Opteron system

Group
Well, I built my dual Opteron system.
MSI K8T Master2-FAR, MB
2x Opteron 246
2 gigs DDR
Quadro 980XGL
I haven't run any comparative SW benchmarks yet, but I did do some Mastercam
tests that involved a very complex 3D toolpath. This measures pure CPU
horsepower. The system I compared it against was a P4 3.2ghz, 1 gig ram,
Quadro 580XGL. Both systems had the same MC configuration with regards to
memory allocations to toolpaths etc.. Neither machine used more than 512mb
of ram. Graphics had nothing to do with the results.
The P4 generated the toolpath in 40 seconds. The Opteron did the same path
in 18 seconds. Now, before anyone gets too crazy, allot of this has to do
with th P4's inefficiency in running pure x86 code. Heavy floating point
tasks don't run well on the P4 unless the application is optimized for it.
Mastercam isn't, but it impressed the hell outa me anyway.
I did do couple of quick SW rebuilds of the same monster model on the two
machines. The Opteron still beat the P4. Not by 100% this time, closer to 25
or 30%.
A quick car analog. The P4 is like a Saab. Weeny little 4 banger turbo
tortured just this side of failure.
The Opteron or Athlon64-FX are like a Mustang Cobra.
Both cars are fast, but the Mustang is faster. The torque monster V8 doesn't
have to work nearly as hard, or spin as fast to get there either.
Regards
Mark
Reply to
MM
Loading thread data ...
Excellent news Mark. Great car analogy and technically very accurate.
Reply to
Eddy Hicks
;-) Great ! But what happens when you get to a corner or have to do the 'elk' test.
If there are any Swedes around they could explain it better, but the 'elk' test is tough and even Mercedes have been caught out by the 'elk' test . The A class rolled over to have its tummy scratched.
In jest, as I would never own a Saab; Architects seem to like them for some reason.
Jonathan Stedman
Reply to
jjs
Mark,
I am also about to buy components for an AMD box, and am a bit confused about which CPU to get.
I am leaning towards a single processor and wonder which would be better, an Athlon 64 FX-51 or an Opteron 148 (I think, but am not sure, that this can work in a single CPU system), each costs $733. Or, maybe an Athlon 64 3400+, at $417. Can you tell me the difference? (SWX workstation setup using CamWorks) I was planning on using the MSI K8T Master1X and an FX1100 video card.
Sincerely, Jerry Forcier
MM wrote:
Reply to
Jerry Forcier
"MM" wrote in news:FbXYb.25746$ snipped-for-privacy@newssvr27.news.prodigy.com:
I've never heard that the P4 is inefficient at x86 code. That makes no sense. I think you are thinking of the Itanium processor which is known to have problems with 32-bit code. The P4 is not Itanium.
Joel Moore
Reply to
Joel Moore
I'm about to build a new system myself. The Opteron 14x series is for single processor systems. The 24x series is for two processor systems. The 84x series is for 4-way and higher systems.
The Athlon FX chips are rebranded Opteron 14x series chips, and have the same requirements (dual channel ECC DDR400 memory, socket 940).
The Athlon64 3200+ and 3400+ are similar to the 14x series Opteron. They do not support multiple processors. But the memory requirements are different. The Athlon64 uses single channel non-ECC DDR400 memory. I don't know if ECC will work. Also, the Athlon64 uses a different socket (socket 754), so you'll need a different motherboard. I'm looking at the MSI K8T-Neo.
The Athlon64 3400+ is very close in performance to the AthlonFX-51, which corresponds to an Opteron 146. It's cheaper to build, and should give almost identical performance. The systems that need ECC RAM will be a little more stable because of the ECC RAM. Also, it seems that you have to be careful when selecting memory modules for Athlon64: some just won't work. Buy quality, preferably something approved by the motherboard maker.
Watch out for the Athlon64 3000+, it has only half the L2 cache of the higher rated chips. Good performance for the money though.
I haven't built my system yet, so there are no guarantees on my selections...
Reply to
Dale Dunn
Joel,
Nope, no mistake. When I said "pure x86 code" I was refering to the PIII and below instruction set. The P4 is very differnt from the PIII. So much so that applications written, and optimized, for the P4 won't run on a PIII. They added new instructions and changed the math core. Some of the changes were to make it cheaper to fabricate than the PIII.
SW has a special P4 optimized install that's transparent to the person performing it. Many programs, like, Mastercam, don't offer an optimized version. In these cases, depending on the tasks, the P4's performance is pretty dismal. For instance, that same test that took the P4 40 seconds, were completed on my 1Ghz PIII at home in 51 seconds. That's 1ghz vs 3.2ghz and only a 29% edge. Not exactly what you'd expect when you do the math. Mhz for Mhz the PIII was much more efficient than th P4. That honor now seems to belong to AMD.
Regards
Mark
Reply to
Mark Mossberg
Mhz for Mhz the PIII was much more
Not only more efficient, but gets more work done too, in most applications. Most importantly, in SW.
Reply to
Dale Dunn
Check out my benchmark posts of the last couple months running Athlon64 3200+. There is little difference with SW running on the Athlon64 3200+ vs the FX or Opteron. The Opteron is the fastest but at half the price I'll wait the extra two seconds :)
There is also PRECIOUS little difference between the Athlon64 3200+ and 3400+ and again, the price difference is considerable. $280 vs $404. If you're planning to do this again next year, go with the Athlon64 3200+ (Not FX).
- Eddy
Reply to
Eddy Hicks
Yes ECC works fine with Athlon64 but if you want 2Gb expect to get non-ecc for price and availability reasons. Also know that different mobo's will advertise up to 3 sticks of ddr400 and up to 3Gb total ram but the ones that can do that are underclocking the ram all the way down to ddr200. I know this for a fact and have reported it in other threads. Don't waste the time or money trying 3 ddr400 sticks in an Athlon64 socket754 mobo. Also, be aware the official spec from AMD on these cpu's is 2 ddr400 sticks and 2Gb total. I only use MSI mobo's and they comply 100% to the AMD spec.
- Eddy
Reply to
Eddy Hicks
Other facts Marks analogy have revealed:
1. Your Opteron chip will literally fall apart within 6 months. 2. You have to overclock it (install a smaller supercharger pulley) to be accepted at the local CAD meetings (or Pep Boys parking lots). 3. You are a young guy with a mullet, wear a miller lite cap while cadding it up and have a number sticker on the corner of your monitor and save files to the hard drive when people are nearby to hear them "spin up". 4. Or you are middle aged guy in a mid-life crisis and have CPU envy. You probably installed a clear lexan side panel on your computer case so everyone can see you have an Opteron. 5. Your Opteron is frequently passed up by turbo'ed 4 bangers also known as P4 Extreme Editions.
Muwahahaha.
Reply to
Jeff N
"Mark Mossberg" wrote in news:c13mvr$lmc$ snipped-for-privacy@www.omnica.com:
Wow. That blows my mind. Am I just out of touch or was this not talked about too much by the press?
Reply to
Joel Moore
The reviewers talked about it some. Traditionally, Intel processors have not been as fast, clock-for-clock, as their predecessors. Intel's philosophy has tended toward increasing clock speed to gain performance. The latest P4, Prescott, is a good example. It doesn't measure up with the previous P4 chips (find a review site to explain which version to buy and how to tell the difference if you want a P4). The Prescott has a collossal 30 stage pipeline, which allows the design to reach speeds much higher than todays 3GHz or so P4 (at 20 stages). To oversimplify, the long pipeline introduces inefficiencies which make the new chip a bit slower on average per clock cycle, but enable much higher clock frequencies. This is how Intel increases performance of their chips.
Reply to
Dale Dunn
Joel,
Most of what you read in the press is marketing BS. This goes for all companies AMD included. The only rag I give any credibility to is Computer Aided Design Report. They're funded by hefty annual subscriptions, and aren't in bed with anyone.
Once Intel broke the 3gig mark it became, in my opinion, a usable CAD chip. The clock speed rendered the ineficiencies less significant. We have a couple of P4 3.2's for SW, and we are quite happy with them. If an application is P4 optimized, It runs at about what you'd expect. I guess Intel figured that all software developers would just jump and rewrite their code. Well, this costs money, and not everyone's done it.
My problem has been finding an upgrade solution for our Mastercam users, who also use SW.
Regards
Mark
Reply to
MM
Jeff,
Sooo... You drive a Saab, do ya..??
Mark
Reply to
MM
*sigh*
I used to consider myself technically competent. In a way I still am, I suppose. I'm usually the one most people I know come to for PC repairs and I still prefer to assemble my own boxes rather than buying prefab (though that's no way to save money).
But it's clear to me that I'm slowly falling behind. Between the various chipsets, RAM modules, and CPUs, the complexity of the topics you need to understand fully to stay sharp in this field can be quite daunting for someone who hasn't got the time.
In some ways I miss the days of the simple, predictable, and relatively infrequent progression of CPU clock speed. The move to EDO RAM sort of complicated things a little (now I have to know what type of memory my MB supports), but at least you didn't have to worry about so many different FSB speeds and dual-channel pairing and etc.
But then again, when I'm sitting in front of my PC drooling over the luscious landscapes of Morrowind or charging into the Reichstag amidst a barrage of tank artillery in Call of Duty I tend to be thankful for today's state-of-the-art computers.
Reply to
Joel Moore
Nah. Just like making fun on Mustangs. I wouldn't be so biased, but the drivers in my area are asses.
Reply to
Jeff N

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.