Slashdot Mirror
Intel Begins Shipping 64-bit Prescotts
Rucas writes "With a minimum of fanfare, Intel has begun shipping a version of the Pentium 4 with 64-bit instruction set extensions. The news came to light not via an Intel press release, but rather through the spec sheet for a new server from IBM. In the midst of the new IBM eServer xSeries servers based on the recently released 64-bit Xeon is a blade server powered by the 64-bit Prescott. This marks the first product appearance of the new CPU."
"Intel president Paul Otellini said that Intel was building the capability for its 64-bit extensions into Prescott. At the time, he said that Intel wouldn't enable the feature until Microsoft released a 64-bit version of Windows; that operating system is expected later this year. "
Does this mean that we will have disabled and enabled versions? Like the old 486SX and DX (SX I understood was a disabled/failed math co-processor). I suppose like all their other chip lines, each will be labled distinctly with some marketing nomenclature.
Have you Meta Moderated t
Which makes it even more surprising that Intel wasn't really hyping this themselves. You'd think they would want every reason to get people to switch back to them for high-end tasks where 64 bits could be useful.
"Bang" is not only raw speed. It's working chipsets, no PCI transfer issues (e.g. sound clicks when transferring via IDE), no lockups. I'm not so sure AMD is ahead in every aspect. Nothing compares to an Intel chipset + Intel CPU combo IMHO.
Of course it runs NetBSD. BTC: 1NT7QvbetmANwaMzhpVL6
It's pretty astounding that major jump from 32-bit to 64-bit processing isn't even mentioned by Intel.
Think about how big a jump it was from the i286 to i386 (16-bit to 32-bit.) That release was a major deal for Intel.
Chew: You Nexus, huh? I design your eyes.
Roy: Chew, if only you could see what I've seen with your eyes.
Go Intel! only 7.5 months behind Apple and IBM who collaborated to put out a nice 64 bit solution in August of '03.
"AMD welcomes Intel to the world of AMD64, said Ben Williams, director of server and workstation marketing at AMD.
It's kind of funny to watch. Intel is choosing their words very carefully. They're saying things like, the new chip "will run programs currently being developed for AMD's 64-bit processors with very little modification." They absolutely refuse to call the new chip "AMD compatible" even though that's exactly what it is. Intel is having a lot of trouble facing the facts: they poured zillions of dollars and years of R&D into an architecture that nobody wants (Itanium), meanwhile AMD got it right (Opteron) and now they're playing catch-up.
You'd think that Intel, moreso than anyone else, would know that you just can't kill x86.
Tired of FB/Google censorship? Visit UNCENSORED!
Funny my AMD 760MP based system has not show any of the issue you describe.
At the time, he said that Intel wouldn't enable the feature until Microsoft released a 64-bit version of Windows; that operating system is expected later this year.
But according to Computer World,
Microsoft Corp. has further delayed versions of Windows for PCs and servers equipped with x86 processors with 64-bit extensions. Analysts said the extra delay will slow the advent of 64-bit desktop computing and provide a head start for rival operating systems on servers.
Windows Server 2003 for 64-bit Extended Systems and Windows XP 64-bit Edition for 64-bit Extended Systems won't be available until the first half of 2005, a Microsoft spokeswoman said yesterday. The 64-bit Windows XP client was originally due early this year but had already been delayed. The server software was scheduled for late 2004. Who's correct?
Begin sarcasm I want my 64-bit Prescott chip desktop with the most awesome world's number 1 OS a.k.a. Microsoft's Windows XP (64-bit) and I can't take these delays any longer. Waaaa. end sarcasm
Seriously, though... why not release 64-bit prescott for retail and let people install whatever they want... e.g. Linux (64-bit)...
486SX - a 486DX with its FPU disabled.
486DX - 486SX with a working FPU.
487 - 486DX with a slightly different pinout for use in 486SX systems and sold as a "math-coprocessor;" actually, it would disable the 486SX and be used exclusively!
Source.
I realize its pretty much hit-and-miss with VIA chipsets (just like with Windows... make sure you're using a "newer revision" or else...), but the KT266A (note the "A" revision) is a pretty damn good chipset under Linux at least. I've never had any issues with it under 2.4 or 2.6 kernels. All the built-in on-chipset crap work. No problems with AGP 1x, 2x or 4x + nVidia GFX. No problems with ATA-100 IDE.
Uh...he did say "within 1GHz of the speed." You're talking about a chip that's well over twice as high in frequency....or, another way, nearly 2GHz faster. Or, a third way, several times as expensive. So, uh....can it.
And while you're at it...what's this about USB being broken and never fixed? I haven't noticed it on any of my AMD machines.
For the record, I have many, many Intel boxes, and many (though not quite *as* many) AMD ones as well. If cost is no object, and insano speed and/or 64bit isn't necessary, I'll choose Intel. If I'm paying for it, it will be an Athlon. If I want the fastest thing I can lay hands on, I'll get an opteron.
There is no point in bickering over vi vs. emacs. Or NVidia vs. 3dfx.
Given a choice between free speech and free beer, most people will take the beer.
I'll take a chipset that work (never had a problem with my NForce2) over one that doesn't work with a major new technology (PCIe) any day.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
Remember, the AMD hammers all have a lot of history going back to an ancient Alpha pedigree - a lot of the Alpha design team went to work on the 64 bit implementations for AMD.
It shows in the excellent IPC scores of the hammer series, and the incredible scaling for SMP systems that Opterons enjoy.
So alpha technology lives on in a small way in the AMD hammers.
I am government man, come from the government. The government has sent me. -- G.I.R.
I've said it before in this article, but I'll say it again:
/. just had an article on the A64 3000+, referring to it as a budget CPU. My wholesale prices also support this, (somewhere around $145 for an A64 2800+, maybe $170 for a 3000+.)
It's not the 64-bitness of the Athlon 64. It's the changes in architecture (more general purpose registers, on-die memory controller, hyperthreading, SSE2, etc.) The Athlon 64 out-performs any other x86 desktop CPU on the market today, including the P4EE.
Couple this with a much lower price than the P4EE. Hell,
All of this adds up to a very impressive package for my customers. And I'm getting nothing but rave reviews on the machines I'm building with Athlon 64s. To hell with whether or not they're doing what they're designed to do! They're fast, and inexpensive. Talk about a gamer's dream. And let's face it, the 64-bitness will be useful in the future, but is only an afterthought for the consumer, right now.
Is it another 80x860?
I am bringing this up because Intel refused for a long time to bring in a 64-bit x86 due to invement with HP for Itanium. It seems so odd that we have a chip based on 1970's technology.
Itanium was supposed to be surpacing x86 by now like NT replaced win95 derivitives. Intel has a notion of sunkin costs while HP would rather beat a dead horse than admit it failed after billions of billions of dollars of development. I guess its the culture of zero accountability and perfection with no room for mistakes that Fiona implemented.
x86 just wont die.
I would prefer to see the Alpha as an eventual replacement for the aging x86 and its a shame it was bought up just to boast the Itanium.
Well long live the Pentium 64-bit and forever x86.
http://saveie6.com/
The cited EEtimes article has the information totally incorrect on the CPU.
The processor is not in-fact a Prescott. IBM Blades infact use Prestonias today. Prestonia are 400/533MHz FSB Xeon processors. These processors have in reality been shipping for over two years. These are also dual CPU blades (can we expect to see EM64T enabled 4-way Foster CPUs?).
This is much more interesting than Noconas and Prescotts having EM64T technology, as it shows that the technology is being retrofitted into older currently shipping CPUs (or perhaps retro-fitted is not accurate, perhaps it has always been there?).
The article hints a the truth though, as IBM did announce a EM64T Prescott, but on the xSeries 306 server, not the blade servers.
The "buzzz" is poor shielding of the audio chipset on the MOBO...not AMD's problem...just a cheap motherboard. Also, the SB live was ENTIRELY creative's problem. They went more than 3YEARS without driver updates for those cards...meaning that they didn't account for the variations in higher-speed/ processor changes in that time... heck, I have an INTEL board/proc [i815 made by asus] from 2001 that simply refused to work with the thing. Anyway you can't blame AMD for SB problems!
486SX wasn't failed, they actually cut the traces to the FPU. Other than that the chip layout was identical.
Later SXs may have omitted the FPU completely but given that the SX was what amounted to a "loss leader" for Intel, intended to hold back AMD & Cyrix from the gates, they never spent a whole lot of time engineering the thing.
The real fun was when Intel sold a "FPU upgrade" for some 486SX systems. The "FPU Upgrade" was nothing more than a rebadged 486DX chip that mounted in a socket close to the original 486SX and, when installed, disabled the original 486SX chip.
That was the start of my hatred for Intel. I didn't buy one, I didn't get burned by one, I simply was a geek who knew what a lame duck looked like when he saw one. Crud, even back then I was doing tech support & MIS work.
Going to more CPU's in the Opteron model still takes a lot of work. From what I've heard, the current chips lose quite a bit of performance due to the cache coherency protocol (ccNUMA memory using MOESI consistancy model) running at 8 processors. The chips themselves only have 3 links coming off of them, which means that at most, you can build 3D topologies of chips. Anything bigger than a cube in 3D gets to be rather slow.
So.... To get beyond 8 processors in a true global-memory ccNUMA configuration, AMD is going to need to a) extend their coherency protocol to deal with that and b) come up with a way (external HyperTransport switch, perhaps?) to get more dimmensions off their chips. 6 is the minimum to do a 3D torus configuration (which is the topology of a Cray T3E and a Cray X1). SGI Origins, which are also ccNUMA boxes, uses a 4D hypercube.
Go Badgers! -- #include "std/disclaimer.h"
It is believed that Intel could turn on 64-bit capability in already shipped Prescott cores just by having the user load a microcode update (just like the ones shipped in the BIOS to fix CPU errata.)
Does anyone know if this is true, or did they have to fix some circuit bugs before it was re-released? It apparently was true for hyperthreading on early Northwoods.
Noticeably no Spec 2000 numbers have been released for Nacona with 64bit addressing!!!
0 4q2/cpu2 000-20040419-02982.asc0 0/results/res2004q1/cpu2 000-20040126-02775.asc0 0/results/res2004q2/cpu2 000-20040518-03050.asc
In case anyone finds them, I've put some contemporary numbers for comparison. These are Spec 2k, Int, 64 Unix OS, base:
Opteron 2.2Ghz 1338 Suse 64bit
Itanium 2, 1.5Ghz 1380 HPUX
Fujitsu, 1.9Ghz 1174 Solaris
Nothing yet on Power5 on the spec.org website, and there are faster 32 bit Opteron numbers.
Spec(tm) links:
http://www.spec.org/cpu2000/results/res20
http://www.spec.org/cpu20
http://www.spec.org/cpu20
I ran into a similar issue dealing with some local banking institutions years ago. EVERY bank in this area was hooked on Token Ring. Now this was understandable from the perspective that Token Ring was arguably better than most older forms of networking, but this was 1997. New installations were still getting 16 Mbit Token installed. In some case we saw twisted pair installations, but they were still running 16 Mbit Token! What the hell? Ethernet over twisted pair was so much cheaper, faster, AND established.
Then I started noticing that EVERYTHING was IBM. The servers, the workstations, even the CABLING. I saw this at every bank we did work for (at least 8 different organizations).
So if it wasn't for the quality, expense, and/or speed, what was it? I later learned that this was a common theme in many larger organizations and it had a lot to do with how much IT stock was owned by the execs.
A friend of mine - a CIO - relayed to me that when a large organization buys a ton of equipment from IBM, the resulting sales figures usually give a bounce to the stock. Better still, if you coordinate your efforts with other execs in other companies, you can often make yourself a tidy profit.
During my time consulting for these banks, management did not want to hear about any other solution that wasn't IBM. I suspect that most Fortune 500 companies play a similar game with Dell product - and that would certainly help explain Intel's entrenchment.
"...Well, there's egg and bacon; egg sausage and bacon; egg and spam; egg bacon and spam; egg bacon sausage and spam..."
For us in .uk a "Prescott" is a byword for a bumbling buffoon.
.uk you should not that Prescott is our Deputy Prime Minister.
Characterised by:
1). Rambling incoherent communications.
2). A violent temper which could blow at any time.
3). A tendancy to do a rapid about about-face whenever challenged by the realities of hard work.
Do we really need a chip like this?
Note: For those not in
They're not hyping it since their pal Microsoft has delayed 64-bit Windows. Intel probably doesn't want to put pressure on Microsoft since Microsoft might favour AMD more.
Intel have already lost out on providing XBox2 with a CPU.
One thing I have noticed is that ibm and hp have 1 opteron cpu boxes with no redundant power supplies, and then 4 cpu boxes with 32 ram slots, 2 PSUs that cost 12k without ram or HDD. Neither of them is selling a fairly normal dual cpu box that is 2U high - they are either web servers or computational nodes for clusters, or big honking 4 way boxes.
ostiguy