Itanium SpecInt Benchmarks at TheReg

Post #5,259 by Another Scott 8/15/01 10:37:04 AM Reply	Itanium SpecInt Benchmarks at TheReg [link\|http://www.theregister.co.uk/content/3/21032.html\|Benchmarks put skids under Microsoft's IA-64 pitch] SPEC's published its latest [link\|http://www.spec.org/osg/cpu2000/results/res2001q3/\|SPECint] results for a number of single-processor systems including a Dell Precision 730 running 64-bit XP on a 800Mhz IA-64. The Itanic steams in last with a base figure of 314, 24 per cent slower than the next slowest surveyed this quarter, which as it turns out is a 750Mhz UltraSPARC II Sun Fire box. The high end chips return North of 600, including Dell's own 1.8Ghz Pentium 4. The Dell Itanium has 2 GB of RAM, 2 MB of L3 cache and a 10k rpm 18 GB Ultra3 SCSI disk while the Dell 1.8 GHz P4 has 256 MB, no L3 cache and a 20 GB 7200 rpm ATA/5 disk. Pretty pathetic. I can hear it now... "But 64-bit XP still has the debugging code in place. It'll be much faster when that's removed! Itanium is much faster than the P4 on optimized code!" I step forward, 23 steps back... Cheers, Scott.
Post #5,482 by inthane-chan 8/16/01 5:46:09 PM Reply	'Bout what I've been hearing. Itanium blows goats. It actually isn't the software's fault - it is just a sucky processor. Supposedly the next gen ia64 processor will be where they get it right, if the market forgives them. "What do I care for your suffering? Pain, even agony, is no more than information before the senses, data fed to the computer of the mind. The lesson is simple: you have recieved the information, now act on it. Take control of the input and you shall become master of the output."
Post #5,487 by lordbeatnik 8/16/01 7:00:45 PM Reply	You're being too kind I'm doing software work on the damn things and boy do they suck! Compilation of programs that take 20 and 35 minutes on my P3 700 take about 50 minutes and an hour and a half respectively on a 1 processor Itanic! Only with a 4 processor 733MHz model will an Itanic be able to perform a linux kernel compile as fast as my little P3. And everything they say about the heat they produce is all too true. In an otherwise cool lab, I start sweating after a while when I'm working between two of the ugly beasts. McKinley better be good because Merced is a total joke. I'm hoping something will raise out of the embedded world to challenge Intel. The power consumption for Itanics is just looney. It's certainly no poster child for efficiency. Dave "LordBeatnik"
Post #5,513 by Another Scott 8/16/01 10:19:49 PM Reply	130 Watts at 800 MHz. It's a nice heater... [link\|http://www.eetasia.com/ART_8800099676_499481,1034362.HTM\|With Merced in hand, OEMs await follow-on] in EETimes Asia. I'm waiting for the PocketPC version. >:-) Cheers, Scott.
Post #5,763 by lordbeatnik 8/18/01 2:47:19 PM Reply	We make many jokes about its spaceheater ability We've been joking about managers turning down the heat when winter gets here so save money. This of course would cause all the developers to fight to get itanics into their offices to use as space heaters. =8^) The frightening thing it thinking about the heat generation a McKinley must have. With 50%+ higher clock rate and 25%+ more die area in the same .18 micron process the chip must be ungodly hot. I've long joked with a college friend that soon computers will start coming with Mr Coffees built in to make use of all that heat. It appears that soon Itanic processor systems will dump enough heat for powering small ovens. Doesn't everyone need an oven that also cruches seti@home? =8^P Dave "LordBeatnik"
Post #5,823 by ben_tilly 8/19/01 10:03:52 AM Reply	Itanium has competition AMD's Sledgehammer for one. Personally I am betting that when games hit the wall with 32-bit addressing, AMD will be what they move to. And with the listed performance numbers, I would wonder if TransMeta can emulate Itanium faster than Itanium runs... And if the performance problems are bad enough that people are willing to throw away x86 compatibility, don't forget that there are several 64-bit chips out there which could potentially move from their current roles in servers to general desktop use. Cheers, Ben
Post #5,894 by static 8/19/01 8:54:13 PM Reply	Certes. I'd like to see how HP's PA-RISC chip fares in a fair comparison. It always surprised me that the CPUs in their T520 ran at only 140MHz. Wade. "All around me are nothing but fakes Come with me on the biggest fake of all!"
Post #6,173 by lordbeatnik 8/21/01 8:12:45 PM Reply	perhaps,perhaps,perhaps The way I see it, the current 64 bit chip makers are too used to fat margins to ever make a volume play. Perhaps one will prove me wrong but right now most appear to be too good a rolling over and playing dead with the approach of the itanic. Thus, I see two paths to nothingness for the itanium. 1) x86-64. The least "painful" path. Everyone gets a "free" backwards compatible upgrade to 64bit. The big downside: it's still x86, just with still more complexity tacked on. 2) embedded assension. Just as intel is using dominance in the desktop to lay seige to high end servers, I can see Intel getting so attached to fat margins that it falls victim to the same play from the embedded world. The risc embedded processors are designed to be everything itanium isn't: simple and efficient. Perhaps I'm dreaming too much, but I hope for something more power efficient than the itanium ( and to a lesser extent x86 ) to emerge. With the way itanium is going, air cooling soon won't be enough. The costs and hassles of a water cooled computer are not something I look forward to. Dave "LordBeatnik"
Post #6,201 by tonytib 8/21/01 11:13:04 PM Reply	X86-64 for masses, 64-bit RISC & Itanic/Alpha for servers Intel loves its margin on server chips, too. Check out Itanic prices, and Xeon prices versus desktop P3/P4 prices. AMD seems a lot more interested in the volume market. I think the desktop market will most likely go x86-64; for one, the Itanic will simply cost too much (die size, heat, and p/s issues). I don't think the current RISC server chips (SPARC, PowerPC/RS6000) will come down much in price -- or will have problems staying ahead of Itanic at least until Intel merges in Alpha technology. And, I don't think you're going to see embedded 64-bit chips (e.g. MIPS) have an impact in the desktop market -- unless Linux really takes off in the desktop. Windows ain't going to run on them. Basically, the PC business is pretty boring now and the excitement is in other areas. I'm more interested in the lastest DSP's than the latest Athlons. Tony
Post #5,534 by tonytib 8/16/01 10:58:13 PM Reply	No, it'll suck until 2005 or so....that's when Intel will get the Alpha technology melded in. I had written off Itanic until the Alpha deal. Now I give them a chance. From what I've read, Intel made some pretty bad choices when the designed the Itanic architecture. And, the benchmarks reflect this. I think by the time McKinley hits, it'll still be behind Sun and IBM. Intel doesn't have a good track record trying to design revolutionary architectures from the ground up -- anyone remember the iAPX-432? Tony
Post #26,041 by Another Scott 1/28/02 10:19:47 AM Reply	Rumblings about Intel X86-64 at The Inquirer. [link\|http://www.theinquirer.net/25010201.htm\|Here]. It seems Intel has a skunkworks project to fight AMD's Hammer if Itanic doesn't start doing well and Hammer takes off. According to a [SJMN] story up there, and based on an engineer, who quite sensibly declines to be named, the "Yamhill Technology" which is essentially based on X86-64 features, will be built into allegedly 32-bit chip Prescott with the option to turn the code on or off, depending on how well the Intel folk think sales of Itanic are still crap, and sales of Hammer are very good. As we have reported before, the work is going on in Ronler Acres and we can assure you that if Intel has to bite the bullet and dump Itanic, it will do so. The newspaper says that the Itanium has cost Intel $1 billion over the seven (actually eight) years it's been in the making, but we believe the true figure is higher than that. Intel rarely, in our experience, puts ideology above market forces and if the Itanium fails to bolster its investment, it will have no compunction in pulling the trigger. But it's a bit too early to call it a day, isn't it? What with the agreement with Compaq over Itanium migration, and the imminent introduction of three McKinleys, which certainly hold out some promise, there's still a lot to play for. Cheers, Scott.

Welcome to IWETHEY!