Best Motherboards With Large RAM Capacity?

cortex writes "I routinely need to analyze large datasets (principally using Matlab). I recently 'upgraded' to 64-bit Vista so that I can access larger amounts of RAM. I know that various Linux distros have had 64-bit support for years. I also typically use Intel motherboards for their reliability, but currently Intel's desktop motherboards only support 8GB of RAM and their server motherboards are too expensive. Can anyone relate their experiences with working with Vista or Linux machines running with large RAM (>8GB)? What is the best motherboard (Intel or AMD) and OS combination for workstation applications in terms of cost and reliability?"

Comments

Tyan?

[therufus]

Have you looked into Tyan mainboards. They're more for the server market, which is really what you're aiming for.

Re:Tyan?

[arivanov]

I would concur. Tyan Opteron motherboards are probably the best choice for this. The only annoyance is that most of them are EATX and fit only in high end huge cases.

The other thing to do is to abandon Windows. Matlab behaves considerably better on Linux or Solaris than on Windows (especially on big data sets). Most Matlab users I know have long stopped trying to run it on Microsoft platforms. They are simply not fit for purpose. AFAIK Vista is no exemption. So if you really make a living off matlab you should move your other windows stuff onto a cheap and cheerfull small PC and switch the matlab monster to a "proper" OS. That is the way I have maintained it for my matlab users in the past and they have been happy with the arrangement.

Re:Tyan?

[Erpo]

If you're running Matlab on Linux, you'd better pick one version of one Linux distribution and make sure the version of Matlab you're using supports it. If you change distros or get updates, expect problems, like crashes when you multiply [1,0]*[1;0].

If you're a free software advocate, you could blame this on the mathworks for not providing the source to Matlab so that it can be endlessly tweaked and rebuilt to keep up with FOSS development.

If you've got any common sense, you can blame this on OSS developers who do things like making binary incompatible changes to libraries and doing nothing to make sure old programs don't accidentally load the new, incompatible libs. It doesn't make one iota of difference if it's "wrong" for a program to access errno directly! There are programs that depend on being able to do it, and taking away "extern int errno" breaks those programs (including Matlab).

Then there was that whole NPTL mess. *sigh*

Re:Tyan?

[infonography]

....The other thing to do is to abandon Windows. Matlab behaves considerably better on Linux or Solaris than on Windows (especially on big data sets). Most Matlab users I know have long stopped trying to run it on Microsoft platforms. They are simply not fit for purpose. AFAIK Vista is no exemption. So if you really make a living off matlab you should move your other windows stuff onto a cheap and cheerfull small PC and switch the matlab monster to a "proper" OS. That is the way I have maintained it for my matlab users in the past and they have been happy with the arrangement. I been craming quite a bit of ram in my Sunblade, the thing to remember is only a fool buys their Sun memory from SUN or a vendor, ebay has tons of it cheap at prices comparable to OEM PC memory prices.

Re:Tyan?

[krilli]

Are EATX boards really that big?

I got an Athlon MP board a while ago. As far as I can remember, it was EATX and fit into a regular old case. There wasn't a lot of space left, but it did fit. If I'm not mixing things up, EATX have the same mounting holes, and the extra board area just flows into normally unused areas of the case.

The listed size is deceiving - I am prepared to be wrong, but I urge you to check again.

Re:Tyan?

[cortana]

If I'm forced to choose between theads and some crappy proprietary application, I'll take my theads thanks!

Re:Tyan?

[Erpo]

Oh me too. But if I'm forced to choose between threads and Matlab, I'll take my Matlab. Especially if Matlab is the whole reason the computer is there in the first place.

Re:Tyan?

[arivanov]

1. Either that or have a good sysadmin. Based on what you are describing I would clearly blame that on the sysadmin not having a clue.

2. I have maintained matlab working in a Debian environment for 6 years in my previous job (along with plenty of other stuff) and never ever seen what you are describing. In fact the system started as linuxthreads (2.4) and moved to NPTL later (2.6).

3. As far as open source and so on, MatLab is extremely well behaved for something that is closed source has such a nasty licensing mechanism. It is written correctly to be used and distributed in a network according to the old school sysadmin cookbook (using centralised NFS shares). If you use it this way and not the default installation you are least likely to see any problems.

Re:Tyan?

[arivanov]

Depends on the case. Most high-end quiet cases like Antec Symphony will not fit an EATX board. You have to go for a proper EATX case. Considering that the MatLab box tends to sit under the desk of its primary user it has to be quiet so putting it into any noisy monster is not really an option.

Re:Tyan?

[TheThiefMaster]

An EATX board tends to be nearly square. There's no way it would fit in to a normal case.

There are quite a few big boards around now which are ATX with extra bits sticking out from the lower part to go behind/under drive bays, where there is space to spare in a typical case.

For AMD's more modern Opteron line, there are ATX and EATX boards available, and the difference tends to be that the ATX ones don't have a full set of ram slots attached to both cpu sockets. Some have no ram slots at all for one cpu. The EATX ones, on the other hand, let you use nearly ludicrous amounts of ram.

I have an ATX dual-Athlon MP machine. It used to be more powerful than my pc, but now it's more of a 200W+ room heater.

Re:Tyan?

[cortana]

It would be best if you bugged your vendor to get off their arses and *support their software*. Which is what you presumably pay them for in the first place...

glibc 2.3 has been around since 2001 (at least... that is the date it entered Debian unstable. It may have been released earlier).

Re:Tyan?

[plague3106]

Nothing like anecdote to pass on as evidence..

Re:Tyan?

[Erpo]

It would be best if you bugged your vendor to get off their arses and *support their software*. Which is what you presumably pay them for in the first place...

Actually it's been several years since I worked there, but I agree: a good vendor ought to support their software. If the stories I've heard are true, the people working in the lab after I left did try to get the vendor to responsibly support the software, but the vendor refused. The vendor decided to continue with the elitist, "we don't think you're doing things the proper way so we're going to intentionally break your applications," attitude. So the lab admins switched from Redhat to OS X.

Re:Tyan?

[jedidiah]

If enterprise database vendors can manage to support Linux, then some desktop application vendor shouldn't have any problem.

Things get deprecated in all manner of environments that coders have to deal with. Linux may be more annoying in this regard but it's hardly unique.

Nothing really forces you to alter a Linux installation once it's been deployed. Running a 5 or 8 year old copy of Linux doesn't quite have the same problems as doing the same for Windows. You can do the same with MacOS or Solaris too (safely run a 5 year old copy of the OS).

Re:Tyan?

[aybiss]

Especially considering that when you go buy that Tyan mobo you're probably going to get 8MB onboard graphics and no AGP/PCIE slots. ;-)

Tyan

[B5_geek]

Look no further then Tyan. The Tempest line (Intel CPUs) can hold 32GB of ram and the Thunder line (AMD CPUs) can hold 64GB of ram.

Now I am curious about one thing you said about Intel mobos:

and their server motherboards are too expensive If you are too cheap to buy a mobo that in your own words was "reliable, and solid", how the heck are you going to pay for the 32GB of ECC RAM?

I run a Tyan Thunder with two Opteron 270's (and 4GB of RAM) as my primary workstation, and I have never been happier. I can honestly say that this is the last workstation I will buy until it dies, I no longer need to worry about "but my computer can't run X".

With the memory sizes and data sets that you are talking about I wouldn't consider anything other then AMD CPU's. The bandwidth that the CPU and memory are shared on Intel boards, and each AMD cpu has a dedicated memory controller and dedicated RAM slots.

You posted this on /. so you know that Linux will be the preferred OS.

Go with AMD, you won't be disappointed.

Re:Tyan

[Dogtanian]

Go with AMD, you won't be disappointed. Yeah, but dont go with tyan, you will be disappointed Since you haven't bothered giving even a basic justification for this "opinion", and also that you're posting as an AC, I doubt many people will take this seriously. Probably a troll anyway.

Re:Tyan

[ZorbaTHut]

Funny thing about this, actually.

I jumped on the dual-processor bandwagon pretty much the instant that commodity CPUs officially supported it. Namely, the Athlon MP. I got a Tyan Tiger motherboard and a friend did the same. Shortly thereafter I lost contact with that friend.

A few years later I went to turn on my computer as usual and it wouldn't turn on. A bit of troubleshooting later and I realized that the PSU connector had burned itself into the motherboard power socket because something on the motherboard had randomly decided to short itself. Four of the pins had fried (in a distinctive pattern, see here and here) and I ended up buying a new motherboard from a different manufacturer and a new power supply (thankfully, the other components had survived fine.)

About a year after that I ran into my friend. We were talking about upgrades and I dug out those pictures. Turned out he'd lost three Tyan Tiger motherboards, with the exact same burn pattern, before changing manufacturers.

So, yeah, I'm not touching Tyan again. I've never actually had a computer component burn itself to death before, and one time was enough.

Re:Tyan

[cowbutt]

A few years later I went to turn on my computer as usual and it wouldn't turn on. A bit of troubleshooting later and I realized that the PSU connector had burned itself into the motherboard power socket because something on the motherboard had randomly decided to short itself. Four of the pins had fried (in a distinctive pattern, see here and here)

What make/model of PSU were you using? Looks to me that the PSU's power connector couldn't cope with the current the board was pulling. Of course, that may or may not be down to a fault on the board, but seeing as you haven't told us anything about the PSU, I'm betting you were using a cheapo one that came with the case you were using.

Re:Tyan

[phantomlord]

S2460?

I just went through the same thing in late October.I built the system in June, 2002. Had a problem with the initial power supply (Antec 430W) that came in my case not being big enough and I was getting random lockups. Switched to another power supply and everything was fine for years. I went away for a weekend and sshed in to read my email. On my way home, a friend called me and noticed that I wasn't logged onto AIM. Hmm, ssh was still up. I must have gotten disconnected and it didn't full reconnect right (gaim likes to do that from time to time).

So I get home and notice that my computer had rebooted rather than having a simple AIM disconnect. Odd, but whatever. A week goes by and no problems. I attribute it to a freak power problem since I notice my UPS battery isn't holding a charge. Suddenly, my computer reboots in the middle of playing nethack. Ok, that was weird. Power failure is set to turn the computer off, not reboot. Computer hangs before LILO runs. I reset it and it hangs at LILO again. Odd. So I take the side of the case off to make sure the CPU/GPU fans are spinning. No problems. I let it sit for a few minutes. Turn it on and everything is fine again.

I grumble about losing my nethack game and start anew. I get about 5 minutes in and my computer reboots again. This time, LILO starts loading Linux and the computer reboots before the image is uncompressed. It does the same thing again. I start smelling that aroma of burning electric and plastic. Ok, it's too late to deal with this. I power off for the night and decide to come back in the morning.

Same problem in the morning, as soon as the computer gets warm, it starts rebooting. Electrical smell is getting heavier. I start taking PCI cards and drives out to make sure they aren't causing a problem. I swap in a known working video card and that's the only thing connected to the motherboard. Same problem. Ok, maybe the power supply is flaking out. I go to disconnect it from the motherboard and it was stuck pretty good. Bad enough that I had to get out some pliers and really start yanking.

Exact same pattern as you, all the red (5V) connectors are burned out. Fearing the worst, I ordered parts for a new computer. Later that night, I decided to see how bad the problem really was. I took a scalpel and small finger drill and cleaned all the melted plastic out of the motherboard connector until I could plug my backup power supply in. Some quick testing showed everything worked, so I reassembled it and used it for a week until my new parts got here.

I think what happened, in my case, is that the UPS batteries went bad (I've since replaced them) and the minor power fluctuations caused something to burn out in the power supply, which in turn, affected the current it was providing. It's worth noting the rating on my power supply says it provides 42A across the 5V wires, so we're talking some significant juice to start with.

That said, it is interesting that we've all had the same failure result.

Re:Tyan

[Lord Byron II]

I've been using Tyan Thunders for the last five years. I've been very happy and had very few problems with them.

Re:Tyan

[toddestan]

I jumped on the dual-processor bandwagon pretty much the instant that commodity CPUs officially supported it. Namely, the Athlon MP.

Wandering off-topic, but you were kind of late for that bandwagon. Dual processors was supported by the original Pentium, though the Athlon MP might have been the first attempt by AMD to do so.

Re:Tyan

[kjs3]

And you'd be wrong, since he specifically said 64-bit versions.

Re:Tyan

[Fweeky]

Heh, I've seen that before. This was an Antec PSU connected to an EpoX motherboard (a single socket SoA) with a couple of doming capacitors. It melted part of the ATX connector into the motherboard socket (these are the 5v lines, same as yours); the board still worked, but it had some stability issues, and the 5v line reading from the on-board sensors was sagging badly (unsurprisingly; much of it's clearly being lost as heat).

I gave the board to a more electrically oriented friend; he cleaned up the socket, and still uses the board today. I don't use Antec PSU's any more, though.

Re:Tyan

[everphilski]

Yup, I was using 2 Intel 366 MHz Celeron processors (overclocked to 550MHz) back in high school, using an Abit BP6 motherboard. At $30 apiece for the processors and picking up the motherboard for $70 at a trade show, it wasn't a bad deal at the time for a gigahertz when most people were running less than half of that.

Re:Tyan

[Fweeky]

I'm running a Tyan Thunder K8WE with 8GB and a pair of 275's using FreeBSD. Excellent expansion, solid hardware, and well liked, though getting on a little bit; you might like to look at some of the newer Socket F options.

These boards aren't cheap, though; here in the UK you're looking at ~£250, which looks to be about the same as Xeon motherboards. You have specialist needs, suck it up.

Re:Tyan

[Skapare]

I have an old Tyan mobo in an AT (not ATX) case, in the garage gathering dust, that has a pair of 133 MHz Pentiums in it, and something like 32 MB of RAM. Yes, dual processors were around for quite a while. Maybe the OP mean "dual core".

Re:Tyan

[asdfghjklqwertyuiop]

Looks to me that the PSU's power connector couldn't cope with the current the board was pulling.

Well, obviously. The issue is why was the board drawing such an abormally large current through it in the first place.

Re:Tyan

[bhima]

I had one of those. One by one over the course of a few months all the power to fain headers on board died.

I used it for months with cover off the case and a box fan from home depot pointed at it.

I can't remember why I finally quit using it....

Re:Tyan

[everphilski]

Yup.. I had mine clocked to 450MHZ and used to read bp6.com.. but it should be noted that Celerons came out around Pentium II. There were Tyan boards that would allow the now fossilized non-MMX Pentium I to run in dual configuration.

Oh, yeah, I'd hang out on bp6.com as well. And you of course are right. I was working for a company that manufactured hospital software, working as a 'junior network administrator' (spent most of my day re-installing NT3.5/4.0 on broken/repurposed machines), and yes, we had a few servers that were dual Pentiums, you know, the processor on the card that plugged into the motherboard with the heatsink built in. I always thought that was a dumb design I'm glad Intel came around :)

Re:Tyan

[Bastardchyld]

To elaborate a little more on the previous answer. This memory limitation is not a limitation with xp/vista this is a memory limitation with 32 bit xp/vista. If you are using 64 bit xp/vista the 3+gb limit disappears.

-matt

Re:Tyan

[cowbutt]

It may not have been an 'abnormally large current' just one that was too much for the build specifications of the PSU in use. Not all PSUs are created equal.

Re:Tyan

[Bill, Shooter of Bul]

Also, this is not a limit on the server editions. As far back as win 2000 advanced server could 'use' more than 4 gigs of memory, but with a limit of 4 gigs per process. I'm not sure when linux go the ability to do that on 32 bit x86, but it was right around the same time. See wiki for more details

Re:Tyan

[phantomlord]

Bad capacitors were my initial suspicion, especially since the board is from that bad stolen capacitor formula era... but all the caps on the motherboard look fine(I checked then and just double checked now).

I just cracked open that power supply for fun and the two biggest caps are slightly bulging and 4 of the 5 medium sized caps are almost domed. The small and tiny caps all look fine.

I'm still leaning toward my problem being with the power supply, but there may also be some kind of defect in the board design which is triggering these faults in general.

Re:Tyan

[canuck57]

The pics were great. But that can happen with mobos that have a lot of power requirements and the connectors are "cheap" or seat poorly on the mobo pins. For example, the main 5v line might have 40 amps through it. If the surface contact area of the power supply contacts don't have enough surface areas touching the mobo contacts, it can and will overheat. That is the quality of the power supply connector and it's seating into the motherboard likely had a lot to do with this. But it is also why they went to multiple connectors for power on newer systems as that connector is pushed to it's design limit.

Re:Tyan

[ZorbaTHut]

High-end PSU (500w I believe, I forget the maker - I'm obviously not using it anymore and this was two entire computer upgrade cycles ago.)

Remember that it was working fine for years, and remember that the pins were actually [i]burnt[/i] - it's hard for an inadequate PSU to burn pins on the mobo if the mobo itself isn't helping quite a lot.

Re:Tyan

[ZorbaTHut]

Yep, S2460.

I like your diagnosis about the PSU. I'll have to dig out that old PSU (I still have it somewhere because the burnt connector is so awesome) and see if it has bad caps. But yeah, even if it was the PSU, the motherboard definitely contributed somehow - I've posted that picture quite often and the only people who've ever said "Hey, I had the exact same pattern!" were people with S2460s.

Re:Tyan

[5pp000]

Huh. I have a Tiger MPX (S2466) with dual Athlon MPs that's now over 4 years old and has never given me the slightest trouble, in 24/7 operation.

Re:Tyan

[Reziac]

I don't normally salvage below P200 components anymore (that being about the bottom end that's useful to me), but I make exceptions for interesting stuff like your garage collection, and for Tyan boards in general -- I've had wonderful luck with them (long-lived and extremely stable).

[thinking] 133MHz CPU probably means it's socket 5 and maxed out, but if by some chance it's socket 7, most can take a CPU up to P266 or K6-2/450. RAM probably maxes out at 128mb if it's purely 72pin SIMM, or 256mb if it's got DIMM sockets too.

Re:Tyan

[Reziac]

Every manufacturer has the occasional bad production run, or defective stuff from their own suppliers. First time I've heard of Tyan getting (possibly) bit that way, but maybe it was their turn on the hot seat.

I use Tyan boards by preference, and they've been nothing but stable and long-lived for me. My oldest is over 10 years old now, and I have two in 27/7 use that are now past 8 yrs. You're soaking in one of 'em. :)

Re:Tyan

[Bearhouse]

But a whole new whole of problems open up...like (no) driver support

Re:Tyan

[petermgreen]

Umm the intel processors built into a riser card were the pentium 2 and it's associated xeon variant and the early celeron and P3. Not the original pentium.

Re:Tyan

[petermgreen]

driver support for major PC components is pretty good afaict. Driver support for crappy consumer perhiperals isn't but most of the time people don't need to use those on thier heavy number crunching machines.

Re:Tyan

[TRS-80]

I have the same pattern http://trs80.ucc.asn.au/W5A00031.JPG http://trs80.ucc.asn.au/W5A00033.JPG from when I forgot to plug a molex into my 6600GT, so it was sucking all its power directly via the motherboard. Looking at your connector, it's a 20pin connector too, so I'd say you and I simply hit the physical current limit of the 20pin ATX connector. Server motherboards have been using 24pin connectors for years now, the S2460 must predate them or it's a design flaw that it didn't use one. Either way, these days server mobos (including Tyan's) generally come with an 8pin socket as well as a 24pin ATX so I doubt you'll see the same problem again.

FWIW I got a hardware friend to solder a new plug onto my PSU and socket onto the motherboard, and it worked fine after that, and I always remembered to check my video card had its direct power supply plugged in. What sort of power load did your motherboard have? Lots of PCI cards, a power-hungy AGP card?

Re:Tyan

[ZorbaTHut]

It was a gaming system, so we're talking high-end AGP card, probably a sound card, I think it had a TV-in card also. So, yeah, quite a bit of load.

I don't remember if it had an auxiliary plug or not - I think it did, and it would have been plugged in because I'm careful about stuff like that :D But I do know that I hadn't opened the case for a few months at that point.

Re:Tyan

[mr_mischief]

To be fair, some PC and server manufacturers put the CPU on a riser card in the 486 and Pentium days, but it wouldn't have come from Intel that way. I have a Dell 486dx33 with the processor and cache SRAM on a riser, for example.

Re:Tyan

[RedDirt]

So, yeah, I'm not touching Tyan again.

Me either. My first Tyan was an S2850 which lost both its ethernet ports (one was dead from the moment I bought it, the other failed slowly over eighteen months). I sent it in for service and they returned it to me claiming that they just had to re-flash the BIOS to make it work. I tried that, but whatever. I put the server back together again, fired it up and got nothing. No beeps, nada. Moved the CPU/RAM back to an old MSI server board and got CPU failure beeps. Grr. Swore off buying their stuff until about three months ago when I tried one of their S3850 boards - it was on sale for about $75 less than the SuperMicro H8SSL-i boards I had been buying. As far as I can tell, it's very, very picky about DRAM voltage. Had to exchange it for a SuperMicro because none of my Corsair or Crucial sticks would even POST in the thing.

I hear tell their boards are wonderful, but I sure don't see it. And their tech-support is awful, especially compared to the service I get from SuperMicro. No more chances from me.

Re:Tyan

[asdfgl]

This was a known problem with the Tyan Athlon MP boards. The problem was that the CPUs drew too much current via the 12V rails. One solution was to solder on an additional 12V connector on the motherboard, not very elegant. In my experience, this was a one off mistake by Tyan, and I still happily use their other motherboards...

Re:Tyan

[TheThiefMaster]

"500W" isn't necessarily high-end, there have been PC PSUs around for ages that quote "peak" power instead of "sustainable" power, with the former being twice the latter. IIRC any that is 80%+ efficient certified will be advertised based on its sustainable load, and any which advertise their colour, number of fans/huge size of fans, or glowing LEDs in huge letters on the box are likely to be advertising "peak" power, or about double what it can sustain. A good power supply will also generally advertise that it has "multiple 12V lines".

The most reliable test I've found is that a good PSU is really freaking HEAVY.

To summarise:
Big brand supplies like any Tagan or any Antec (particularly TruePower) = Good
qTec 650W Triple Fan Gold = Bad, can only sustain around 300-400W.

Re:Tyan

[Medieval_Thinker]

I have a BP6 motherboard sporting dual 500 MHz processors and 512 mb of ram that I built when all this equipment was new (1999 sometime I think). It is still in service and works pretty well given the age. When I built this machine, my main box was NT4 233 MHz and 32 mb of ram, so I was suddenly living very large.

Re:Tyan

[ivan256]

If you don't count specialist applications of the 80386, and you're limiting yourself to x86 architecture, you still missed the boat by 4 generations.

The original Pentium supported 8-way SMP.

As for Tyan, my previous two systems ran with Tyan boards. One was a Dual Pentium 200Mhz, and the other was a dual Athlon. The dual pentium still runs well today (and has been running well since 1997), and the Athlon probably works fine, but I don't power it up because I can't afford the electricity it sucks down. The board I replaced it with is a DFI that gets excellent reviews for reliability and features, but I find it severely lacking the stability and features the Tyan boards had.

They're not overclocker boards, but they are consistently reliable. I'm always bothered by people who will write off an entire brand of product because of one faulty model, especially when they don't even give the manufacturer a chance to correct the problem. It usually happens with hard drives (I'll never use "Brand X" drives. I had one crash! Those die for everybody!). Generally every manufacturer out there will put out a flawed product eventually when they're releasing dozens of newly designed products a year. That doesn't mean everything they make is trash...

Now if you give them an opportunity to fess up and fix the problem, and they don't step up.... Then you have a point.

Incidentally, after the huge batch of bad capacitors that entered the market in the late '90s/early 00's, I've had several machines that I'm responsible for either blow up (capacitor popped), or fry themselves; generally 4-5 years after they entered service. Sometimes they just turn off and never come back on, and sometimes the capacitor leaks in a way that yields more spectacular results. This is with both el-cheapo boards, and name brand boards from a variety of manufacturers. Products with thousands of components can fail in ways that send the blame in many directions...

I doubt the need for that much ram.

[djcapelis]

Is your working set honestly over 8GB? Your dataset might be extremely large... but I would think that for the most part you'd get along just fine with swapping out to a decently fast device and your working set would be considerably below 8GB.

Consider swapping to and from a flash device or a series of flash devices. That will get you better latency over a spindle. If you want bandwidth though, you'll need to go with a hard drive. I find it very unlikely even with matlab (bloated as it is) that you honestly will improve performance considerably with >8GB of physical memory... Then again, I have no idea how good Vista is at swapping these days. But they talked about ReadyBoost and all that, so I assume it doesn't suck at it completely. :)

If you really are worried about I/O performance, you should consider getting multiple chips (and cores, but mostly multiple chips) so you have more L1/L2 cache available to access. Though this assumes your applications are somewhat parallelizable...

Generally this question is a lot more complex than simply assuming throwing more ram in the box is going to be the best use of your money.

Re:I doubt the need for that much ram.

[arse maker]

Is your working set honestly over 8GB? Your dataset might be extremely large... but I would think that for the most part you'd get along just fine with swapping out to a decently fast device and your working set would be considerably below 8GB.

What on earth are you basing that on? Someone says they need/want 8gb plus but your argument is that he doesn't?

Re:I doubt the need for that much ram.

[jacquesm]

I love that attitude...

Some guy comes and asks an honest question. Then people go and tell him that can't be right and then go and give all kinds of suggestions taking into account that he isn't right.

Let's just for a second assume that the OP has a dataset that large. I can easily imagine it:

- complicated physics model
- computational biology problem
- datamining

and any one of a thousand other not so trivial computational problems.

If his 'luck' is the problem is not trivially parallelizable (I hope that's spelled right) then he's got two choices:

1) try to set up some kind of pipeline
2) get a single machine that can handle all the data

Apparently he has chosen for door #2 because that seems to be just about feasible.

There are some top of the line dell machines that will hold up to 128G of ram, the R900 series.

Re:I doubt the need for that much ram.

[16384]

Is your working set honestly over 8GB? Your dataset might be extremely large... but I would think that for the most part you'd get along just fine with swapping out to a decently fast device and your working set would be considerably below 8GB.

When doing computer simulations it's really easy to need that much RAM. I currently have 4 GB (2xQuad Xeons on a Tyan motherboard -- To the OP: Get Opterons instead if you can), but could sometimes use much more. Swap is not an options: When the memory hits the swap the performance simply drops to pathetic levels.

In computational physics you are in a constant struggle between the need for more accuracy and the limits of the machine.

Re:I doubt the need for that much ram.

[cnettel]

Flash might give you better latency than a HD. It's a very far shot from a proper DRAM memory controller, if only so due to the fact that it will be connected by the SATA bus. Hey, any data transfer from flash will need to go through DMA! If the problem is not inherently serial, latency is very expensive and even the hughest CPU cache might help little. The fact that he uses Matlab also (might) indicate that this prototyping, or at least "once-off" analysis. You don't want to optimize heavily for memory locality and VM-friendly prefetching if you can just throw more hardware at the problem and get a solution.

Re:I doubt the need for that much ram.

[dan dan the dna man]

Without you having any idea of what his dataset is, how can you suggest that he doesn't need that amount of RAM? I have a machine in one of my racks with 32GB RAM set aside for when the machines with 8GB RAM and however many GB of swap just don't hack it anymore with our datasets in MATLAB. I'm a bioinformatician/computational biologist, and compared to some of the sciences I don't think our datasets are 'large' but they sometimes certainly require large amounts of memory to process.

Re:I doubt the need for that much ram.

[Jeff DeMaagd]

If it's that kind of data, then it's really worth paying more for a solid workstation class board. And it almost assures you of ECC compatibility. ECC isn't necessary for home use and gaming, but if you have a need for 8GB+ of memory, then you probably should protect that data, and it's not terribly expensive either, in my opinion, last year's FB-DIMM pricing notwithstanding, but even that's very affordable now too.

Re:I doubt the need for that much ram.

[failedlogic]

I was about to ask something about ECC myself for a system with 4 GB RAM. I've seen the prices and shuddered as its being considered for SOHO use. Would ECC help prevent bit-flipping errors or would my money be better spent elsewhere when building a new system (e.g. better power supply)? I would think that higher or over-clocked memory could cause more 'errors' then would be accounted for from bit-flipping.

As a SOHO user, my understanding is that random bit flips while writing data can cause major problems down the road like when restoring from a backup if the original or backup file were bit-flipped at some time. I can't recall ever having a problem with a file I've restored or helped others restore so I know this to still be a rare instance in a SOHO environment. But my computer is quite an important tool (as I'm sure it is for most people) and holds many important files. Does it make sense to get ECC to help prevent bit-flipping in backups? I usually RAR and PAR my important files and copy using rsync (with error checking). And, similar to original parent's question, I also use large data sets with math programs at home (more for learning and nothing serious).

Re:I doubt the need for that much ram.

[Jeff DeMaagd]

ECC might not be that important for you. ECC memory only helps resist bit flipping while the data is in memory. It won't make your backups much more reliable as it's mostly the reliability of the medium, when backing up, the amount of time data is in memory during the transfer is very short. If you keep gigabytes of data in RAM for days at a time, or if the data is valuable, then ECC would be one step, in conjunction with mirrored or RAID-5 storage and off-line backups.

Re:I doubt the need for that much ram.

[failedlogic]

Thanks for the kind reply! :)

Re:I doubt the need for that much ram.

[jacquesm]

even the not-so-wealthy can have skills that allow them to ask difficult questions but may not have a corresponding budget.

I used to be in that position. Now I run my more interesting software on a 5 node dual opteron cluster (small for a cluster, I know... see that's those budget constraints again), each node has 8G of ram and 3TB storage. Before that it was 10 pentium machines at 600 Ghz (See http://clustercompute.com/ , which has inspired numerous people to build copies) and before that it was 10 pentium 225's (overclocked 200's :) ). What used to take weeks now takes at the most days. My applications are mostly in datamining, but I find computational biology to be very interesting.

You have to love it when people overcome their financial limitations with cleverness, why not give the guy a break and simply help him to solve his problem, starting out from the assumption that his problems and limitations are real.

It would have been nice to have a few more bits of information about the kind of data and the nature of the calculations, I'm pretty sure that 'cheap' is also relative but it seems that cheaper is better for this guy. How many people are at their most brilliant periods in their lives when they're also poor is not easy to figure out but I would not be surprised if it was the majority.

Re:I doubt the need for that much ram.

[TheSkyIsPurple]

Maybe he figured that most everyone else would answer his direct question, but he thought might have deeper insight into the problem?

I don't know how many times I've been focused on a problem for a long time, ventured down a solution path, and ended up asking for help for something complicated; only to have that guy ask me what I was thinking. When I explained the problem, it turns out I had missed something that drastically reduced it.

Sort of like the ol' America space pen vs. Russian Pencil story.

In other words, he was getting at the underlying concern, not the question asked. (think "Do I look fat?"... that's really not what they're asking)

Re:I doubt the need for that much ram.

[Detritus]

Sort of like the ol' America space pen vs. Russian Pencil story.

Which is a myth. Do you have any more pearls of wisdom?

Re:I doubt the need for that much ram.

[Glonoinha]

Well, that and right now even premium memory is crazy cheap - you can pick up 2G sticks of Kingston for $40 apiece at NewEgg, when two or three years ago we were paying twice the money for a fourth of the memory (like $80 for 512M). All of a sudden big boxes become viable options, upgrading to 8G of quality memory for less than $200.

Re:I doubt the need for that much ram.

[try_anything]

Is your working set honestly over 8GB? Your dataset might be extremely large... but I would think that for the most part you'd get along just fine with swapping out to a decently fast device and your working set would be considerably below 8GB.

...

more L1/L2 cache available to access. Though this assumes your applications are somewhat parallelizable...

That's a big assumption. Give the guy a break! Maybe he's just working on a problem where there's no known way to achieve predictable data access patterns. After all, not everyone doing math on computers is solving differential equations. When somebody says their working set is over 8GB and you make the jump all the way down to L1 and L2 cache, it's obvious that you are used to working on nicely behaved numerical problems. Not everybody is so lucky! And, indeed, a lot of heavy work goes into making those problems so "nice." Differential equations have been the center of the applied math world for over two hundred years, and they have important military and industrial applications. Centuries of brilliant mathematical work, massive investment, decades of clever programming, and all this for problems that naturally lend themselves to partitioning and parallelization anyway. The field is so mature that people who work on these kinds of problems get used to the idea that arbitrarily large datasets can be processed in arbitrarily small chunks just by using common sense and known techniques. In general, this assumption is much too optimistic. There are plenty of problems that are not so nice or not so well understood.

Re:I doubt the need for that much ram.

[djcapelis]

> Without you having any idea of what his dataset is, how can you suggest that he doesn't need that amount of RAM?

I certainly can suggest it.

I'm not telling the OP that they don't need more ram, I'm simply suggesting that they take a close look and ensure they actually do. Given the way the question has been written, this is the OP's first foray into getting this type of machine and the OP might just want to reconsider their original premise that this is what they need. Then I provided a series of suggestions on alternatives that might be suitable *if* the OP finds that they don't actually need this entire set of data in physical memory.

I don't see why everyone thinks I'm evil simply for suggesting that perhaps the OP's original question is flawed.

Re:I doubt the need for that much ram.

[djcapelis]

I think it's presumptuous of you to assume that I lack enough knowledge to provide a suggestion that he examine his I/O hierarchy closer to ensure that additional RAM will actually yield the performance he's after.

This isn't making fun of the questioner, nor even arguing with him.

Re:I doubt the need for that much ram.

[djcapelis]

It very well might be.

But it also might be that he's a fellow like you who simply is after more RAM so he can go on slashdot and talk about how much he loves having a machine with 16GB of ram in it.

I find that the latter is far more likely among the folks that aren't willing to purchase a server board when they're buying a new system...

Re:I doubt the need for that much ram.

[djcapelis]

Where did I tell him he can't be right? I told him he *might* not be right.

I then explained that he might find better performance improvements by addressing other portions of his memory hierarchy first. Just because you seem to assume the only place for improvement is in the expansion of RAM capacity doesn't mean that some people couldn't be better served with faster swap or more L1 instead. If his working set was a terrabyte then it's not likely to fit in any amount of RAM he's got in his budget and his best bet might be to improve his swap performance and use an operating system with very tuned read-ahead functionality for his type of application.

Re:I doubt the need for that much ram.

[djcapelis]

Absolutely. Flash is of limited use, but if he was having a capacity problem and a limited budget there are instances where it could significantly improve performance for random data access over an extremely large dataset.

Re:I doubt the need for that much ram.

[djcapelis]

I made no assumption about what his problem was. If his problem has no locality and it's working set is so large that it will never fit into RAM no matter how much he buys (consider a working set the size of a terrabyte, not easy to fit into ram even with a large budget) then improving his swap performance will be most helpful if he has very little locality. If he happens to have lots of locality (yes, this requires getting lucky) and if he happens to be running something that could be parallelized then on-chip cache would likely be even more valuable than any RAM improvement.

You clearly didn't understand what I was saying if you thought I was assuming that his problems had significant locality. I was encouraging the OP to explore other sections of his memory hierarchy as people tend to neglect to plan for any level but RAM when buying new equipment.

Re:I doubt the need for that much ram.

[TheSkyIsPurple]

I wasn't meaning to represent it as the truth. I assumed everyone here already knew. (Otherwise I would have actually told the story)

The moral of the story was the point.

Re:I doubt the need for that much ram.

[jacquesm]

Ok, I think I see where that went wrong, I think you missed the 'matlab' bit. That pretty much implies matrix manipulation or something close to it. Matlab is afaik not 'clusterable' so if he needs more speed the only way he's going to get it is by using a 64 bit box and oodles of ram.

As soon as you hit the swap it's game over, suddenly your run of one day can be a run of several weeks or more.

Another option would be to get rid of Matlab and learn how to really program the problem but for many people that are trying to do stuff like this that's too big a hurdle and they try to stretch the tools as far as they can go.

Re:I doubt the need for that much ram.

[try_anything]

You're right. I hadn't considered the possibility that his working set might be much bigger than 8 GB. Any time you can fit an appreciable amount of the working set into memory, but not all of it, it's a big win to add memory. It doesn't help to add memory when the working set already fits or when physical memory is very small compared to the working set. But, I still think we can give the original poster a little bit of credit and assume that he wouldn't be sweating gigabytes while working on a terabyte problem.

Re:I doubt the need for that much ram.

[tlhIngan]

Or it doesn't have to be some super-complex data set, either.

I recall doing some calculations using MATLAB, and MATLAB would do it (out of memory errors). When I hand-calculated how big the dataset would be, it turned out it needed around 4.5GB of memory. And given this was 7 years ago and a university-level assignment...

I reduced the dataset (basically, lowered the sampling rate) and managed to do the calculations in a more agreeable 400-500MB dataset. Still excessive, but hey, it worked.

Sometimes it doesn't take a very complex dataset to suddenly consume gobs of memory.

Re:I doubt the need for that much ram.

[bobcat7677]

He needs that much RAM because he is using Vista. If he used XP 64bit edition he wouldn't need as much. My $.02

Re:I doubt the need for that much ram.

[Stephen Samuel]

If all you're doing is playing Quake, then 8GB should be more than enough

If, on the other hand, you're doing 3D visualization of a 2Kx2Kx2K cube, then you're looking at 8GB per 'picture'. In a case like that, you'd probably want 16-24GB of ram so that you can keep your source, destination and some working data in RAM all at the same time. Once you start working with multi-dimensional information, it's really easy to get into a situation where you've got over 8GB worth of working set.

Re:I doubt the need for that much ram.

[Reziac]

Wow, that's clever. Do the motherboards/CPUs all have to match? if not, that looks like a good use for random outdated hardware. Love the "string 'em all on wires" approach, very creative. So what is its mission in life?

(I know nothing about beowulf clusters, so use small words :)

Re:I doubt the need for that much ram.

[Deliveranc3]

datamining

Since he doesn't want to buy high end commercial gear I think we can assume this one is the case.

For this increasingly common problem you want a LOT of commodity ram, are there any solutions that provide for this?

As far as performance storage goes isn't the ideal for price/performance a raid 0 of usb hubs and flash disks?

Re:I doubt the need for that much ram.

[jedidiah]

It also makes a dang fine personal writing instrument.

Sometimes a good quality product is what you really want. Not everything has to be velveeta.

Re:I doubt the need for that much ram.

[pionzypher]

I don't know how many times I've been focused on a problem for a long time, ventured down a solution path, and ended up asking for help for something complicated; only to have that guy ask me what I was thinking. When I explained the problem, it turns out I had missed something that drastically reduced it.

I believe that was his point. The moral of the story being that asking for the opinion of others can provide insight and direct you to things you may have overlooked.

Stop being a dick and obsessing over his example.

Re:I doubt the need for that much ram.

[jacquesm]

well, since he specified matlab I think the that datamining is out but you never know.

In olden times we used a trick called bankswitching and tons of ram to address memory beyond what you could address directly with the cpu. This was used to give 6502's up to 256K of ram (which was plenty expensive) and later using similar techniques to increase the memory in pc's.

The basic principle is really simple, you add an extra set of io mapped registers (on an x86, on most other architectures you'd have it all in memory space somewhere) that you can write to, then reserve a small window in your memory space that you will be using as your data transfer window. Based on the contents of the registers a selected block of memory is paged in to the window and then you can transfer data in and out.

Incidentally this is how most graphics cards work.

Re:I doubt the need for that much ram.

[jacquesm]

thank you :) It wasn't creativity out of choice, more out of necessity.

Motherboards / cpus can be anything you want but it's easier if they're all the same. After all as long as you plan your rods only in the standard pc stand-off holes they'll go right from the one to the next. It was a bit noisy though ! When it got switched off (somewhere in 2005) it had been in continuous operation for 5 years amazingly enough we never had a node crash in all that time, so the boot floppies (which were a total pain to make) got used only twice!

It's mission in life is (or rather was, because it has now been supplanted) to function as a test bed for a search engine and other massive parallel projects. I also used it to stress test switching hardware (with sometimes spectacular results).

Ever since I played a bit with transputers in the 80's I've been in love with parallel computers.

Re:I doubt the need for that much ram.

[Reziac]

Necessity and poverty drive the most practical inventions :) The outcome actually struck me as very professionally done... and that it worked so well for so long vouches for that.

I imagine if a person wanted to design ducted cooling, the noise could be significantly reduced. But if you're not living next to it, I don't suppose it matters!

Chipsets

[niceone]

To narrow things down a bit, it's not about Motherboards - it's about chipsets. I've only been looking at Intel (AMD don't have the performance right now for music stuff) - Intel's current P35 and X38 chipsets both support 8GB memory max. If you need more then you have to look at one of the Xeon chipsets: the 5000X workstation chipset is the one to look at if you want to be able to run 2 processors (not sure what the equivalent one is for a single processor) - it supports up to 32GB of memory.

Re:Chipsets

[niceone]

AMD don't have the performance right now for music stuff

What I meant to say there was Intel is ahead for music stuff right now. Last time I went with AMD (X2 4400+), but this time it looks like it will be an Intel (Q6600 probably).

Re:Chipsets

[thegrassyknowl]

Unless you're trying to parallel process stacks of tracks of audio and put effects on top of all of them then I doubt there is much difference between AMD and Intel for the same stuff.

Re:Chipsets

[downix]

Hm, for audio work I'd be grabbing a dual G5 setup and skipping the whole AMD/Intel thing entirely. 35GFlops per G5 vs 40Gflops for a single quad core Xeon, and you can pick one up for less.

Re:Chipsets

[Joe The Dragon]

the 5000x chip needs FB-DIMMS that cost more then DDR 2 ECC.

A dual cpu dual quad or dual dual-core system with 2 to 4 gb per cpu will cost less + you can get a board with the nforce pro chip set.

up to 32 GB DDR2 667/533/400 ECC ram + on board sas hardware raid also High-End PCI-e Graphics (SLI Supported)
http://www.supermicro.com/Aplus/motherboard/Opteron2000/MCP55/H8DA3-2.cfm

or this one

http://tyan.com/product_board_detail.aspx?pid=541

Re:Chipsets

[NeoThermic]

Not really. For video encoding or just outright FPU performance, the latest Xeon range will crush an Opteron. Even on price the Intel is cheaper; an Opteron 2222 (at £447.99 each) vs a faster Xeon E5345 (£300.09 each). Ok, so call me unfair for pitching a dual-core vs a quad core. The highest quad-core Opteron is the 2347, £275.70 each. Picking the same one at the price range Xeon side gives you the E5335 at £213.35 each, and it'd still be faster.

So, with Xeon you can have performance and cheap prices at the same time. I see no reason to go to AMD, do you?

NeoThermic

Re:Chipsets

[gazbo]

That's pretty much exactly what you do. And have you seen the CPU requirements of convolution reverb, or component-modeled effects/synths?

Best motherboards for 64-bit and large memory

[redstar427]

Standard motherboards are typically limited to 8 GB of ram, since they are designed for home users and gamers.
Server/workstation motherboards are the best solution at this time to go beyond this. Most people are only running 32-bit software, with 1-3 GB of ram, so it's not a problem for them.

Currently at work, I use a Tyan Tempest i5000XT (S2696) motherboard, with dual quad-core Intel Xeon cpu's, and 8 GB of ram. I will expand to 16 GB in 2008. This board can upgrade to 32 GB of ram, with 4 GB Dimms, which should be available sometime in the future.

I dual boot with 64-bit Fedora 8 Linux, and 64-bit Windows Vista Ultimate. I run Fedora 8 for all my productive work, and use VMWare with different versions of Linux and Windows, for testing and standard Windows work. I dual boot into 64-bit Vista Ultimate when I need Windows with direct hardware support for some multimedia apps and gaming. 64-bit Vista Ultimate seems a lot more compatible with current apps than 64-bit Windows XP Pro.

For my next home computer, I will choose a similar, but different Tyan Server/workstation motherboard.
The Tyan Tempest i5400PW (S5397) is also a dual socketed motherboard for dual quad-core Xeon cpus.
It has 16 memory sockets and can be expanded up to 128 GB of ram, with future dimms of 8 GB each.
I believe this is the best long-term solution for those that really need a lot of ram, at a reasonable price.
Even with just reasonable priced 2 GB dimms, it can hold 32 GB ram, which is a lot, even for large 64-bit apps.

While $450 for these motherboards is fairly expensive, they provide a lot of value, and good quality desktop motherboards cost $150-400, so it's not really that much more.

Battleship

[eddy]

>Consider swapping to and from a flash device or a series of flash devices.

Good performance. Gets expensive though. $7000 for nine Mtron 16GB Solid State Drives alone, then you need very high end RAID cards to cope with the throughput.

Re:Battleship

[bgat]

Indeed. For that kind of coin, just get enough RAM to eliminate the need to swap altogether instead. And with the money you still have left over, rent someone to babysit the dataset-crunching while you sip something cold on a sunny beach somewhere. (Posted from Midwest USA, where we've been below 32F for a while now!).

What's "music stuff"

[Joce640k]

I mean, if Microsoft thinks that doing all audio mixing in software is ok for gamers (Windows Vista) then it'd be interesting to know which audio tasks would bog down a multicore CPU.

Re:What's "music stuff"

[fbjon]

which audio tasks would bog down a multicore CPU Multiple tracks@24bit/44 or 96 khz, complex realtime effects and virtual synths, low latency.

Audio recording

[Bromskloss]

Ooo, music person!

You seem to be into audio recording. I'd like to build a computer for that purpose. Do you have any links (or direct information) about what I should keep in mind when I choose my components? I would really appreciate that!

(I intend to use Linux, if that makes any difference.)

Re:Audio recording

[gazbo]

I'm not the OP, but you may want to check out the Studio Central forums. Filled with equal measures of twats and great advice. You'll probably want to check out the DAW forums (digital audio workstaion, in case you've not come across that term before).

Don't expect to find very much about Linux though.

Re:Audio recording

[niceone]

I use XP for music stuff (linux for everything else)... I'll probably get an Intel DP35DP motherboard (pretty popular with DAW builders) and Core 2 quad Q6600 (best bang for buck). There's good advice to be had over at the SoundonSound forums (the PC Music board) and there's even a Linux section: http://www.soundonsound.com/forum/postlist.php?Board=LinMus . The other way to go is to look at what the pro DAW builders are using - www.adkproaudio.com are pretty well respected and seem to be on top of the issues.

Good luck!

simple solution

[drew_92123]

if you want that much memory stop looking at desktop boards and look at workstation/server boards.

Re:Check your app first

[hakr89]

Well you know, Matlab does run on linux as well, so depending on what limitations linux has, the examples you mention may just be reasons why you don't use windows for high end computing.

Your AMD Options

[this great guy]

All current socket AM2/AM2+ AMD processors (Opteron 1000 series, Phenom, Athlon X2, etc) support a maximum of four unbuffered DDR2 memory sticks. All current socket F AMD processors (Opteron 8000 and 2000 series) support a maximum of eight registered DDR2 memory sticks. (You can find this info in AMD's public datasheets).

As of today, unbuffered and registered DDR2 memory sticks of 4 GB or more are extremely expensive because the technology cannot be inexpensively mass-produced (yet). Only 2-GB DDR2 sticks can be found at reasonable prices.

For these financial and technical reasons, your are restricted to a total of 8 GB per socket AM2/AM2+ processor, or 16 GB per socket F processor. Therefore the cheapest option for an AMD mobo supporting more than 8 GB of memory is to buy a single socket F model. Newegg sells one for $136 (open box, though). Add a $180 Opteron 2212 processor, $240 for eight 2-GB sticks of registered DDR2-667, and you end up spending only $556 for a dual-core 2.0 GHz 16 GB barebone server assuming you have a chassis and a PSU lying around.

I'll leave other people comment on your Intel options. I am not very familiar with Intel server motherboards.

Re:Your AMD Options

[Fweeky]

Just be careful what you put on a ServerWorks HT1000 board, they have some nasty bugs that need to be worked around (Linux and Windows should be ok):

Implement a workaround of the datacorruption problem on serverworks HT1000 chipsets.
The HT1000 DMA engine seems to not always like 64K transfers and sometimes barfs data all over memory leading to instant chrash and burn. Somewhere there's a QA team which needs to be set on fire.

A different algorithm may be needed

[mangu]

Is your working set honestly over 8GB? Your dataset might be extremely large... but I would think that for the most part you'd get along just fine with swapping out to a decently fast device and your working set would be considerably below 8GB.

My thoughts exactly. When doing physics simulations, one often needs to manually optimize the code in order to use the cache correctly, so optimizing the swap shouldn't be such a problem.

Personal computers do not have support for more than 8 GB for a good reason, there isn't I/O capacity to use that much memory. There's no use having memory if you cannot transfer data to and from it.

However, the problem is that he uses Matlab. Perhaps he could get better performance using Octave with Atlas optimization, but in the end, only compiling in C with assembly language optimization will guarantee the best results. I have heard from several people that Matlab has problems when the data sets become large.

Re:A different algorithm may be needed

[EsbenMoseHansen]

However, the problem is that he uses Matlab. Perhaps he could get better performance using Octave with Atlas optimization, but in the end, only compiling in C with assembly language optimization will guarantee the best results. I have heard from several people that Matlab has problems when the data sets become large. Well, looking at the price list, switching to octave should buy him a good deal more hardware, even if the performance is the same :)

Re:A different algorithm may be needed

[try_anything]

Personal computers do not have support for more than 8 GB for a good reason, there isn't I/O capacity to use that much memory.

Sure there is, if you're using all 8GB over and over in an unpredictable pattern. You can't optimize memory access unless you can predict it. That means insight into the problem and insight into the data, combined with intelligent reorganization of the data, either in preprocessing or at run-time.

Given the information we have, as far as I can see you can only argue against needing more than 8GB by arguing that 1) it's implausible that his problem might be resistant to such optimizations, and 2) it's implausible that buying 16GB of RAM is a better solution than doing the R&D to discover and implement the optimizations.

Re:A different algorithm may be needed

[petermgreen]

He probablly already has the matlab license.

Also many matlab users are in universities, they won't be paying anything like list price for matlab and will probablly have a central pool of matlab licenses that anyone from the university can use.

Intel 5100 chipset, Tyan, and Supermicro

[MojoStan]

I also typically use Intel motherboards for their reliability, but currently Intel's desktop motherboards only support 8GB of RAM and their server motherboards are too expensive. Intel recently released their 5100 chipset for "value" 2-socket Xeon servers, which can use up to 32GB of "standard" DDR2 (not FB-DIMMs). Unfortunately, they haven't released an Intel-branded motherboard based on this chipset.

Tyan and Supermicro, which both focus on the server/workstation market, are the only motherboard makers I've heard about releasing motherboards based on the 5100 chipset. If you trust the Intel brand for reliability, then I think this Intel chipset on a Tyan or Supermicro motherboard might be a decent compromise.

• Tyan Tempest i5100X (S5375) - Seems to be out of stock everywhere (searched Google Products), but it's being listed at about $320 to $400. It's a server board, but it looks like it would make a decent workstation (PCIe x16 slot, integrated audio, extended ATX size).
• Supermicro X7DCL-i - In the same price range as the Tyan board, but seems to be available at a few online stores (like this). Standard ATX size, but lacks PCIe x16 slot.

Re:Intel 5100 chipset, Tyan, and Supermicro

[Wesley Felter]

That Supermicro board has only six slots; how do you get 32GB? The Tyan board only has eight slots, which would require quite expensive 4GB DIMMs to get 32GB.

I think a motherboard with 16 slots would be a better choice.

Server boards are cheaper than you think

[dbIII]

There are server boards without SCSI and a variety of other features - they'll be described as "bare bones" servers and can still support large amounts of memory. Supermicro stuff is good as is Iwill and a variety of others. I don't really understand why you want to run something that has unix versions on Vista - this is really a problem solved by having two machines; a low end server with something decent to run the software well and a display terminal running whatever you want. X-windows software is available on MS platforms and Macs so you'll be able to see the stuff as if it is running on your own machine only it will be running with less overhead. Server 2003 64bit is another option - I have not heard anything good at all about how Vista handles memory intensive applications.

Apple Mac Pro or XServe?

[foniksonik]

The latest Mac Pro supports 16 GB of RAM and the latest XServe (a better option IMHO) supports 32GB of RAM.

Mac Pro Specs

XServe Specs

XServe is a quad-core XEON 64bit at 3GHz as is the Mac Pro

They will both run Matlab w/ stunning execution.

Here's a nice case study for the XServe w/ Matlab: Induquímica Laboratorios

Re:Apple Mac Pro or XServe?

[WMD_88]

OWC is selling a 32GB RAM kit for the Mac Pro now, so it must support that much, even though Apple doesn't offer it out-of-box.

Re:Apple Mac Pro or XServe?

[petermgreen]

Apple don't offer the mac pro with 32GB of ram but by all accounts it handles it fine (which is not surprising given that it is essentially server class hardware in an unusual physical layout) and there is at least one vendor selling 4GB modules with suitable heatsinks (the mac pro has less agressive fan cooling than most server boxes so the heatsinks need to be bigger).

The mac pro is much cheaper than the xserver for basically the same hardware and probablly quieter too. Unless space is at a premium it seems like the more sensible choice. The mac pro is also availible with dual quad core processors whereas the xserve isn't.

Re:Apple Mac Pro or XServe?

[keeboo]

The latest Mac Pro supports 16 GB of RAM and the latest XServe (a better option IMHO) supports 32GB of RAM.

The guy asked about motherboards and OSes (citing Vista and Linux).
XServe-whatever supports 16 GB of RAM -- so what? Dozens of Dell, HP, IBM xyz beige-servers-inc etc machines do aswell but that's not what was asked about.

Go back sniffing Steve Jobs' ass.

Re:Apple Mac Pro or XServe?

[foniksonik]

Actually the guy asked about running Matlab data sets (large ones) and cited Vista and Linux, probably ignorant to the fact that there is an option with a high quality hardware supplier which also has excellent software for what he's doing (unless he left out a requirement of some other software that only runs on Vista and Linux ;-p)

If the guy is willing to use Vista, he's probably open to using OS X (both non-free OSes)

Maybe he's looking for an inexpensive upgrade to his current hardware, in which case you're half correct... not likely though, as the RAM itself is going to cost him several thousand dollars... but since he's looking I thought I'd mention these facts so that he and others like him could know that Apple has a solution at a decent price point (considering how much your going to spend on the RAM anyways).

Thanks for the comment though... it's nice to hear how the other half thinks.

Re:Check your app first

[ChronosWS]

No, if Matlab was compiled as a 32 bit app, the compiled code doesn't have the correct instructions for accessing a 64 bit memory space, no matter what the OS is capable of.

Traditional Flash Poor Choice for Swap

[DDumitru]

Flash devices are fast at small reads but very slow at small random writes. Swap is usually a 50-50 mix of reads and writes with no pattern to be seen. Thus if you swap to Flash, at least traditional Flash, then you will be very unhappy.

Your comment about bandwidth needing to go to a spindle seems strange as well. I have a 4 drive raid-5 flash array here that just tested at >400 MB/sec on reads and >150 MB/sec on random writes ...

      http://managedflash.com/news/papers/07-12-01_mtron-benchmarks.pdf

I don't know of many "spindles" that can keep up with that. Plus total power draw is 12 watts operating and 2 watts idle (actually the raid controller draws another 15, sorry).

If you want to see how Flash "can" be used for swap, see:

      http://managedflash.com/

Even compact flash cards are effective for swap yielding about 2000 4K random read/write IOPS.

ps: sorry for the advert. at least it was short.

Re:Traditional Flash Poor Choice for Swap

[Nutria]

Even compact flash cards are effective for swap yielding about 2000 4K random read/write IOPS.

But why spend the money on Flash RAM swap, when you can spend the money on more "regular" RAM, thus reducing the need for swap?

IMNSHO, flash RAM swap is the STUPIDEST idea ever in the 55(?) years of commercial computer hardware sales.

Re:Traditional Flash Poor Choice for Swap

[DDumitru]

Swap to Flash is "stupid", but the stupidest in 55 years. Surely you can come up with something worse.

Anyways, swap of any kind only make sense if adding memory is inconvenient. It all depends on whether system memory is even available at the size you need.

Right now, RAM prices are so low that flash as swap does not make much sense.

Tyan (Tiger S5197G2NR)

[FliesLikeABrick]

All of my favorite motherboards have been Tyans, lately the Tyan Tiger S5197G2NR. I now own ~10 Tyan-based machines, including some rackmount machines based on their 2U TA-26 barebones systems. I really can't think of any other brand I can recommend, but they've certainly got something to satisfy what you're looking for.

Re:Tyan -- welllll?

[Krishnoid]

... with the exact same burn pattern, before changing manufacturers --

-- (drum roll) to whom?

Re:Tyan -- welllll?

[ZorbaTHut]

Honestly I haven't stuck with one since. I think I've used Gigabyte, MSI, and a small family of VIA boards since. The MSI board turned out to have an incredibly weird defect (it would cause memory faults if it was under low load, so I used it for a while with distributed.net running in the background before I proved it was the mobo and not the RAM or CPU) and I was hesitant to buy Gigabyte because I had nothing but trouble with one of their video cards. Mobo works fine though.

Besides one with the cap problem, the VIA boards have been the most reliable computers I've ever had - I've got a screenshot of WinXP hitting 400 days uptime on one, and I only rebooted it because the hard drive was getting noisy and I needed to replace it (which led to some interesting issues itself.)

Dual CPU

[meehawl]

I jumped on the dual-processor bandwagon pretty much the instant that commodity CPUs officially supported it. Namely, the Athlon MP. I got a Tyan

Actually, I got a Tyan a while earlier (~1995) that supported dual Pentiums. That board ruled. Tyan kicks arse.

Re:Check your app first

[dezert_fox]

He's running Vista x64, the address space allocation isn't the same. The address space is gigantic (2^64 possible locations, rather than the 2^32 = 4 GB VM locations on a 32 bit machine). The OS doesn't have the same limitations.

Re:NCR

[mr_mischief]

Indeed. I nearly bought a dual 486dx2-66 (or a Pentium 60) instead of my dx4-100. The Intel line has had at least rudimentary support for multiprocessor installations since the 386. The boards used to do much of the work that the CPUs do now, though. Not a lot of people cared about dual processors in the PC world yet (other platforms had a head start). The boards were therefore pretty rare and pretty expensive.

The fact that NT 4, SCO, or a very immature (at the time) Linux would be my only OS choices on the dual pretty much solidified my choice of the single dx4. It gave the best bang for my buck for a Windows 95 box so I could play the games for that which NT 4 wouldn't. If Microsoft had a dual-processor capable OS with full DirectX support in 1995, I'd have had one of those dual 486s.

Re:NCR

[mr_mischief]

Actually, it was 1994 that I got that system. So it was mostly DOS games at the time. But there was press out about "Windows 4"...

Mobos with large ram

[softcoder]

Check out Dell Servers.
They come in towers. They use dual and quad 64 bit intel chips. They are reasonably priced for a server.
They have 64GB of RAM, and you can get them with supported RH Linux.