Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Hardware

Dual Socket 370 Card for a Single Slot 1 MoBo 95

Vertigo1 writes "This is a dual Socket 370 card that will allow TWO (2) Socket 370 Celerons into a single Slot 1 motherboard. The pic is here.The company is QDI. From what their page say it requires no special motherboard, but I have yet to confirm it. I dunno if this is really on par with what usually is posted but to go along with your overclocking and cooling stuff I thought this was a nice addition! "
This discussion has been archived. No new comments can be posted.

Dual Socket 370 Card for a Single Slot 1 MoBo

Comments Filter:
  • Isn't it strange how Intel can have a run-away sales hit like the Celeron to do nothing but stamp it out for, "our protection."

    It is odd to think that if we were a small chips manufacturer that had a hit like the Celeron, it would be our bread-and-butter: Intel can afford to just change the line in order to keep people in their "correct processor class."

    I wonder what the margin is that they make on each type of processor? I would venture their net profit on two 370 Celerons is about 1/4 of the net profit they make on a Pentium III-- or maybe 1/6 of the profit on a Xeon III sale.


    AP

    (All quotes "made up.")
  • Like to see you fit 2 PPro's on that motherboard =)
  • by gavinhall ( 33 )
    Posted by Neothi:

    I don't know how much you all know about dual Celery systems, but bxboards.com gives some really interesting info on it. Of course, I would suggest as they do "don't try this at home kids."
    For those weaker of heart. ComputerNerd is offering dual Celeron systems that are 'supposed' to be overclocked.

    I think it is kind of tall, I mean it does have to fit 2 370 side by side, right?
  • When running on a real, 32-bit operating system, the TwinMagic's dual processor can increase the performance of the desktop PC.

    Now... what exactly do they call a real operating system? Linux ? :-)

    ---

  • I think you meant to say "...bigger than 2 CPUs..."



  • Absolutely true.

    I think the reasoning for placing the Power Supply in that location was to help with "fresh" airflow taken from outside the case by the power supply fan while directing it in the direction of the processor. Which in my ATX Mid case has the power supply about 1 inch over my pII chip.

    It would be time for a new case.
  • I don't know about Abit, but a company named Promise has them. Look here:

    http://www.aberdeeninc.com/abcatg/MB6601.htm

    I use this model at home with a 40 Gig RAID.
    Also, Promise is reported to be introducing a newer version that supports ATA66 that will be released sometime in Aug of 99.
  • This sounds immensely a dodgy hack.

    I think I'll go for the ABit BP-6, but I need to find another Celery 300a PPGA first :(
  • Okay lets insert some common sense here, the board is just about bigger then 2 CPUs, which IMHO is not really that tall.
  • >so forget about compatability, forget about case space, forget about this.

    I think you're being a little overly harsh.

    Sure, many people took a glance at it and thought it was a slocket you could plug into any motherboard. If that's what you wanted, indeed, that's not an option.

    However, a motherboard like this has real potential. With the single slot 1, the motherboard can be smaller -- the same size as a single processor system. While some cases may have power supplies in the way of a tall riser like this, others should work just fine. A single slot 1 is cheaper to build than 2, and a single large riser card is cheaper to build than two single CPU riser cards. So viewed as a possible dual-processor motherboard choice, it may be an attractive option. Heat is potentially an issue from having two CPUs so close together, but on the other hand it may make arranging a cooling system easier.
  • Would it be possible to have a Quad CPU on a Dual Board???
    I just bough an old Tyan Thunder II (66 Bus)... would be nice to put 4 Celeron 400 in it :-)
  • It looks like MTRR support may or may not be really hairy.

    I mean, you're installing two CPU's on a motherboard that's designed for one. Obviously there's no SMP support in the bios.

    Even some SMP bioses make the error of only configuring the MTRR registers on the 1st cpu. I'd be surprised if a uniprocessor bios tried to configure the MTRR registers on a second cpu.

    I can vouch for the fact that Linux SMP's just fine on Celerons. I'm using a pair of modified MSI MS-6905's and a pair of Celeron 366's in a dual slot 1 LX based motherboard.

    However, I can also tell you that MTRR problems can be ugly. My dual celeron refused to load X until i replaced the PCI video card with an AGP video card. For some reason the PCI video card wasn't being set up correctly. I could have fought with it, but i didn't think a 2 meg #9 Motion 771 was worth all that much trouble, and got a G200.

    This isn't to say it probably couldn't be worked around. They also may have figured out some way of tieing the MTRR's together. What I'm saying is, if there isn't a workaround in the hardware, or some kind of a bios update (unlikely), it probably won't work out of the box until the kernel is updated to match it.
  • Does it mention anywhere there a list of features of the motherboard? Does it mention the chipset? The IDE/ATA speed? PCI and AGP configuration? I find that strange if it was an entire motherboard.

    I beleive it will still require a motherboard with an APIC onboard, but then again, I didn't think the signal lines for both processors were on the same connectors. Infact, I kind of doubt there would be, as how would a processor know which one to take. My option is they found a way to fake these signals to the system and have their own method of running SMP....maybe with provided drivers. Then again, I have't seen the product itself, so what does my opinion have to do with anything?

    Dan Guisinger
    Quantum Warp Laboratories
    http://www.qwlsoft.com
    http://www.networkedcomputing.com
  • if you are going to say that you should say why you dont like it
  • Pay attention to ;)
  • I got dual 400 celerons on an epox board....if quad celerons is possible, contact me too :) (shaldannon@usa.net)


    Who am I?
    Why am here?
    Where is the chocolate?
  • I have been running dual S370 433's now for a while using the ms-6905 card, it allows running the celeron at 100mhz, voltage changing, and dual proccessors all with a couple jumper changes. Doing seti@home is more fun that way ;-)
  • Yeah, I'd think the motherboard the card would attach to would have to have bios support for 2 cpu anyway.
  • I know people that have compiled kernels on unprocessor machines in around 2 minutes... If it takes that long with 4 CPUs, each running at that speed, something is dreadfully wrong.

    - A.P.
    --


    "One World, One Web, One Program" - Microsoft Promotional Ad

  • by Anonymous Coward
    Nuff said.

  • Sigh

    Looks like you're right...

    Silly how small things can make you happy then not....

    Message on our company Intranet:
    "You have a sticker in your private area"
  • Yeah, it looks like someone shoved a PCI NIC into the slot. Sure it might not need a special motherboard but it may need a different type of case...;)
    Seriously, if this thing works with my Abit BH6 motherboard I would buy it in a second. I'm kinda leary about the compatibility issue though. There has got to be some sort of weird requirements for it.
  • by Anonymous Coward on Tuesday June 08, 1999 @06:02AM (#1860150)
    If you read the rleas a little more closely,
    that riser card with the 2 chips works only on their MB.
    so forget about compatability,
    forget about case space,
    forget about this.

    now the new abit dual 370 MB,
    and the one with built in IDE RAID,
    now those are interesting.

    this is just eye candy.
    (and how long is it gonna be until intel decides to cut this little easy dual celeron thing off?)
  • Heh, I knew someone else must have an LX...

    AFAIK the Celeron's won't work in an LX. In fact, I don't think much of anything will--I wouldn't even trust a 333 PII ;-) I finally broke down an bought a dual Celeron system--now I just need to DL Quake III...

    To keep this on topic: how exactly is better than a dual slot 1 with socket 370=>slot 1 adapters? It can't be any cheaper; the board may be (single slot and all) but the cost of the custom card should more than make up for that.

    I also get the impression that who ever wrote that product description doesn't speak English as their first language (or second, or at all...). They seem to be using "motherboard" to describe both the dual card and the real mobo.
  • But it's not going to be cheaper than the dual Socket 370 motherboards that are starting to emerge. And these are from established makers, not random fabs that are wholly owned by never-heard-of-them holding companies.

    -jwb

  • Posted by Dr Evil:

    You mean you'd rather go with a 'moderately dodgy hack' instead? ;-)
  • If you have a decent motherboard they should have come up with BIOS updates to support the celerons, and the dechutes(333mhz PII and up). I've installed a celeron on a Tyan LX motherboard with no problems.

  • I suggest you read it one more time.

    "Legend QDI, one of the world's largest PC motherboard manufacturers, today announced the
    launch of TwinMagic, the first dual processor motherboard on slot 1 interface"

    Here they are referring to the Twin Magic sockets to slot interface as a motherboard.

    "Users can simply plug into one Slot 1 or the motherboard's single socket to realize substantial gains in the power and performance of their desktop systems."

    This sounds to me like they are saying you can plug it into one slot on a muti slot MB or into the single slot on a standard MB... to increase performance on an existing system.
  • If someone does determine that this is possible, PLEASE contact me. The first thing I thought about was 4 400s. I have the same Tyan board w/2 Celeron 400s, and if I could get 2 more, I would be extremely happy (kernel compilation in 2 min)!!
  • It looks no bigger than a Xeon (not that Xeons are exactly small...)

    And of course this comes out right after I bought a shiny new Socket-370 mobo... :P


    --

  • sounds from the press release like you need a smp capable OS to use the riser card. a white paper or some more specs would be nice. would be a nice way to get a cheap smp server up.
  • Isn't it strange how Intel can have a run-away sales hit like the Celeron to do nothing but stamp it out for, "our protection."

    Ahh, the Microsoft model. Don't give the customers what they want; give them what you want them to have.

    I wonder what the margin is that they make on each type of processor? I would venture their net profit on two 370 Celerons is about 1/4 of the net profit they make on a Pentium III-- or maybe 1/6 of the profit on a Xeon III sale.

    I came across an article a while back (can't remember where) that said it cost Intel approx $65 to make a Celeron processor (margin's a little thin), vs about $70 for a Pentium III 500 (nice fat margin there). I would imagine the Xeon is more expensive to manufacture than these, but not that much more.

  • Save your money and use Linux's software RAID driver. I get much better performance, using both ide channels and one channel of a Promise ATA/33 ultra-ATA non-RAID card. And, sadly, under NT I find that somehow, with two different sets of drives and Promise FastTrak RAID cards, data corruption seems - universal - when running SMP. No problem with uniprocessor configurations, though. Or ever with thier nonRAID card.
  • I suspect those estimates forget that it's quite an investement for Intel to make the .25u or the upcoming .18u fabs.

    Guess which processor is paying for .18u and coppermine? Celeron or PIII?

    That way I see it is that the Celeron is just Intel's 'trick' to keep AMD from making profits...

    (Running a stable dual 300A/S370 @ 450 here, and happy with it).

  • Yes, I am afraid that this [www.qdi.ca] would interfere with my power supply.
    Oh, well. Guess I will have to skip this "hot new product" as it were.
  • Posted by Lord Kano-The Gangster Of Love:

    >>2) Multiple CPUs (under an SMP-capable OS) operate in PARALLEL. This means that 400 MHz + 400 MHz = 400 MHz. (Now, as for FPU performance, that's a different matter...but the overall clock speed is the same.) PLEASE stop talking about your 1 GHz machines!

    Thank you for saying this again. About 3 years or so ago when Daystar had their 4-way SMP Mac clone on the market (Genesis MP) there were arseholes bragging that at work they had "800 mhz machines".

    You can't take two one carot diamonds, squeeze them together and call it a two carot diamond. It doesn't work that way.

    >>3) YES, you need an SMP-capable OS (i.e. NOT Win9x) to use two processors. Just because they're plugged into a single slot doesn't mean that they automatically operate in parallel.

    Not necessarily. When the Genesis MP was released the MacOS (Still called "system software" then) was not SMP capable(Aside from X,it still isn't) but apps like Photoshop, some 3d rendering programs had plugins that allowed them to take advantage of the additional CPUs.

    I don't see why this can't be done on Win9X. I know that Carmack is working on SMP support for Quake3.

    LK
  • Celeron, like the normal PII only supports two way SMP. Only a Xeon (II or III) can do four way.
  • looks kinda... TALL. Or that an optical illusion?
  • Anyone have these yet?

    Thanks,

    Dave Manning
  • It may not be true 1GHz speed, but you forget one thing myfriend. Both processors are NOT running the exact same code at the exact time. What does this mean? It means that you are executing 800MHz (from your example) worth of instructions at once. While your CPUs physically are not that fast, your net performance / instruction execution is. You say FPU performance is a different matter. Duh, if FPU is different, so is Integer and all other operations. Why? Again they are running different sets of instructions at the same time. In fact, 800MHz worth of instructions. So, my point is....Net MHz is what couts....not individual. A single processor system with a 400MHz chip has a net MHz of 400. A dual 504 like mine has a net MHz of 1008. It is always net MHz that determines throughput of instructions. (Besides for wait states and cache misses) Just because net MHz is not common enough for most people to use the term doesn't mean its not what really is there.

  • No it is tall. Between 5 and 6 inches. The Motherboard may be fine, but you must have a case to clear the height. Many cases have the P2 slot situated under the power supply as to make this upgrade impossible.
  • Strictly speaking, the APIC is built in to the processor, one in each. It is the I/O APIC that resides in a separate chip. Confusing, huh?

    Anonymous pedant.
  • It is only a few centimeter taller than a Pentium III XEON processor. And usually an ATX case will have the power supply out of the way of the processors, usually with the PS fan blowing across them..(in a full tower case) But then again in a mini-tower or mid tower this would definately be a problem.
  • I don't see why this can't be done on Win9X.

    Simple: The Win9x kernel can not granulate.

    I have no idea how they're expecting to be able to do SMP under Win9x. They'd have to come up with their own SMP engine and tack it onto the kernel I suppose. But then I'm no SE.
  • Posted by Lord Kano-The Gangster Of Love:

    >>I have no idea how they're expecting to be able to do SMP under Win9x. They'd have to come up with their own SMP engine and tack it onto the kernel I suppose. But then I'm no SE.

    It should be possible for software to take advantage of whatever hardware is available. When Win95 first shipped there was no such thing as a "3D Card" but we were able to use them. I don't see why SMP should be any different. As long as you've got a kernel that can support multiple threads SMP should be possible in one form or another.

    LK
  • i don't see anything on that page saying that the dual celeron slot adapter works on any other board, it is always referred to as a motherboard, not an adapter board.
  • Would anyone know what chipsets this setup supports ? I'v been looking to upgrade my LX Pentium II 266 to something faster. Celeron is my only viable option because of the 66 Mhz bus restriction.

    I want this !!

    Message on our company Intranet:
    "You have a sticker in your private area"
  • by Anonymous Coward on Tuesday June 08, 1999 @05:57AM (#1860182)
    Does that QDI board have an APIC chip?

    Abit Dual Socket 370 [abit.com.tw]

    Picture of Motherboard [impress.co.jp]

    Computex piece both of these were listed [impress.co.jp]

    ABIT Announces the Release of the BP6, The World's First Dual Socket370 Motherboard!

    Taipei, Taiwan, May 31th 1999--ABIT announces the release of the World's First Dual Socket 370 motherboard, the BP6. The ABIT BP6 doubly defies conventional limitations by offering both Dual Socket 370 and UDMA/66 on a BX chipset board, once again proving that with ABIT, "Yes, It's possible". The BP6 is based on the award winning design of ABIT's BX line of motherboards. All the great features of our flagship models have been kept, and a lot of amazing new features have been added.
  • Cute... pink parallel and serial connector block :) ... I want one too! :D

    Much more sensible idea than this hoofer card sticking out all over the shop.

    Being Abit, should prolly be overclockable as hell too ..
  • If i'm right, you wouldn't sanely expect to use this in a desktop, right (or you could have fun and cut it up)? In most towers (in mine, atlest) the powersupply is above and behind the CPU. Since this sticks out from the slot like a p2 does, as long as the power supply doesn't cover your p2 in any way, it should work. And if it's too wide for my case, i shall simply have to continue not bothering with putting the side panel on. Of course, the matter of cooling the damn things is another issue. How much weight can that thing hold? and will it use the locking clips? (I will HAVE to get this!)
  • by Hey_bob ( 6104 )
    I might off here...
    But that lil setup still wouldn't allow
    you to use 2 CPU's like that. I think It might
    allow you to have a redundancy (sp) within you system... or not? (one chip fries, the other still goes??)

  • Yes! I also have a working system.. Asus P2B-DS,
    pair of Cele 300a's running 450 or 464.

    I was doing the drilling/jumpering mod as a side buisness, for a while there.. till the socket 370 converters killed interest. (tho I hear they challange overclocking with extra trace length and connector effects)

    Both 2.0.35/36 and 2.2.5 kernels run fine, detect processors, etc... I've been too busy to do much benchmarking since converting to RH 6.0, or even much decent testing back then, but on the 2.0 kernel, I would get compile times (-j unlimited)
    of 1 min 28 seconds, running between 10-20 copies of gcc! This on a 64MB system, with (at the time) a old Quantum Fireball 1.08Gb ide drive..

    Sometime I'll get some tests run with the 2.2 kernel and the new UDMA drive.
  • Ahem... there I go, talking stats and parts like a real bithead. Ya'd tink I used to hop up cars...
    :-)

    Meant to comment on the dual Abit board. I think it would be a MUCH better solution to SMP 370's than the plugin cards, for reasons mentioned in my last post. Cleaner signals, less clock/signal skew possibilites.. gotta be more stable.

    I still suggest slot 1 chips modified, over the 370's in the slocket solutiion, tho the native dual 370 should be as good or better.
  • The 2-CPU card can only work with a special motherboard. Big Deal. So Buy 2 socket MB from a reputable manfuacturer...this whole things seems very deceptive...
  • Posted by Lord Kano-The Gangster Of Love:

    >>I have no idea how they're expecting to be able to do SMP under Win9x. They'd have to come up with their own SMP engine and tack it onto the kernel I suppose. But then I'm no SE.

    It should be possible for software to take advantage of whatever hardware is available. When Win95 first shipped there was no such thing as a "3D Card" but we were able to use them. I don't see why SMP should be any different. As long as you've got a kernel that can support multiple threads, SMP should be possible in one form or another.

    LK
  • I bought to 300a's from these guys about 3 weeks
    ago. I ended up getting them both overclocked to
    504 mhz each in a dual proc system:) Only needed
    to add two extra fans and of course some MONSTER
    cpu fans. 1+ ghz !

    Mike
  • Hehehe... wondering how long till some braindead counterfiter tries to make a Xeon plastic case to glue around it.. good thing it dosen't work in regular motherboards.

    But, if you see a guy in a trench coat, with chips ALLMOST the size of Xeons, for $300, saying "Pssst..." you know which way to walk...
  • I couldn't find any mention of an Abit MB with support for IDE RAID on their website. Anyone care to include a URL?
  • I would be intersted in that too...
  • by Anonymous Coward
    For quad cpu operation, you need four
    arbitration signals (BR#0-BR#3).

    PPro and Xeon have them on slot 2,
    but PII only have BR#0 and BR#1 on slot 1.

    Celeron has BR#0; BR#1 is not documented,
    but it is there (AN15, but maybe removed from
    intel without notice).

    As there are many undocumented ("n.c.") pins
    on the celeron, it is imaginable, that there
    are BR#2 and BR#3 somewhere as well, but
    that doesn't help: the LX and BX chipsets
    only support BR#0 and BR#1 (but the chipsets
    have "n.c."-pins, too :-) )


    But with some simple hardware and software tricks,
    it migth be possible to use quad celerons:

    You need an additional PLD (~$2) on the
    Dual Socket 370 -> Slot 1 Adapter, that
    performs a kind of "sub arbitration".
    This would cheap and simple to do, so
    that memory and i/o accesses might be
    performed correctly by all 4 CPUs, but:

    Two CPUs would have id #0 and the other
    two would have id #1. That will cause
    problems, using the standard APIC-comunication
    functions. But maybe with the right
    little piece of software ... :-)
  • What about SMP under linux? Yes? No? What about overclocking? How does a fan fit?

    I'm doubtful.
  • Only a Xeon (II or III) can do four way.

    You're forgetting, of course, the venerable PPro, which can also do 4-way SMP.

  • This looks like something that escaped from
    some derranged engineer's lab.
    It is still a single-socket motherboard. How the Hell is the OS (a good one, say...NT? ;)) going to know that it has two processors to use?
    Weird.
  • I wonder if yo have a dual slot 1 MB if you can make it a quad cpu MB with that?
  • Does it work without any BIOS stuff though.....Will it work an any mobo? I have a BH-6...seems like having two processors would confuse it; what would the clock speed be? Normal? Excuse the dumb questions, this just seems too good to be true.
  • the LX and BX chipsets only support BR#0 and BR#1
    That's what I would have thought. But looking at the data sheets does not reveal that the BR pins even connect to the chipset.

    The FX chipset supported four-way SMP. I have not been able to find any technical reason why the LX and BX chipsets don't, other than Intel's assertion that they support two-way SMP.

  • SMP support is in the chipset in question, for example, the Intel 440BX. On dual processor motherboards, you don't set the number of processors in the CMOS - the chipset scans the processor bus and determines itself how many CPUs are present.

With your bare hands?!?

Working...