Forum OpenACS Q&A: Development machine recommendations sought...
Two years ago, on a recommendation, I based my machine on a SuperMicro
P6DBE motherboard. Boy was that a recommendation I have learned to
love over the years. It's been a rock solid machine from day one, and
if it had a 133 Mhz bus and supported much faster CPUs, I would be
upgrading it today.
So what's a reasonably wonderful motherboard and CPU combination these
days? I am looking to piece together a machine for say 2K. I value
good (not necessarily great) performance, stability, and support. I
would like a motherboard that enables upgrades to technology expected
down the road. As an example, the machine I bought two years ago, I
filled with two 500Mhz CPUs. I recall at the time, 700 or 800Mhz CPUs
were the very expensive top end, and two 500Mhz CPUs could be had for
much less than one 800Mhz CPU.
This machine should be well qualified for running Linux, Windows, and
Secondarily, what's the name of a good hardware reseller with
reasonable return policies and knowledgable techs?
As an alternative, I could actually upgrade this board to two 800Mhz
CPUs of the 100Mhz bus flavor. It appears such an upgrade of CPUs
would be about $300. A very nicely priced upgrade. Aside from just
raw HP, is there any real value in a two year newer system?
My biggest hesitation has been in selecting which laptop would work on Redhat 7.2. There is a good website on this, something like laptop-on-linux.org or something, but it's a laborious process figuring out what to get.
I also haven't worked out what to do with CVS, etc... to coordinate my development with the production servers...
Actually, the raw CPU speed may not matter as much as the higher memory bandwidth of DDR RAM, which would serve to make everything faster overall. I think that PC2100 DDR gets you about 1.4GB/second of memory bandwidth - pretty zippy.
That's what I do, and it works great on an 800 MHz laptop. Of course, you probably want to load it with around 512 MB RAM, but you would anyway, wouldn't you? :)
Jade knows my lifestyle ...
VMware's great, though I run it the other way (windows under linux).
What kind of disks do you have, what kind of work are you doing, and what makes your machine feel slow to you? Or do you just lust after a new machine (I do! I do! I *always do!)?
I'm asking because depending on where your bottlenecks are, upgrading to 15K RPM SCSI disks might do more to boost performance than anything if you're still running on old 7200 or 10K RPM disks.
Last time I looked (about two months ago) the best Athlon motherboard out there looked to be one of the Tyan dual-processor 760MP chipset based boards. The biggie with onboard SCSI and dual NICs takes a special power supply with a non-ATX connector, though, so one of the more modest dual-processor mobos might be a better choice. By now ASUS or one of the other majors has probably added to the selection list. Furfly has a dual-processor Athlon system Mike built that apparently has been extremely solid thus far.
You might also look into one of SuperMicro's ServerWorks boards. While they only take registered ECC PC100 or PC133 DIMMs (which means you might or might not be able to use your existing DIMMs depending on what you bought) they're two-way interleaved which will provide better performance if memory is one of your bottlenecks. If they support the PIII Tualatin performance would be close to Athlon performance on a clock-by-clock basis. I'm speaking of the Tualatin with 512 MB L2 Cache, not the Tualatin Celeron, They're hard to find because they outperform the current P4s even worse than a plain old PIII with 256 MB L2 cache and Intel killed them almost as soon as they were released. The dual Supermicro ServerWorks boards may not support them because they require a different voltage level than previous PIIIs (this was the last gasp of the PIII line). As you state, supermicro boards tend to be rock solid. My downtown server is the same board you've got. I've got another single-processor board here that lost one of its IDE channels but everything that still works on the board is absolutely solid as a rock.
As you probably know, stay away from P4 and RDRAM for now. It's still not a good price/performance choice and won't be for some time until products are released compiled especially for it. And even longer for Linux since gcc doesn't really optimize for it and it's not clear when that will happen.
I've been toying with the idea of a laptop for a developmental machine. Sure, it wouldn't have the horsepower, especially the disk speed, but wouldn't it be a joy to work on projects in a coffee shop,
You're right, about the only thing I can think of that would be better, would be to work on my projects from "the airport cafe" overlooking the tarmac, or else to work on my projects from a van or something perched along Highway 1 one week, parked in a Tahoe ski resort the next week, and maybe even, parked down in the Guaymas harbor the following week.
Sigh, but that just has to be a fantasy, right?
I usually judge speed by Emacs keystrokes, project compiles, and mandatory Windows reboots. Don, I think you're right, there's a good chance that adding modern SCSI drives to my current machine is the best bang for the buck.
Any reason I shouldn't do this for my developmental computer? It seems like once Postgres and Aolserver are installed, there shouldn't be any difference, right? CVS should play well with Linux, shouldn't it? Anything I should be worried about?
Side benefit for OACS community: I can help maintain and test the OS X install docs. :)
I shuffled through a number of boards for my primary box, after UPS gave me the impetus by destroying my machine. I've settled on Epox - Tyan & ABit are as stable, but Epox gave me more on-board; maybe you'll see differently. Those VIA KT266A chipsets are nice, but I'm still leery of VIA after their DMA bug last year that trashed data. SiS opened up something cool with the 735; keep an eye on them.
Go with Athlons for CPUs -- they're 60% the price of P4s, clock for clock, and have a higher instruction dispatch per clock (AMD hired big chunks of the Alpha team after DEC went to Compaq). And more importantly for upgrade paths, AMD keeps the pinouts and electrical standards longer than Intel. Lately it seems like Intel intends for every upgrade to involve trashing both CPU and mainboard.
Buy SCSCI drives, cover them with fans. Almost every hardware failure of mine has been heat problems on drives.
Buy ATI for video if you dislike monkeying around with your kernel. Not that monkeying with kernels is bad (or I would have a big problem), but 'monkeying' usually entails 'distros do not include it'. My debian machine is riddled with little hand-configs and monkeyjobs I have done on it to do things not in the distro -- NVidia is one of those things.
As for a hardware reseller who isn't a pain in the ass, I'm having good luck with newegg.com. I haven't tried anything in terms of support or techs, though. I don't think they do that. For full machines, Talli mentions Penguin Computers. I've only worked with their rackmounts, but those were some very nice machines, for very cheap considering the competition. Might be worth looking at.
As for your last line (new CPUs)..... $300 can buy you a new motherboard and Athlon XP. Then reuse all your cards & such. But junk that 100Mhz fossil.
Get a 1U rack mount server, stuff it with 2 15K rpm SCSI disks using
hardware RAID 0. Get a nice LCD panel. Buy a medium sized travel
bag. Although it will blow your 2K budget.
Oh well just buy an Athlon XP coupled with a KT266A mboard (check
out tomshardware or anandtech which is the current top Athlon
boards). Stuff in 1GB of DDR, 512MB for Linux 512 for VMware
Windows. RAID IDE with 7200 rpm IDE disks. That should be pretty
Although Jade's suggestion is pretty tempting.
As I hinted in my response to Jade (we've drunk beer together, once over my laptop) I do work much of the time on my laptop.
But ... you still can't beat a machine with decent drives, ECC (for stability, my laptop has occassional mysterious Oracle problems that I don't see anywhere else, and I suspect the lack of error correction, and besides, I'm an old-fashioned old-fart).
Via's popular chipset for Athlon has tested out to be great, but I'm guessing Jerry's an SMP kinda guy. There's something cool about dual processor machines, though my only one is my webserver (which is vastly underutilized but which I love anyway).
I suspect Jerry's into the cool factor, as I would be, even though he agreed with my assessment that switching to 15K RPM SCSI would be the most cost-effective thing to do (for me, that has high coolness factor since I'm running 7200 and 5400 RPM ATA and SCSI drives).
DDR and RDRAM have been shown to provide surprisingly little performance improvement over SDRAM alternatives for most applications. You need to be doing a lot of memory shuffling and little computation.
Meanwhile Oracle is doing directly buffer-to-disk I/O bypassing the kernel buffer cache (at least Linux 2.4 has this feature) which diminishes the improvement in Oracle performance you can get with higher speed memory. In PG, increasing the buffer cache to reasonable levels means you'll reduce shuffling, too.
So ... it's not clear in practice how much better DDR will perform than SDRAM for the kind of systems we deal with.
Better, obviously, but the "bang for buck" issue isn't clear.
I do agree with the comment that Intel's forcing a lot of mobo and memory rebuys. I don't see it as being intentional, rather I see it as being a result of reactionary responses to the totally unexpected (by Intel and frankly everyone else) ability of AMD to not only challenge Intel on price/performance, but on raw performance. About two years ago Intel was thrown into a tizzy by the Athlon and (worse, from their POV) the roadmap, and realized that their slowdown on incremental improvements in order to concentrate on a far-off "next generation" left them vulnerable. So ... scramble, scramble, scramble and at the same time the rate of incompatible upgrades were accelerated. Most consumers don't upgrade, after all ... they buy a new box. And more importantly Intel had no choice from an engineering point of view.
Another reason was that this dual processor Supermicro motherboard was only $30.00 more than the equivalent single processor Supermicro motherboard. At the time, it seemed that putting a single processor into a dual processor board (that's the webserver configuration) was a cheap investment that yielded a pretty reasonable future performance upgrade extending the longevity of motherboard and system. That may or may not have been valid. At the time I thought it was valid, but since then I've heard several times that for many of these dual processor systems, you really need to get two processors made in the same batch, or else you can run into timing issues. So you can't just get another processor of the same speed as the first processor. That seems dubious to me, but has just the ring to it that means it might be true. If it is, then the future upgrade has to be two new processors, not just one. So is that truth or myth? Regardless, if the current difference in price between single and dual processor board is not much, I do think that getting a dual processor board and one processor is a good choice.
The third reason I had for going for a dual processor system was that two 500Mhz processors were more or less the same price as one 850 Mhz processor and theoretically could offer better performance under the right circumstances. I recall the 850Mhz was the top of the line then or very close to it, and I prefer to buy many things at a reasonable bang per buck part of the curve -- that keeps me from being disappointed a week after buying the 3Ghz machine when the 3.05Ghz machine comes out. (Offline, ask me how to buy a used car in terms of most bang for least buck.) I don't believe AMD was anywhere near competitive in terms of top end performance back then. That's not quite true. I spent a week with a dual processor AMD motherboard from ABIT, it was the first of its kind, and I believe it was an unauthorized, unstable hack that could be made work with enough spit. I didn't have enough spit so I refunded the thing and went with the Supermicro.
Coda: yesterday for $240 I ordered two 850Mhz processors, and wanting to be able to run VMware decently at the same time I am running Oracle and potentially NetBeans, I picked up a gig of registered ECC SDRAM for $200. Why registered? The board requires registered mem if you want to put a gig in. Okay, the joke is on me, of the three pieces of software I just mentioned (Virtualizer, RDBMS, Java IDE), which has the largest amount of memory suggested/required for decent performance. Yeah, the Java IDE which I recall suggests 512M, but many folks say it doesn't get honest until you have 768M. So for $450 I should have reasonable performance and I don't have to figure out how to recycle the old machine. SCSI drives may be in the future, but being a bit of a tightwad, I may opt for the latest in IDE controller and drive technology. (I have two or three 1G SCSI drives that well, I could buy a car for what those things cost me then.)
don't have to figure out how to recycle the old machine.</i></p>
<p>Too bad. The TiBook is really sweet, and you can use VPC to
run Windows on it.</p>
Now pricewatch tells me I can get a Ultra160 SCSI controller and 30G drive for about $335 and an Ultra SCSI-2 setup for about $260. Am I going to notice the difference between 160Mb/s vs 80Mb/s on my older development machine, or is my $70 better spent at the movies?
Sounds like you've got a good system started, if you don't mind could you please pass along the model of the motherboard and processors you chose. Also the source you used would be appreciated.
While I know systems integration and software, I must admit I'm a little weak on hardware specs.
The motherboard I am using is from Supermicro, and the two takeaways I have is that your motherboard investment is perhaps the most important part of your system, and once you've priced your motherboard, an inexpensive additional investment would be to get a dual processor motherboard. I used to get those Fry's specials: $49 mobo. But I have found the Supermicro motherboard (at $150 maybe?) to be absolutely rock solid, and almost as important, and this may just have been luck, it has remained a useful system for almost three years now.
But as to which motherboard, or which chipset you should get, well that's some other guy (maybe Don), not me.
About a year ago(?) I was thinking of building an ACS site where folks could just enumerate and detail specific motherboard/controller/vendor/application solutions and how well they have worked out. Sort of a geek/niche version of epinions. And then the dotcom market collapsed and my $49M funding vanished. Yeah, that's the ticket.
can you say the reason, why it was not possible to develope a new processesor(more than 3.2Ghz speed) without dual architecture?
please send the reason to my e-mail id: mailto:firstname.lastname@example.org
thank you sir.