2CPU.com » News » November 2002 » Intel Ships Gallatin-based P4 Xeon MPs

In an attempt to be more competitive with Sun Microsystems in the mid-range server market (four- and eight-way SMP machines), this article at NewsFactor states that Intel has begun shipping their new version of the high-end Xeon CPU based on the Gallatin core.

Intel said this new version of the Xeon MP provides a performance boost of 19 to 38 percent compared with the chip's previous version, dubbed Foster.

The performance increase is partially due to a smaller manufacturing process than previous Xeons and an increase in L3 cache size.

IBM, Dell and HP are already on board for Gallatin-based, Xeon-powered servers, so other manufacturers can't be far behind. Could Gallatin be what Intel needs to make inroads in to the mid-range enterprise server market?

Jim_ Administrator Posts: 3574 Joined: 2000-03-15

#19518 Posted on: 11/05/2002 05:06 AM /me golf claps. Well done. Well done. Welcome aboard. [url="http://www.2cpu.com"][size=1]2CPU.com[/url] - Because two are always better than one! [url="http://www.jimkirk.org"]jimkirk.org[/url] - Not a Myth any Longer. Just a Dad.[/size]

mattsteg Senior Member Posts: 981 Joined: 2000-05-07

#19519 Posted on: 11/05/2002 05:08 AM Hey HEMI, good to see you posting on the front page, Hello.

JohNEE <font color="#E22000"><b>News Monkey</b></font> Posts: 363 Joined: 2000-05-10

#19520 Posted on: 11/05/2002 05:10 AM Nice first post, chap! I picked this subject for my news post sample too - I have my fingers crossed that Hooz likes what he saw and that I can join you! J. The Toaster Dual PIII 800E, T100, Win2k, 512MB Crucial, 2940U2W + 2x 18.4 Gb 10k II's, GF2 Ultra, TBSC, Pioneer DVD-ROM, Plextor 24/10/40A, 3com NIC - in a modded Aopen HX08.

Jemmott Unregistered

#19521 Posted on: 11/05/2002 05:54 AM Good 1st post HEMI!

DPF No one of Consequence Posts: 730 Joined: 2000-10-15

#19522 Posted on: 11/05/2002 06:07 AM The 2GHz - 2M ones are only a measly 4 grand? Hell, 8-ways for everybody!!!!! And beauty post HEMI -DPF [color=blue][size=1]"The little BX that could"[/color][/size] P2BDS, S370-DL's|2x1GHz@1.12GHz SL5QV|1GB Mosel Vitelic PC133|Herc GF2GTS Quadro Mod|UltraPlex 40TSi|Plexwriter 8/20|Audigy2|Sony DRU510A "If you give a man fire he will be warm for a day -but- If you set that man on fire, he will be warm for the rest of his life"

LRSeriesIII Aspiring Rocket Scientist Posts: 1120 Joined: 2002-08-29

#19523 Posted on: 11/05/2002 07:03 AM Nice post, very nice I wonder how the Xeon MP compares to the Suns on a bang to buck level. ->Computers ->Folding for team 3074

mdenfp Registered User Posts: 11 Joined: 2002-02-19

#19524 Posted on: 11/05/2002 07:26 AM Only 2 MB of cache? The Pentium II's could be ordered with 2MB and that was, what 4 years ago. If cache matters so much, why hasn't it increased with the newer chip families?

morpheus777 Burninating the peasants Posts: 1664 Joined: 2002-07-02

#19525 Posted on: 11/05/2002 07:43 AM W00t HEMI! So 2cpu has got so far at least one new news guy. 20" iMac, 2.66GHZ Core 2 Duo, 2GB RAM, 500GB HDD, ATI Radeon HD 2600 Pro 256MB, OS X 10.5.3 Leopard Apple MacBook 1.83GHZ Core Duo, 1GB DDR2, OSX 10.4.11 Tiger

LRSeriesIII Aspiring Rocket Scientist Posts: 1120 Joined: 2002-08-29

#19526 Posted on: 11/05/2002 07:56 AM The L2 cache on the PII was not on die, while the L2 cache on the current processors is, so I think that has an effect on it. I would imagine they don't have the room on the CPU core they had when it was off die like the PII (the PPro has on die L2 cache, some of them have alot of it, ever looked at the size of those suckers?). Of course, the on die cache is full speed... ->Computers ->Folding for team 3074

questionlp the Cowardly Tech Posts: 323 Joined: 2002-02-14

#19527 Posted on: 11/05/2002 08:15 AM Originally posted by mdenfp Only 2 MB of cache? The Pentium II's could be ordered with 2MB and that was, what 4 years ago. If cache matters so much, why hasn't it increased with the newer chip families? As LRSeriesIII mentioned, the cache in the P2/P3 Xeon processors were not on-die and it was structured differently (the cache was L2 rather than L3; the Gallatin processors also have 512K on-die L2 cache in addition to the L3 cache). By having the cache located outside of the die, latency increases and it becomes even harder to ramp up since it needs to keep in-line with the processor and running the bus out to the cache at anything over 700Mhz can be a PIA. Intel did the same with the Itanium and the Itanium II. The original Itanium had external cache (same brick, different chip) of either 2MB or 4MB. The Itanium II brings the cache on-die but also decreases it to either 1.5MB or 3MB. Another reason might be is that anything above 2MB might not increase performance a whole lot, except in certain circumstances. The additional cost and decreased yields due to increasing the on-die cache is probably something Intel didn't want to deal with just yet. With regards to the Pentium Pro, the cache was in the same package but it was a separate core. I think cramming 256K, 512K or 1MB of cache into the same 0.5/0.35-micron die would have been too risky for Intel (now if it was HP and their PA-RISC processor... they might have taken the plunge). Source: http://sandpile.org/impl/p6.htm Just my US$ 0.01 [http://closedsrc.org/] My Rig: 2x 2.4GHz Xeon, 1GB PC2700, Supermicro X5DAL-TG2, ATI AIW 9000, SB Audigy 2 Plat/EX, 2x 80GB 7200.7's, Plextor 40x and 8/20, Pioneer 106D, Antec 1000AMG, Enermax 460W EPS-12V

todder Unregistered User Posts: 2058 Joined: 2002-01-30

#19528 Posted on: 11/05/2002 08:17 AM Eventually there will be no need for Ram or HDDs cus it will all be cache 50 Thz with 100TB onboard Cash [COLOR="Blue"]". . . I think a general Government [is] necessary for us, and there is no form of government but what may be a blessing to the people, if well administered; and I believe, farther, that this [The Constitution] is likely to be well administered for a course of years, and can only end in despotism, as other forms have done before it, when the people shall become so corrupted as to need despotic government, being incapable of any other. " -Benjamin Franklin [/COLOR]

questionlp the Cowardly Tech Posts: 323 Joined: 2002-02-14

#19529 Posted on: 11/05/2002 08:20 AM I sure hope that the cooling device for such a processor/chip doesn't flake out one day or you can just kiss all of that data good-bye. Oh wait... you can already do something like that with the DeathStar hard drives... my bad [http://closedsrc.org/] My Rig: 2x 2.4GHz Xeon, 1GB PC2700, Supermicro X5DAL-TG2, ATI AIW 9000, SB Audigy 2 Plat/EX, 2x 80GB 7200.7's, Plextor 40x and 8/20, Pioneer 106D, Antec 1000AMG, Enermax 460W EPS-12V

bimo Special FX and mod freak Posts: 322 Joined: 2000-06-13

#19530 Posted on: 11/05/2002 08:35 AM Keep 'em rolling HEMI. Well done!! Waaahahahahahaa!!! Moommmyyy!!! Want that procs for my rendering slaves. Cheers!;) Main: XW9300, 2xOpt285, 8GB, 600GB HDD, QFX 4600, 640GB Medea RTR320, Decklink HDXtreme || 2nd: HP 8510w, 4GB, 320GB, QFX570m || 3rd: Dell Precision M90, 2Ghz, 2GB, QFX Go 2500|| 4th: HP nw8240, 2GHz, 2GB, FGL V5000

Athemeus SMP Decepticon Posts: 576 Joined: 2002-03-02

#19531 Posted on: 11/05/2002 12:50 PM Originally posted by questionlp above 2MB might not increase performance a whole lot, except in certain circumstances. It does in the big enterprise database stuff that is at discussion here. That's why my ancient SPARC box has 8mb of cache.. On a tangent, anyone else keep reading/seeing "Gelatin" with the name of this CPU? I can't help but thinking of a jiggling bowl of green stuff. 2 x 2.8 Xeons on a PC-DL I am an out of control, hardware junkie.

BuyALambo SMP Decepticon Posts: 171 Joined: 2000-03-28

#19532 Posted on: 11/05/2002 12:52 PM /me raises the roof keep the news flowing HEMI_

HEMI Administrator Posts: 2716 Joined: 2001-12-18

#19533 Posted on: 11/05/2002 04:36 PM Originally posted by Athemeus On a tangent, anyone else keep reading/seeing "Gelatin" with the name of this CPU? I can't help but thinking of a jiggling bowl of green stuff. Oh no...Now every time I think of these CPUs I'm going to think of a certain scene in the movie "Airplane." I'm happy to be helping out with the news for this great site. Hopefully things will go well. Unix is user-friendly; it's just picky about its friends.

Sink Registered User Posts: 1189 Joined: 2001-08-20

#19534 Posted on: 11/05/2002 05:40 PM The News article mentions Xeons as 32-bit, does this new Xeon chip stay at 32-bit. If you want to keep your 8GB of data in memory you really need 64-bit OS's and CPU's (aka Sun Sparc Solaris with Ultra Sparc III CPU's, lots of them or HP-UX etc.). However, as a middle-tier machine supporting a Portal, CRM, or CM (Content Mangement), these might be just fine and better than what is currently available for sure. 1. Dell Precision 670|2x2.8GHz|4096MB (quiet). 2. Sun Blade 2000 |2x900MHz UltraSparcIII-cu(64-bit)|4096MB Heat: zinken255

anatolli Registered User Posts: 1881 Joined: 2001-07-18

#19535 Posted on: 11/05/2002 08:06 PM Not quite. Current Xeons use something called PAE (Page Address Extension). Allows the xeon processors to address up to 64GB of memory. Windows NT Server (enterprise?), win2k adv server and datacenter can use PAE. The problem with PAE is that it comes with massive performance losses. Check www.aceshardware.com for more info. anatolli Life's short and hard, like a body building elf

LRSeriesIII Aspiring Rocket Scientist Posts: 1120 Joined: 2002-08-29

#19536 Posted on: 11/05/2002 08:45 PM Also, Intel already has the Itanium for 64 bit stuff. I think this is more along the lines of something to go along with Win2kAdv. Server and Datacenter Server. 32-bit stuff, but still big iron. (at least compared to what most of us have ever had the privilige to work on) ->Computers ->Folding for team 3074

evil_bunnY Registered User Posts: 5 Joined: 2002-11-05

#19537 Posted on: 11/06/2002 02:37 AM a "smaller manufacturing process" aka smaller die size doesnt help in chip performance, it only reduces costs and possibly allows increased clock speed. nice post otherwise.

evil_bunnY Registered User Posts: 5 Joined: 2002-11-05

#19538 Posted on: 11/06/2002 02:38 AM Originally posted by LRSeriesIII The L2 cache on the PII was not on die, while the L2 cache on the current processors is, so I think that has an effect on it. I would imagine they don't have the room on the CPU core they had when it was off die like the PII (the PPro has on die L2 cache, some of them have alot of it, ever looked at the size of those suckers?). Of course, the on die cache is full speed... on die or not doesnt really matter, but the P3 and P4 cache is full speed (compared to the P2's half speed). and that *really* helps

SP Registered User Posts: 435 Joined: 2000-07-07

#19539 Posted on: 11/06/2002 05:42 AM Will these be socket 603 or 604 Xeons and will they run on a 400MHZ fsb or 533Mhz?

HEMI Administrator Posts: 2716 Joined: 2001-12-18

#19540 Posted on: 11/06/2002 06:44 AM Originally posted by evil_bunnY a "smaller manufacturing process" aka smaller die size doesnt help in chip performance, it only reduces costs and possibly allows increased clock speed. nice post otherwise. Smaller process == requires less power. Less power consumption == cooler CPU. Cooler CPU == generally able to run faster. Smaller processes do indirectly affect chip performance in a positive way in modern CPUs. Unix is user-friendly; it's just picky about its friends.

evil_bunnY Registered User Posts: 5 Joined: 2002-11-05

#19541 Posted on: 11/06/2002 03:05 PM Originally posted by HEMI Smaller process == requires less power. Less power consumption == cooler CPU. Cooler CPU == generally able to run faster. Smaller processes do indirectly affect chip performance in a positive way in modern CPUs. thats exactly what i said.

ambit LoveYourFruit Posts: 1243 Joined: 2000-10-15

#19542 Posted on: 11/06/2002 07:25 PM Originally posted by questionlp As LRSeriesIII mentioned, the cache in the P2/P3 Xeon processors were not on-die and it was structured differently (the cache was L2 rather than L3 I thought i read somewhere that the "fake" pIII Xeon's on the 133 FSB (that couldn't do more than 2-way SMP and only had 256k cache) had that cache on die. From what i read they where just glorified CuMine's. by anatolli Windows NT Server (enterprise?), win2k adv server and datacenter can use PAE. We have 9 NT4 Enterprise servers here, and according to everything i have read from Microsoft they will only "use" 4Gb of RAM even with the /PAE enabled.

2CPU.com » News » November 2002 » Intel Ships Gallatin-based P4 Xeon MPs