Is AMD Worth A Professional Reputation? 41
heyetv asks: "AMD has finally proven itself strong in competition to Intel. For over a year now. Old story; read TomsHardware or Sharky's. For overclockers, hackers, and the rest of us, this is great, but what about high volume, mission critical environments with hundreds, or even thousands of machines? What about high-performance clusters? I'm in a growing University/College Intel house of several hundred and trying to figure out why we are still as such. Are AMD's fast, cheap Athlon processors ready for production situations where a lack of support or seemingly minor failure could cost hundreds of thousands of dollars and reputations/careers? I'm sure some of you have rolled out Texas-style processors in large-scale corporate situations. Have you had positive or negative experiences in doing so? I'm not interested in a flamewar over which is the faster or more technically superior, but opinions on which one is a processor to base a professional reputation on and given that AMD has only been performing on par with Intel for a year now... is this long enough?"
Re:Crippled Karen (Score:3)
Anyway... good story, just possibly the wrong the place to post it.
--we are the music makers, and we are the dreamers of the dreams.
Do you trust Intel? (Score:2)
The real question about high availability situations are the rest of the system. Motherboards...power supplies... disk subsystems...etc. I don't think anyone is building real Athlon servers yet. I know Dell, Gateway, and Compaq don't yet. I wouldn't even consider putting in a clone in a situation like you describe.
we ship AMD (Score:1)
--
T-Birds FLY over PIII (Score:1)
I do think that mission critical information could be kept on a Athlon based system, when running the RIGHT OS(Linux or Solaris).
AMD works... (Score:3)
I'm not sure I understand the question. You're a College/University with severl hundred PCs? That's hardly a mission-critical environment. In fact, I would think it's the sort of environment where you owe it to the institution to look for the best bang for the buck.
But even in production environments, I haven't had any problems with AMD and would certainly trust them at least as much as Intel. If the environment is really that critical I probably wouldn't use either.
Sun and IBM both make rock-solid hardware (Sparc and RS/6000, not x86) and IBM at least has _incredible_ service contracts (Sun probably does too, but I have no personal experience with their service dept.). You pay _way_ to much money for that support, but in a truly mission-critical environment it's worth it.
But again, for a university environment, Athlon is definately the way to go.
Not quite ready (Score:5)
While the Athlon is a great chip, you can't get a 4-way SMP system with it yet. And most high-end boxes need SMP. They're so expensive to build that customers need a clear upgrade path - and if dropping in a second processor isn't easy, they don't want the solution.
I build LVS/HA clusters for a living, and one thing I don't think I can do without is the EMP, (Emergency Management Port) which isn't available on AMD motherboards.
The chip really is a smaller portion of the decision. When I build a cluster, I usually recommend an Intel L440GX motherboard, which has all the necessities like onboard dual SCSI, EEPro and EMP. Once you pick out the best motherboard for the lifecycle of the system, you look at the processors that it supports.
If AMD had a motherboard similar to the L440GX that supported SMP thunderbirds, they'd break in to that market. But they don't.
Their motherboards are designed for low-cost deployed workstations or gamers.
Really, motherboard choice is more important than chip choice if you're building LVS, HA, Beowulf, etc. PPC is an option, though.
AMD = cheap (Score:1)
For the gamer, an amd is the way to go. It's cheap, fast and get's the job done. Graphics designer? Spend the money on a P3.
To be tottal honest, its a matter of opion. Some people (like myself) like Intel. Other (*cough* cheaper) people like AMD. But don't forget that alot of stablility issues deal with the other hardware you have (motherboard, ram, video card etc..). You CAN'T blame everything on your cpu.
just my two cents
--we are the music makers, and we are the dreamers of the dreams.
VIA (Score:3)
What many seem to neglect to speak about is how quickly the 4x AGP and PC133 memory advantage is tossed out the window on the first crash. Who needs speed if that speed is unusable. Give my 2x agp and pc100 8 days a week if I can run my computer for more than 6 days a week.
Just MHO.
Trust AMD (Score:2)
I trust AMD enough to be stable for my needs; that my 300 is stable enough for my parents needs, and that my 1GHz AMD will be stable enough for my future needs. Admittedly, there were some problems with AGP cards with AMD chipset implementations (a lot of problems with 3rd party AGP implementations with the K6-X chips, and the GeForce cards on the ASUS K7M motherboard; problems, AFAIK, that have been worked out), but at a university, high-end graphics is not high on the priority list, at least from my background as a part-time computer technician at my University for 2 years (the CAD labs were the exception, of course).
Re:Do you trust Intel? (Score:2)
AMD has been in some disfavor with the reputable vendors because in the past machines that didn't have an Intel Inside sticker on the front were typically big old hunks of shit (well, cyrix even more so) the vendors who chose to save a few bucks by buying a non-Intel processor usually chose to cut costs on other parts too and the quality suffered. This, combined with a lack of SMP support by AMD really ruled them out of the high-end/business market.
If you have the time and are able to support the machines yourself I'd be the first to recommend AMD chips and hand building your machines. I suspect that you don't have those kind of resources so your choice of vendor is going to be much more important than the CPU that's inside.
_____________
Re:AMD = cheap (Score:2)
The article linked above is a fairly in-depth explanation of why Athlons are good servers (for some applications at least). Personally, since that configuration was put in place, I haven't seen Anandtech down or slow even once. Used to be crawling around major release time (GeForce, Athlon, that magnitude of news).
-----
#o#
Re:AMD works... (Score:1)
I do agree that saving money is of major importance, and that using AMD is the best bang for the buck. I just don't agree with your assesment of what a mission-critical environment constitutes.
Re:AMD works... (Score:2)
I didn't mean to imply that student labs aren't important, but the failure of a single machine isn't going to cost millions of dollars.
Put another way, would the students rather have 100 PCs that performed OK and had a 5% per year failure rate, or a 120 PCs that were somewhat faster but had a 10% per year failure rate? I bet the students would prefer the latter. They would have more PCs available on average and they would be faster too.
Ask a stock-broker on a trading floor the same question and he'd definately go with the more reliable, less performant setup. In his case a failure is really very expensive
Re:Pretty good story. (Score:1)
Nobody said good stories had to be tasteful. But I guess I should have figured out that it was plaguarized (sp?). It seems unlikely that somebody would take the time to write something like that just to post it anonymously to Slashdot. Oh well, time to look up alt.tasteless and see if there's anything else interesting there.
AMD, compatibility problems. (Score:1)
I bought an AMD Athlon 700. Then I tried to install RH 6.2 on it. Imagine my surprise when the kernal paniced when it tried to boot. Seems that RH 6.2 thought I had a PIII and was attempting to disable the CPUID.
The "solution" I was given? Build a new kernel (sorta difficult if it won't boot in the first place). So, I sent the Athlon back, and bought a PIII 800, which ended up being _cheaper_ than the Athlon (for _exactly_ the same system except MB and CPU) anyways.
AMD has neat CPUs, but the execution leaves much to be desired.
Yes, I do blame AMD for RH not working. At the very least, they should provide RH with example systems to use for testing. Software providers are frequently provided with free hardware for compatibility testing.
Jason PollockRe:AMD works... (Score:2)
TCO is a VERY big issue when you have hundreds of machines. Now, I agree that "mission critical" is a term usually reserved for much more failure-sensitive environments than a university but that doesn't mean that the cheapest box is automatically the best.
However, it may very well be that AMD boxes are as reliable, or more reliable than Intel -- if so you're lucky.
My daytime employer uses pure Intel for our x86 based servers -- Intel motherboard (with rare exceptions (820 chipset) they tend to be VERY reliable), Intel chip (with rare exceptions (original Pentium), very solid), Intel network cards (always great). However, for our server environment we are more memory bound than CPU bound -- we can afford to go with a slower (but more proven) processor and spend less money that way.
-JF
Worth your reputation? (Score:2)
But that doesn't make it worth your while staking everything. I've run into problems with AMD CPUs before, back in the 486 days. I haven't run into any with the Thunderbirds and Durons but who knows, they may exist.
So should you stake your reputation on Intel CPUs? Most definitely not. There have been far more problems with Intel CPUs recently compared to AMD CPUs. Far too many to risk your reputation.
Except, of course, you almost certainly won't be. If you go AMD and you run into any problems, your reputation is shot. If you go Intel and you run into problems, your reputation is likely as not unaffected.
Fear, Uncertainty, Doubt. 'Nobody ever got fired for going Intel'. (Of course, this used to be 'IBM').
Would I go AMD over Intel? Yes. Would I risk my reputation on it? No, though it is clearly the better (and safer) option theoretically.
What really matters (Score:3)
Fretting over one's CPU, if reliability is at stake, is a real waste of time. There are four high risk components in any computer.
Hard drives fail. Raid arrays and hot swap can reduce the danger here
Power supplies fail. Having redundant hot swappable power supplies are the only way to go
The CPU fan will eventually stop and you should have software monitoring this and reporting to you when it starts slowing down.
The OS is the most complex and error prone part of the system. It's very important to have a good one and very hard to find. Heck, that's why most of the people at this site are here. You won't find a slashdot site for power supplies or CPU fans.
If your hardware is going to fail the CPU is just as likely to blow as a network card or some RAM. It has no moving parts, just pushes electrons around the same way again and again and as long as its well cooled will give you no grief. As for CPU quirks, just about any CPU will have them. Once a CPU has had a few months to season it's bugs are either well known and the chip is avoided or fixed with a BIOS upgrade or OS patch (never to be an issue again). The Athlon is well seasoned and stable.
But honestly, if hundreds of thousands of dollars and people reputations are on the line there is no alternative but high availability clustering. None. Zero. Nada. It's even better if the nodes are in different time zones. So for Pete's sake, make a nice fast cluster of Athlon boxes with RAID 5 and three power supplies a piece running Linux or QNX. Then pat yourself on the back for a job well done.
Re:AMD, compatibility problems. (Score:2)
Re:AMD, compatibility problems. (Score:1)
its nice to know why it never worked,
thanks
hopefully RH7 will work (haven't tried it yet)
AMD =~ Intel, and has for ever (Score:4)
That is just plain incorrect.
AMD has tracked _every_ major device made by Intel for over 25 years. Back in the Plestocine era (1973-1975), Intel & AMD made a technology exchange agreement, wherin AMD got the masks to the 6104 (4Kbit DRAM) and Intel got the masks to the 2704 (4KBit EPROM)(I'm guessing here, anybody with better data is welcome to supply it). Later this deal grew to include the 8080/82xx uP/peripheral family and by the time the 8085 & 8086 came out they were solid partners in competition against Zilog and their Z80. You see, in those days, they had a thing called "second-sourcing", which meant that if you wanted to sell your microelectronic-based devices to the military, you had to establish at least two parts suppliers, so the DoD wouldn't be invested into a proprietary (or outright unavailable) part. The 8086 technology partnership was supposed to be "for the lifetime of the iAPX86 product family", which Intel decided ended with the 386. Since the 486, AMD has been forced to reverse-engineer Intel's CPUs, and has been generally drop-in compatible with Intel, except for occasional issues. Look at AMD & Intel's OEM price lists. THEY MAKE THE SAME CHIPS! (Many of which, like the 8051, you've never heard of.) Except that AMD usually has smaller dice and better yields, which translates to faster and cheaper parts. I guess AMD has drawn the line at licensing Intel's proprietary socket, and now they're no longer drop-in compatible. Intel has from time to time done other things to break AMD compatibility, but they catch up, and AMD usually offers comparable or better parts for less, both because they _have_to_, and because they don't spend millions of bucks for TV commercials with people dancing around in tinted bunny suits. That's what jars me the MOST: the unwashed masses now _know_ that Intel makes superior parts, because they've seen silly blue men advertize the PIII on TV, but they've never heard of that AMD outfit.
To build a High-Availability system, I would:
AVOID the bleeding-edge technology. It COSTS TOO MUCH, and has compatability and reliability issues. Anybody in chip manufacture can tell you that it takes a year to _really_ get a new chip really rolling off a line. Then they come out cheap.
Use AMD for a more reliable CPU, assuming that other factors (such as motherboard chipsets) are equal, which I gather from the discussion they may not be.
Spread the load out among a bunch of cheap machines if possible, rather than build a single expensive world-killer and single point of failure. If the job can't be spread among several machines, forget the x86, you need (or will need in the future) a bigger gun.
Think about this statement: "The Intel Pentium III processor will make the Internet COME ALIVE!!!" Now that's a blatent lie. I hope I'm addressing an audience that's well enough informed to know that _yer_connection_speed_ has one whole lot more to do with the quality of your Internet experience than your CPU speed.
Please guys, leave the engineering to the engineers, and quit wasting money on Intel, even if they _do_ have pretty graphics.
Re:VIA (Score:2)
You don't have to go whole hog (Score:1)
If things go badly, much better to find out with 40 machines then 400.
Re: Nice attempt at a plot ... (Score:1)
You must have missed the most important keyword in the story, ``Carbondale''.
Re:AMD, compatibility problems. (Score:1)
In the meantime, explicity enable the serial numbers by passing the kernel "x86_serial_nr=1" at boot time through lilo. Then make add an append line to linux.conf to add that permanently.
Re:AMD, compatibility problems. (Score:1)
Yes I do blame AMD for it not working with RH6.2 (it isn't a bug, since it isn't Linux's fault). They are supposed to produce a drop in clone. It isn't. They could have easily avoided the problem by providing hardware to the OS developers, but that obviously hasn't happened. Therefore, yes it is their fault. In this case, it has cost them a customer.
It is perfectly valid for Linux to panic on a SIGILL. You can't blame the OS developers if your "clone" equipment doesn't work, it's the clone's fault. Do you blame Creative Labs if an SB clone doesn't work with the standard driver? Nope. You don't blame Linux either.
Re:Worth your reputation? (Score:2)
"..virtually nobody has run into any problems with AMD CPUs, at least at the Athlon or better.."
Well I would like to put a word in as a site which has had problems. We currently run 100+ workstations for students in the UK which run RedHat Linux or Windows 2000 (roughly 50/50 split). A while ago we bought new machines with Intel 733 Pentium III processors which all run Windows 2000 just great. We also bought a AMD machine of similar spec (700 Mhz) which was flaky as hell! We run all our machines 24/7 and the intel boxes running windows were ok (not as good as the linux boxes which practically never crash), but the AMD box was rubbish crashing out 3 or 4 times a day! - after that experience we are going to use nothing but intel.
It makes our department look to bad if our machines aren't very reliable, we cannot risk using AMD chips as in our experince they just don't work. It seems a shame, but I wouldn't risk it when your reputation is at stake.
Re:Pretty good story. (Score:1)
Re:VIA (Score:1)
I run a VIA motherboard (the ABIT KA-7) in my home workstation, exclusively with Linux. I've been running it 24/7 for about 3 months, and I haven't seen it crash (well, except when playing with older Mesa drivers - but that's a separate, non-via issue I believe).
I don't run it particularly hard, it doesn't usually do cpu-intensive things, but I've had it running seti@home for 48 hours straight, ripping and encoding a few CDs in a row, compiling Mozilla etc without any problems.
The longest time I've kept it up for is over 3 weeks, and it was only switched off because I was going away for a few days.
So, I'm not exactly a demanding user, and this isn't a server application, but the stability of the system has yet to prove itself to be anything but perfect in my case.
/james.
Re:Pretty good story. (Score:1)
In the US, it's "plagiarized".
(that's a horse of a different colour)
Re:AMD, compatibility problems. (Score:1)
Mandrake had no problems with this and they are a Red Hat dirivitive. How is this NOT a Red Hat issue. Athlons had been out for a while before RH6.2 was released. If they didn't bother to test it with AMD then it is their problem not AMD's. I seriously doubt that Red Hat didn't have any Athlons to test with.
i440GX sux, go ServerWorks instead (Intel did!) (Score:2)
The Intel 440GX is way behind the curve. Intel has poured so much into the Rambus avenue they forgot about the high-end where 4GB RAM and 2 standard PCI busses don't cut it. And the MTH (memory translator hub) failed to produce the SDRAM alternative they needed with the i840.
Enter ServerWorks' [serverworks.com] (formerly Reliance Computer Corporation, RCC) ServerSet III chipsets. They product chipsets for the big-boys, now for mainboard OEMs like Tyan, Asus and SuperMicro. 2 to 3 PCI busses (1 or 2 are 64-bit x 66MHz -- NOT slots, but whole busses!), 2 to 4-way PC133 SDRAM (supports upto 16GB), DDR SDRAM on the way, just awesome. The massive PCI I/O blows anything Intel's got away, and meets or beats most RISC vendors. Cheap too as the 2 CPU, 2 PCI bus, 2-way PC133 bus ServerSet IIILE can be had for just over $250 in SuperMicro [supermicro.com] mainboards.
ServerWorks is so good, Intel has adopted their chipsets for their own branded mainboards. Again, check them out!
P.S. As far as AMD, stay _away_ from Gateway 2000 -- the cheapest/worst components. Stick with a vendor that builds quality AMD systems, with AMD-approved components. Try Micron PC as they just introduced systems based on the new DDR SDRAM AMD i760 chipset mainboards and PC266 CPUs. [micronpc.com]
-- Bryan "TheBS" Smith
Short answer... (Score:1)
Thank you,
Re:AMD works... (Score:1)
Walter
Re:Worth your reputation? (Score:1)
Make sure you have a decent power supply on those AMD machines. 300 Watts or higher. Make sure you are not running into heat problems (keep the CPU below 50C). Make sure you don't have faulty RAM (use MemTest86 or something similar).
Just some suggestions. My AMD machine is rock-solid.
Re:AMD works... (Score:2)
-JF
Re:What really matters (Score:1)
You forgot memory guy.
With no ECC corrective memory you will stay up even less.
-gc
Re:AMD works... (Score:2)
Contrast that with what you are saying and it's a weak comparison, at best. As a college student I think I'm well qualified in saying that public use/lab computers certainly aren't mission critical to my education, or anyone else's I know. When my internet access goes AWOL, I somehow still manage to educate myself by doing things like (god forbid) reading books. When my Sun Netray terminal dies for no apparent reason in the computer science labs, I do math homework. Certainly it's an annoyance but computers that crash periodically are of no real concern to me from an educational standpoint.
If you can't tell I'm squarely in the corner of the original poster. When the computers do work, which is most of the time, they are godawful slow. I would gladly trade a little bit more uptime from something to work on besides a PPro 200 running Slowlaris x86, a 486 with Windows, or one of 250 Sun Netray NCs all feeding into the same antiquated Sun E5K. Hooray for public schools, I guess
Silly Blue Men? (Score:1)
Bingeldac denies any responsibility for the
spelling and/or grammatical errors above.
Re:Do you trust Intel? (Score:1)
With my AMD Athlon Slot A processor, I my system has never run better. I have not even had Windows crash on me once yet. I have been running it for about 1 month with no problems. I would take AMD over an Intel any day. I will not buy, or look at a PC with a P in it. I hate work because we use PIII, the only computer that we have in the office that doesn't have any problems with the server. That might also be because it is not running Windows!
Do you want a chip that is made by a manufacture that had to recall a whole line of processors because they didn't listen to the problems they were having before? Or one that moved to Rambus memory and an ATX formfactor 2? You need to buy a whole new case for them, and spend at least $1,000 more for the PIIII! Why? You can get a chip that runs Faster, Cooler, and one that takes up less energy! Why pick anything else?
Re:Silly Blue Men? (Score:1)