it good to have a government sponsored FritzInside processor?
There is no FritzInside processor. If you are talking about Intel’s processor, it is not government-sponsored. AMD and many other companies are implementing similiar technology, and neither are they.
Once a government gets too paranoid, it breaks, trust me. I bet Bush would go in the way of his father. The government is paranoid, I guess…
How come the same thing never happens to the media? When so many journalists decried the horrible death toll that would result from the Republicans’ school lunch program, and that plan was implemented, and millions of children did *not* die, where were all the people demanding to know where all the propagand came from? Where? Where were all the dissenters? The politicians? The public outrage at being stuffed full of horrible lies? What happened?
People just got mad at the next big lie.
Sensationalism is somewhat like a large corporation suing a little company; they don’t do it because they were *wronged*; they do it because regardless of the outcome the legal fees will crush the little company. Popular opinion works similarly, except that since no one holds journalists accountable they take potshots at bigger companies. Whatever strikes their fancies and gets them recognition for helping – and get this! – the “little” guy.
Oh! I just realized that you may be referring to Senator Fritz Hollings. Well, in any event, he is an outspoken opponent of the Bush administration. Hopefully he and all the other liberal Democrats will go the way of the dodo.