Slashdot Mirror
Intel's 64-Bit Pentium 4s Hit The Streets
ThinSkin writes "Nearly 18 months after rival AMD released its 64-bit processors, Intel quietly added its first 64-bit Pentium 4 microprocessors to the market on Sunday. Four versions of the Intel Pentium 4 6XX series were announced at speeds up to 3.6-GHz, a frequency grade lower than the existing 5XX series. Prices will range from $224 to $605. Intel also added the 3.73-GHz Pentium 4 Extreme Edition to its lineup, a $999 chip that is fabricated on a finer 90-nm process than its older 130-nm P4EE components. As Slashdot previously reported, the 64-bit series will likely be the major enhancement to the Pentium 4 line before the introduction of the Pentium D "Smithfield," Intel's first dual-core part, which is slated for next quarter."
Um - A little slow? I've had an EM64T "630" for about 3 weeks now.
The Inquirer is reporting that Intel will counter the AMD Turion 64 Mobile Processor with a Pentium M Extreme Edition. It is an alphabet soup of potential Intel Pentium M releases and you'll probably have to read it through twice or thrice to understand it all, but an interesting and inciteful read nonetheless.
There's another interesting article about the future of 64-bit as it relates to Intel here.
And of course, we can't forget our beloved Celeron.
I believe it was because they found they couldn't claim IP rights to numbers such as '486' in order to prevent other manufacturers releasing products with the same name.
According to some tech sites, it hits a maximum of 52deg Celsius under full Prime95 load. That's a lot better than the 65+ hit by the .09 micron based Pentium 4's.
Intel has done its homework on these Prescott-based EMT64 chips. They allow a reduction in voltage and die size, which results in a cheaper core too.
Sure, there is loads of programs that make use of the full 64 bits. Several vendors offer 64bit flavours of their distributions, such as Fedora, Ubuntu, Gentoo, and Suse.
The performance gain is found in how the chip itself works. 64-bits breaks the 4GB memory process limits of 32-bits. In 32-bits, a single process can access 2^(32) bytes -> 4 GB of linear memory. In 64 bits, you can therocally access 2^(64) bytes. Pratically, in Windows x64, a process is limited to 16 TB of memory. Plus, there's extra registers that a program can use.
WTF does "Extreme" mean in relation to a CPU?
"EE" stands for "Extremely Expensive", "Centrino" means "doesn't suck on laptops". Other than that, I also am now completely lost as to what Intel's lineup actually is. Their marketing department are fucked.
Dave
I write a blog now, you should be afraid.
Wow, and only 10 years after Sun's UltraSPARC, 13 years after the DEC Alpha, and 14 years after the MIPS R4000
Donate anything 300mhz or above to your local school district. I am a computer tech for my local district and we need all the help we can get.
Does anyone remember what a VAX-780 cost? That was a 1 MIPS CPU.
Bruce Perens.
Emphases
If a job's not worth doing, it's not worth doing right.
It's sort of like throwing a few billion dollars in the fireplace.
Bruce
Bruce Perens.
embedded systems.
there are far, far, FAR more embedded computers in the world than there are PCs. millions more.
the z80 originated in 1976 and is still mass produced.
-
The problem with comparisons in MIPS (millions instructions per second) is that for different architectures, it takes a different number of instructions to accomplish the same amount of work. This doesn't matter for AMD/Intel (Athlon/64/Pentium/Xeon) as they all use the same (mostly) Instruction set. It would matter if, as the gp said, you wanted to compare Sun, Apple, Intel Itanium, or Intel Pentium, since they use differnt ISAs.
Flops are a little cleaner, but still only test the Floating Point units. What if a chip has good floating point, but really shitty integer (it could happen). It would still suck to use for a lot of things. The only decent way of testing something (AFAIK) is to run it with the programs you intend to use it for and compare.
Pentium Pro, Pentium II and Pentium III have all been based on the same core. The PPro was ridiculously expensive to produce at the time because of the on-die cache so they moved to the P2 with separate cache chips on the board the slotted chip was on. As they got the process down, they went to the P3 which returned to the on-die cache. Orginally the P3 was slotted for compatability reasons but they went to a socketed chip for cost purposes later in production.
The P4 was a completely different architecture (NetBurst) which was intended, from the ground up, to hit high clock speeds, without concern for actual performance (granted, once they hit 800MHz FSBs, the P4 finally started showing its stuff). This was one of the biggest mass-market counterexamples to the MHz Myth, with first generation P4s (1.5-1.7GHz) getting solidly beaten by cheaper, lower-powered, lower-clocked P3s (1GHz-ish).
Granted, the P4 wasn't a complete waste - there were some very good technological advancements in it. The Pentium-M is essntially a P3ish core that has some of the enhancements from the P4 (quad-pumped bus, SSE2, awesome branch prediction) added to it but retained the P3's lower power consumption & clock-efficiency. Not to mention that a 3+GHz P4 with an 800MHz FSB is going to be an absolute monster at number crunching, given software that properly uses the SSE/SSE2 (vector math) extensions.
For the most part, however, the launch of the P4 was a disappointing event that helped AMD grab mindshare & marketshare in the CPU market, particularly with those who actually care about more than cute commericials and buying the cheapest thing Dell is pushing out the door.
my sig's at the bottom of the page.
It's possible that Linux might be quicker off the mark to support the chip - you might even be able to patch it yourself with the right gcc & kernel - but since you said Dell you're really talking about XP. And XP isn't going to see support unless Microsoft produces another service pack to shore up the gulf of time that exists between now and their next release. Even if they do release SP3 with 64 bit goodness, you're still talking 6 months or more and the chances are that it will only offer token support in a limited number of places.
On top of that, buying unsupported bleeding edge hardware is always a bad proposition. Its expensive, doesn't do anything and will probably be obsolete by the time that something comes around to supporting it. It would have been as foolish to buy AMD64 back when it was released.
Specifically, it shows two things (note, the clock throttling wasn't working on the Opteron processors mind):
You are not Joe. You want to know what the designations mean while Joe is just confused by 'nerdy' specs and therefore takes advice from the 'more educated' retail dude. Big mistake.
... :)
AFAIK:
P4 is the bog standard current Intel chip (32 bit)
Prescott is a P4 cheap and nasty P4 with even higher clock speeds with lower power/cycle (with the innevitable exceptions). Similar to the difference between P3 and P4, only a smaller difference.
Celeron is a cheap P4 with stripped down cache
Centrino, same as Pentium M is a low power version with more power/cycle and lower clock speeds (similar to AMD chips) that has low power draw and is good for laptops (the only good thing to come from the Intel line-up in recent years imho, only it wasn't really designed by Intel hee, hee)
P4EE is a P4 with a ridiculously large cache (and high price) to get more from a basically crap design, so that they can even compete with AMD on the top end of comparitive benchmark tests.
P4-64 has an extended memory addressing range ie., can take more RAM. Irrelevant and slower for the vast majority of users
Dual core is next-gen tech that AMD and Intel are working on that will be equivalent to dual-processor systems that have been around for ages. Not much more bang for much more buck.
Hyperthreading is psuedo-dual core on one chip, very good for repetitive operations (like video encoding), but no good for unpredictable code (almost everything else)
You bought proprietory tech (RAMBUS). Your mother was a hamster and your father smelt of elderberries!
Hope this helps de-muddy the marketing waters.
Something that i have seen here that is a bit of a misconception is that some people seem to think 64 bit processors are twice as fast, or can handle twice as much data as a 32-bit processor. This is not so. 4 types of data run through a processor: Integer Address String Floating Point in a 32 bit processor, the ranges are as follows Integer - 32 Address - 32 String - 64 Floating Point - 128 in a 64 bit processor, however, the ranges are as follows Integer - 64 Address - 64 String - 64 Floating Point - 128 All that is simply saying, by saying a 64-bit processor, is that the processer can handle 64 binary instructions per clock cycle, twice as many as a 32 bit processer, only in the integer and address ranges. so it is not truly twice as fast, and neither can it truly handle twice as much data, just 2 of the 4 types of data have been doubled. theoretically, a 64 bit processor could address 16 exabytes of ram (i believe this is 16 million gigs of ram) but obviously there are limitations to this. and interesting fact i thought i would stick in here :p
ps. - my ranges for floating point and string might have been reversed, now that i think about it i cant remember whether floating point is 128 bit or whether string data is 128 bit.
"Potpourii doesn't taste as good as it smells." - Dark_Link2135