2CPU.com » News » May 2003 » Dell Considering Opteron

Near the end of a mostly-financial article over at MSN Kevin Rollins, Dell's Chief Operating Officer, mentions that Dell is giving AMD's Opteron a serious look.

"The issue for us on Opteron is to make sure there is enough industry momentum and customer demand so that when we get into it, we get a return," Rollins said.

An article at PCWorld focuses more on the possible Dell-AMD relationship:

"We are very interested, and we are looking," Dell said about the Hammer chip, during a keynote presentation. "There is not much more to say," Dell paused. " . . . in public."

Having a second tier-one vendor (IBM is already on board) would be a big win for AMD and a big change for Dell, who has used only Intel CPUs in the past.

jives BP6 User Posts: 2419 Joined: 2001-05-18

#22753 Posted on: 05/02/2003 12:14 AM There was talk from the Dell camp about Athlon based laptops...never came to be. Have to see if Dell can break away from Chipzilla.

HEMI BP6 User Posts: 2467 Joined: 2001-12-18

#22754 Posted on: 05/02/2003 12:28 AM Yeah, I'm taking the news with a grain of salt. I really would like to see Dell (and other manufacturers) start using more AMD CPUs, but Intel has a lot ouf clout. Unix is user-friendly; it's just picky about its friends.

Buub Old SMP codger Posts: 1276 Joined: 2002-11-19

#22755 Posted on: 05/02/2003 12:37 AM I'm taking it with a grain of salt as well, but... These statements might sound very non-committal, but they are actually the most positive statements Dell has ever made about AMD. I think it signals that there is some very real wheeling and dealing bubbling under the surface.

ReMeDy Old SMP codger Posts: 12521 Joined: 2001-02-10

#22756 Posted on: 05/02/2003 01:17 AM Well seeing as Dell has a large discount on products from Intel. I don't see what they would be gaining. The name of the game is too buy it for cheap and sell it for 10x more. That's how Dell became the biggest threat in the first place. With !ntel undercutting them in price for position, How can these guys with the Big O compete?:confused: |[size=1][url=http://forums.2cpu.com/showthread.php?s=&postid=279831#post279831][color=yellow]FAQ's[/url][/color][/size]|Heatware-Remedy{WcS}|Dual760|Beerology|

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

#22757 Posted on: 05/02/2003 05:11 AM Although Intel has used only Intel processors in their desktop, notebook and server line of products, but they used to use the DEC Alpha in one or two of their older SAN/storage controller devices. I guess at the time, the Alpha just had better performance and I/O capabilities than anything Intel had available. [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

SnowGhost SMP Calaquendi Posts: 1569 Joined: 2000-05-10

#22758 Posted on: 05/02/2003 05:12 AM It wouldn't surprise me if this is just Dell say'ng, in an under handed way, "Intel, we want cheaper CPU's or we sell AMD CPU's. We want a further 10% discount." or some other form of incentive. TigerMP|2x MP2800+|1GB Crucial Reg RAM|MSI NX6800GT|Audigy 2 ZS|Insufficient storage|Antec SX1040 & SeaSonic S12-550w PSU

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

#22759 Posted on: 05/02/2003 05:20 AM If you're wondering where I got the Dell and Alpha bit... http://www.ap.dell.com/html/ap/nlc/ap/en/bsd/vectors_1999-nas.htm [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

anatolli Registered User Posts: 1885 Joined: 2001-07-18

#22760 Posted on: 05/02/2003 06:54 AM I'll believe it when the guys from up the street come in talking about them. I work in Round Rock Texas, (litterally right behind the Dell campus). anatolli Life's short and hard, like a body building elf

plympton Registered User Posts: 147 Joined: 2001-06-21

#22761 Posted on: 05/02/2003 09:21 PM Originally posted by SnowGhost It wouldn't surprise me if this is just Dell say'ng, in an under handed way, "Intel, we want cheaper CPU's or we sell AMD CPU's. We want a further 10% discount." or some other form of incentive. That's EXACTLY what they're doing. Leverage Opteron & Intel's paranoia to get cheaper processors. Dell doesn't give a lick about performance - they just want to make money (duh!). -Dan

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

#22762 Posted on: 05/02/2003 10:25 PM Two perfect examples of Dell not really caring about performance, mostly in the server side of things: - Dell PowerEdge 350: Still stuck with a Celeron or P3-850 on a 440GX motherboard (the server is a modified version of the Intel 1U boxed server) - No dual Xeon in 1U, even though HP/Compaq and other server manufacturers have been pimping them for months now. Of course, some people don't need the latest and greatest, but some do. [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

stmok SMP Veteran Posts: 490 Joined: 2001-06-25

#22763 Posted on: 05/02/2003 10:30 PM What do you guys/gals think about this... "SSE2 makes Opterons slower than Athlon XPs" http://www.theinquirer.net/?article=9278 Basically, do not turn SSE2 ON and you'll see the Opteron kick ass. 2x PIII 1Ghz (100Mhz FSB) + BX|2x PIII 933Mhz + i840|1x Celeron 500Mhz + BX for now

Manatoba Retired Comp-Tech Posts: 328 Joined: 2002-09-02

#22764 Posted on: 05/03/2003 08:07 AM You may find it's slower due to it being higher precision, more like Intel's upcoming SSE3, or may be down to compiler optimisations still more in tune with the P4's implementation, and also you would need to know if the tests were conducted using WindowsXP 64bit Edition or otherwise. I could well imagine it slowing in 32bit, but just think how appalling the score might be on an Itanium in 32bit mode... Xeon - Experiment: 2x SL6EP 2.4/400@3450+|PC-DL 1.04-B05 BIOS1009|2Gb GeIL Ultra-X PC3200 2-2-2-5| Gigabyte 6600GT Fanless Heatpipes|silent Samsung HD's|Antec Titan550|stock voltages|IWT+Agilent cooler| AMD - Experiment: Opteron165 1.8@2.25 CCBWE 0551WPMW|stock heatpipe HSF|HP/nVidia QuadroFX4500| Antec TPII 550W|Asus A8N32-SLI Deluxe|4Gb Corsair TwinX2048-4000PT|2x Dell 2405FPW's|more 2 come..|

Buub Old SMP codger Posts: 1276 Joined: 2002-11-19

#22765 Posted on: 05/03/2003 09:33 AM What do you guys/gals think about this... "SSE2 makes Opterons slower than Athlon XPs" http://www.theinquirer.net/?article=9278 Basically, do not turn SSE2 ON and you'll see the Opteron kick ass. Not sure to be honest. It's possible that if the code is too heavily optimized for the P4 maybe it hurts the Opteron. One thing I was thinking is the Hammer has more floating-point units than the P4 right? So if the code is optimized to only keep two units busy maybe that's leaving the Hammer's third unit twiddling its thumbs. But when you use standard floating point all of them have plenty of work to do. Just a theory...

Insp. Gadget Aspiring Duallie Posts: 89 Joined: 2003-04-17

#22766 Posted on: 05/03/2003 10:22 AM Unless the Opteron's SSE2 instruction decoding was inspired by the Itanium's x86 emulation , I suspect there's something fishy about TMPGEnc's SSE2 code. Personally, I'd like to see some tests with Lightwave (that has a lot of "mature" SSE2 code - TMPGEnc's support for SSE2 was only added a few months ago). Also, this kind of contradicts another article I've read about some companies compiling code on Xeons using Intel compilers (to get SSE2 optimizations) and then running them on Opterons (the result being faster than if they had compiled it directly on the Opterons, with a different compiler).

stmok SMP Veteran Posts: 490 Joined: 2001-06-25

#22767 Posted on: 05/03/2003 05:33 PM Originally posted by Insp. Gadget Unless the Opteron's SSE2 instruction decoding was inspired by the Itanium's x86 emulation , I suspect there's something fishy about TMPGEnc's SSE2 code. Personally, I'd like to see some tests with Lightwave (that has a lot of "mature" SSE2 code - TMPGEnc's support for SSE2 was only added a few months ago). Also, this kind of contradicts another article I've read about some companies compiling code on Xeons using Intel compilers (to get SSE2 optimizations) and then running them on Opterons (the result being faster than if they had compiled it directly on the Opterons, with a different compiler). Now that's interesting...Could it be the way AMD implements SSE2? (Or is there just ONE way to do it?) I guess we'll have to wait a few weeks (to months) on the full story of this SSE2 issue and Opterons... 2x PIII 1Ghz (100Mhz FSB) + BX|2x PIII 933Mhz + i840|1x Celeron 500Mhz + BX for now

Insp. Gadget Aspiring Duallie Posts: 89 Joined: 2003-04-17

#22768 Posted on: 05/03/2003 08:09 PM The way the processor handles things internally idepends on its design, naturally. All that's necessary to claim "SSE2 support" is for it to be able to understand the instructions and produce the right results. Internally, each SSE2 instruction is converted into a series of K8 instructions. Same goes for all other instructions, for that matter. But that's nothing new. The Athlon does the same, and so does the P4, I assume (the CPUs may "speak" x86, but they don't necessarily have to "think" in x86). SSE2 (and MMX, and other "multimedia extensions") don't suffer much from long pipelines (because predictions rarely fail, and the pipeline doesn't need to be flushed), so the P4's higher clock speed should give it an advantage. Given the Opteron's "low" clock speed, I don't expect it to be faster than the P4 / Xeon when running SSE2-only code. But, apart from synthetic benchmarks, and very specific routines, SSE2 code is usually mixed with "traditional" FP code, and there AMD's better FPUs should be able to give the P4 / Xeon a run for its money, even at half the clock speed. First because it's much faster (but hey, even the PIII's FPU is much faster than the P4's, in terms of IPC), second because the shorter pipeline means that flushes are a lot less expensive. I suspect TMPGEnc's motion search algorithms rely a lot on traditional FP, so at the highest quality the Opteron will probably be closer to the P4 / Xeon than at the lowest quality, where less time is spent on motion search and more time is spent on delta / DCT (which are probably the areas where SSE2 is used). This is pure speculation, of course (I've never seen TMPGEnc's code and I don't know enough about MPEG compression to understand it even if I saw it), but it would be interesting to compare the relative performances of an Opteron and a Xeon using "Highest quality" and "Lowest quality" (or "Motion estimate"), leaving all other settings untouched. There are two differences between the Athlon and the Opteron (longer pipeline and SOI) that are there basically to let the Opteron run at a higher clock speeds. So it's a bit strange that it was released at 1.4 GHz. It could be due to manufacturing problems (likely) or it could have something to do with power consumption (less likely). Either way, now that IBM is involved, I suspect the Opteron / Athlon 64 will be running at least as fast as the Athlon XP (in MHz) in 3 or 4 months. Remember what happened with the P4...

Buub Old SMP codger Posts: 1276 Joined: 2002-11-19

#22769 Posted on: 05/03/2003 09:01 PM Well it's pretty obvious SOI was their problem. And because of it yields were poor. The reason I say it's obvious is: why ditch the UMC partnership to switch to IBM (I mean AMD and UMC were even working on a jointly-built fab)? Well because IBM essentially invented SOI and is the only company really using it in mass production. And because of the delayed GeForce FX, which was originally supposed to be SOI, but nVidia eventually scrapped because TSMC couldn't make it work with acceptable yields (proving that SOI is harder than it looks). However two things look good for AMD: First, even without IBMs help they had SOI limping along, so they were close to making it work. Second, with IBM's help I'm sure all the remaining problems have been rectified and they should be able to ramp production easily. And of course there's the added benefit of IBM producing Opterons if demand starts to exceed supply.

2CPU.com » News » May 2003 » Dell Considering Opteron