My First Opteron :D

13

Comments

  • ThraxThrax 🐌 Austin, TX Icrontian
    edited August 2003
    Later AMD CPUs have no folding problems....
  • FAH_WWFAH_WW Training in Indianapolis, IN
    edited August 2003
    Urm - well yeah they do actually - Barton2800+ still dies every once in a while. Not often, mind you :D

    I'm sure the later Opterons will behave better - looks like the 240 is definitely the first one, anyone remember the P60/66 :D

    I've only got a few Intel-type machines here: Celeron 533s, 900 and a VIA C3-800 I seem to remember. All others worth mentioning are AMD chipped.

    Without getting involved in the AMD/Intel thing (I have a life (yeah rite) hehe) I've found both produce nice CPUs.

    A P4 on an 800 bus is fast. I'm thinking those nice Opterons will be faster though, in fairness. Especially if Vijay and his team deliver on what they once hinted at towards me, an Opteron-optimised core. I just cant wait for that to hit release :D

    What did you guys think about the machine, except for the drive rails bits, did you like how stable it felt right out of the box?

    I'm kinda tempted to get a single board like you guys and slap a 1xx Opteron in it. I'm sure it will be a stonker :D

    My PC is using 4*DDR333 registered (it's a duallie Opteron 240) on the MSI board.

    The Asus board was not available at the time :( but if I could afford to slap an 9800Pro in it I'd be well happy. Won't be able to do so for a bit tho - mortgage and servers need paying for :(
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited August 2003
    The opteron did feel quick, and responsive. We had a blue screen when Prime faffed the Promise raid manager installer (Heh), but that was his bad, not the system.

    The BIOS seemed very thorough, the OS showed no sluggishness, and yeah, it felt stable. Time will tell if it really is when we punish the hell out of it on monday.

    I really want to find out of the arithemetic scores were low because of the benchmark, or because of something in the BIOS. Low arithmetic scores means slow folding. Don't want that.
  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited August 2003
    Yeah, the thing felt very responsive once we got the chipset and video drivers installed. I am a little curious oncerned about chipset support and early bios revision though.

    I mean, the bios is v. 1001. I am sure some of these problems will be solved when they work out some of the kinks in the BIOS. The bios reports the chip as an AMD Engineering Sample. Also, the nForce drivers don't seem to be optimized yet. For example, there are no "nForce xxxxxx" drivers listed in device manager after installing the platform drivers. I get the sense that nothing is optimized yet. I will expect bigger and better after a few months when driver and bios are optimized.
  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited August 2003
    ...however, if you decide it's not worth the trouble I can PM my shipping address anytime. :vimp:
  • EnverexEnverex Worcester, UK Icrontian
    edited August 2003
    Templar said
    I bet he uses the machine for pr0n video editing. :(

    Or even better, Pr0n editing..... :respect:

    NS
  • Omega65Omega65 Philadelphia, Pa
    edited August 2003
    FAH_WW said
    Urm - well yeah they do actually - Barton2800+ still dies every once in a while. Not often, mind you :D

    You need better cooling on your Barton 2800+. I recommend any heatsink by Thermalright.

    My PC(s) fold 24/7 as do many on this forum.

    FAH1 - Barton 2500+ (1.83ghz) @ 2.4ghz Watercooled

    attachment.php?s=&postid=30911

    If your having FAH stability problems, it might be Temperature, PSU or memory related (Probably Temp). What are your computer specs?
  • mmonninmmonnin Centreville, VA
    edited August 2003
    No I can back WW up on the later AMD Athlon troubles. Hate to say it but AMD took some shortcuts when implementing SSE onto the later cores, esp TBred B cored. There have been crashes on normally stable sysems running stock speeds. Some WUs will not run on those CPUs but will others.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited August 2003
    Without being an engineer, I'd like to know how anyone is aware of what's causing the crashes. :rolleyes:
  • mmonninmmonnin Centreville, VA
    edited August 2003
    OK Thraxy, wait 3 years and I will be one. Major: Computer Engineering Technology. Getting the basics of transistors right now. Thinking about going back for IT degree when I am done, but thats 3 years down the road.

    Anyway I dont remember everything that was said but I think AMD didnt implement SSE to the exact specs made by Intel.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited August 2003
    AMD engineer.
  • Omega65Omega65 Philadelphia, Pa
    edited August 2003
    mmonnin said
    No I can back WW up on the later AMD Athlon troubles. Hate to say it but AMD took some shortcuts when implementing SSE onto the later cores, esp TBred B cored. There have been crashes on normally stable sysems running stock speeds. Some WUs will not run on those CPUs but will others.

    I've had FAH crash "seemingly" stable systems. Turned out that the Heatsink couldn't handle the temp spike when running FAH. Upgraded the cooler(s) and FAH ran normally. (2 Palominos, 6 Tbreds & 4 Bartons).

    When I build a system I use FAH and 3dMark2001 to test stability. And unless it passes both, a mistake has been made somewhere.
  • FAH_WWFAH_WW Training in Indianapolis, IN
    edited August 2003
    I've got 17 machines folding at home right now ;)

    There ARE problems with some of the cores. Temperature was definitely excluded, so was memory and motherboard.

    The systems were using Crucial LL memory, CPU temps are way lower than my Thunderbird core ones.

    The Thoro cores cant handle the higher temps as well. It's strange. I've only got the problem on the dual Opteron 240 system (and ONLY on some WUs, used to happen all the time till Pandegroup fixed something without announcing it GRR). The Opteron 240s I've got are very early batches probably, as they were pretty hard to get in the UK at the time.

    PSU is good quality and proper spec for the board. The system doesn't have a very high load beside the board, as it only has a 40GB 7200rpm drive, a DVD drive (hardly ever used, just for installs) and a nice set of fans to keep the 2U case cool :D

    Omega - I can see your point & I would agree that you're right most of the time. I would always recommend checking these things out first as well :D

    I might get some of those Zalmans for the Opteron, apparently they cool it a tad better.

    The number of lockups has gone down since Vijay and his team sorted out something. I am running 3.25, like you :D

    One thing I noticed however which worries me is that the GUI client can destabilise the system, even when minimised. I will be taking that up with Vijay, see what the reasons might be.

    Unfortunately watercooling is a little too risky for me right now (plus I cant afford it :( ) as I've got a big rack with 8 machines in it that needs to go in the loft. Watercooling that would need a webcam monitoring it plus a remote turnoff switch of some kind. I'd never leave a watercooled system running unattended for a week. That's coz I'm paranoid LOL :D
  • mmonninmmonnin Centreville, VA
    edited August 2003
    And thats the reason for the 1.48/1.49 core and 3.25 beta. To make AMDs run with extra 3dnow! optimizations instead of SSE.
  • Omega65Omega65 Philadelphia, Pa
    edited August 2003
    I would tend to blame the software instead of the CPU.....

    Now back to our regualrly scheduled Opteron Lovefest....

    Show us some Benchmarks Prime! ;D;D;D
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    The benchmarks are disheartening.

    Sandra arithmetic scores are lower than a 2.8GHz p4. Sciencemark is slower than a 3.2 p4, it only did 4 WUs in 3 days for folding.

    We updated to the 1002 beta BIOS and went through a LOT of **** (And PFM) to get it working, and it didn't improve anything except the PID in the BIOS and Windows.

    Prime, SMJ, and I believe that the Crush K8 chipset needs another iteration (IE, A1/A2/A3/C1 revision on the nForce2), needs new BIOS, better drivers.

    The system is responsive, but it's nowhere near as fast as the AMD chipsets are for the 2xx/4xx/8xx. It's just not optimized, not mature like the important enterprise-class chipsets had to be for release.

    The 1xx series of boards and chips just needs time.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited September 2003
    It's just not optimized, not mature like the important enterprise-class chipsets had to be for release.

    Aren't you guys working with an engineering sample?
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    Apparently not.

    Once we got BIOS 1002 installed, the chip's ID was correctly read as an Opteron 144.

    The version of the SK8N we were using is the newest retail board (v1.03).
  • edited September 2003
    Well I can't believe we did everything we could, there must have been something hidden in the BIOS that would have gotten the performance we expected.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    You know, though...nVidia's chipsets never work right the first time.

    A1 nForce2 - Never made it to market.
    A2 nForce2 - Made it to market, tends to fail after 3 months of use. A maximum of 180MHz FSB. Glitchy LAN.
    A3/C1 nForce2 - ULTRA400 part. Works like the chipset was designed for initially. It has incredible performance, no glitches. Yes, it comes with minor eccentricities, but those eccentricities when exploited are beneficial to us.

    Should we really have expected the C0 revision on the Crush K8 to be as good as the 2xx boards from AMD which are much more refined?

    Omega linked us to the fact in news articles that the CK8 chip isn't being massed produced. It has bugs, bad ones.

    I'm expecting the C2/C3 revisions to be the good ones.
  • FAH_WWFAH_WW Training in Indianapolis, IN
    edited September 2003
    I sure hope it will be a killer chipset. My experience with folding on the Opterons is that they're decent, but equivalent to a Barton at the Opteron clock speed, so 1.4GHz real for my Opteron 240 equates to roughly a 1.4GHz real clocked Barton. It was faster than a 1.4GHz Thunderbird by about the same amount as I'd expected.

    I reckon AMD should start at a 2GHz real clock for anything to make sense. Below that it's quite frankly a way to help Intel...

    As for the NF3 chippie, I reckon you guys are totally right there, first revisions tend to suck badly. I'd like to see ATI do a chipset for this as well, as they've got a P4 license I believe as well now. Would be cool to get some proper competition there. AMD does a decent job with their chipsets, but the real performance has always come from others, SIS/NV/VIA hehe :D

    I can't wait :D
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    Well, the thing is, existing Opterons on the AMD8000 chipset are better than a 3.2 p4C and faster than any athlon available.

    And supposedly the A64 FX is supposed to be even faster.

    This chipset left me disappointed.
  • Omega65Omega65 Philadelphia, Pa
    edited September 2003
    I expected that for unoptimized code a lot of 32bit proggies aren't going to see a performance jump clock for clock going from the Athlon XP to the Opteron - except for L2 Cache limited application benchmarks.

    The Opteron is going to need 64bit apps to shine brightly in the desktop arena.

    But in the server market with >4gb memory requirements, it's already making waves. AMD stock price crested the $11 level last week.....
  • BlackHawkBlackHawk Bible music connoisseur There's no place like 127.0.0.1 Icrontian
    edited September 2003
    FAH_WW said
    I'd like to see ATI do a chipset for this as well, as they've got a P4 license I believe as well now. Would be cool to get some proper competition there.
    The A3 and Radeon IGP320 chipset. For low budget boards and AFAIK they royally suck.
  • FAH_WWFAH_WW Training in Indianapolis, IN
    edited September 2003
    They sure do :D - just hope they will do a high performance one at some point. NV certainly has the lead there at the moment :D

    It's a shame NV didnt get this chipset totally right. I reckon the VIA chipset might actually be better (scary right). The AMD one is extremely stable, fast and relatively up to date. I just hope they continue developing it further, as the 760MPX was not developed any longer way before the dual Opterons were available. I know it must have been tough to see loads of people switching over to the Intels as a result. Can't blame them though, it's been too long to get to the Opterons.

    I'm quite pleased with my dual 240 system, although I must admit it's not going to be an Intel killer by any means, not even in 64 bit probably. The 244 and above should be very good. I have this feeling 64 bit AMD on Windows is being delayed for a reason. It may well be political... Which would be a total shame.

    I'd love to see FAH be the first true 64-bit optimised bit of code for the Opteron that's in pretty common use. If that gets the points flowing in, Opteron (and A64) sales will go up, just because sites like this one, IC and [H] to name a few will have loads of members that will want to try one out.

    If nothing else, for me it means fun times ahead. The new BIOS options are fun again, trying to tweak it out, pushing it to the limit. Someone will watercool it no doubt & get it to go faster :D

    What impressed me most was the stability and total lack of compatibility problems spotted so far. Very impressed with that.

    Do you guys reckon in a couple of revisions we'll have a killer support chipset for it besides the 8000?
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited September 2003
    It's a shame NV didnt get this chipset totally right. I reckon the VIA chipset might actually be better (scary right).

    Wow, now that just makes me want to rush out and buy an NF3 board. :eek2:

    Let's hope nVidia gets it right, and fast. We are all in trouble if Intel doesn't have long term, first rate competition.
  • FAH_WWFAH_WW Training in Indianapolis, IN
    edited September 2003
    VIA does make some decent chipsets - they forever seem to need an 'A' revision though LOL

    I'd hang in there, I'm pretty sure things will improve over the next few months :D
  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited September 2003
    Well my first impression is mixed: Performance is not what I had expected, but then we were warned of this months ago, in MediaMan's article about 64 bit computing: "I wouldn’t expect to switch on your brand new 64-bit processor computer the next morning and cry out “holy crap my old 32-bit computer I had yesterday is a piece of garbage!”

    My choice of the Opteron platform was primarly based on the bus speed (technically, there's no bus, I know.. I know..) Because of my customer's application - pro audio editing. In pro audio recording, people are constantly fighting with latency latency latency, and improvements in FSB are hailed as godsends. Well, going from 333 or even 400 fsb to hypertransport is a light year leap. I'm not too concerned with processor performance, as audio and video recording is not really that processor intensive. The opteron offered the platform that I could buy RIGHT NOW that fit the bill of a gigantic system bus. Hard drive performance is superb, which is what I'm going for. I needed a workstation board since his choice of video card, Matrox P750, requires an AGP slot, so the server offerings weren't valid. I am still happy with the platform. As FAH_WW stated, it does feel very very stable and respon sive.

    I just wish it would have folded me past t1rhino one last time :):ninja:
  • t1rhinot1rhino Toronto
    edited September 2003
    primesuspect said
    I just wish it would have folded me past t1rhino one last time :):ninja:

    It's not that powerful! :rolleyes2

    :D:fold:
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    I just wish the Crush K8 was as fast as the AMD8000.

    The Opteron on the AMD8000 DOES make me go “Holy crap my old 32-bit computer I had yesterday is a piece of garbage!”
Sign In or Register to comment.