TR Forums

Anyone know if the PCIe Raid boards from Promise will seamlessly allow drives to move over from the PCI-X Promise 8300SX SATA Raid controller (controls four 500GB SATA II drives in RAID 10)?

just brew it! wrote:

AFAIK Microsoft OSes and server products are licensed per processor, where a processor is defined as something that fits into a single socket (regardless of how many cores/threads/modules that processor may have). I believe Oracle (which along with Microsoft and IBM is one of the top three software vendors) uses the same model.

Yes, you have to run Windows Professional (or a sever version) for the OS to see the processor in the second socket (but if you're putting together this kind of system, you're probably doing that anyway -- and the extra $40 or so for Pro is a drop in the bucket). But MS isn't quite as rigid (or complicated) as the other two vendors mentioned when it comes to paying a software fee for each and every core on server software: with MS SQL Server, for example, you pay for a particular "edition" of the software which will recognize up to a given number of processors, regardless of core count; whereas with Oracle (IIRC) you end up paying according to a formula that counts both processors and cores per-processor (for their "Enterprise" editions; it's different for the "Standard" editions -- like I said, it's complicated).

Or you could run Linux, and not worry about licensing...

Yep. And if Linux supports the software you're interested in running, that's the way to go. With the number of cores you can reach with even just two sockets, you're probably going to be running VMs and more than one OS anyway.

mnecaise wrote:

I'm typing this on a dual socket xeon workstation made by HP. You might be able to find one of these used on ebay or something...

Like this?

http://www.geeks.com/products_sc.asp?Cat=1250

Refurbished xw8400 with dual Xeon 5150s for $400, or with only one Xeon installed for $260 + ~$50 on eBay for a second. It'll take quad-core Xeons, too, though they can be quite a bit more money- about $100 each for X5355s.

The whole thing probably makes a solid, stable, space-heating workstation at a reasonable price, though you're stuck with awful FB-DIMM memory.

Mr Bill wrote:

[For my dual MP setup I had to link together two ATX power supplies, one for the motherboard and the other for the storage. This setup draws 600W when running. The one in my sig is pulling 240W...

That is a great build. I hadn't heard of anyone using two power supplies like that in a home build. It makes sense to do that, but it's still crazy you need to.

Flatland_Spider wrote:

Mr Bill wrote:

[For my dual MP setup I had to link together two ATX power supplies, one for the motherboard and the other for the storage. This setup draws 600W when running. The one in my sig is pulling 240W...

That is a great build. I hadn't heard of anyone using two power supplies like that in a home build. It makes sense to do that, but it's still crazy you need to.

Well, given the vintage of the equipment I am assuming this was back in the days before 700W+ PSUs were readily available in the DIY market.

Part of the problem is that those Dual MP motherboards needed something like 40A in the 5V + 3.3V but all the newer PSU's only put out their power at 12V.

just brew it! wrote:
AFAIK Microsoft OSes and server products are licensed per processor, where a processor is defined as something that fits into a single socket (regardless of how many cores/threads/modules that processor may have). I believe Oracle (which along with Microsoft and IBM is one of the top three software vendors) uses the same model.

Microsoft will be offering SQL server 2012 on a per core basis:
http://www.microsoft.com/sqlserver/en/us/future-editions/sql2012-licensing.aspx

I recall VMWare having funny rules about cores/processor and pricing the licenses accordingly. But that was two years ago...

edited: so you could see to whom I was replying.

FuturePastNow wrote:

mnecaise wrote:

I'm typing this on a dual socket xeon workstation made by HP. You might be able to find one of these used on ebay or something...

Like this?

http://www.geeks.com/products_sc.asp?Cat=1250

Refurbished xw8400 with dual Xeon 5150s for $400, or with only one Xeon installed for $260 + ~$50 on eBay for a second. It'll take quad-core Xeons, too, though they can be quite a bit more money- about $100 each for X5355s.

The whole thing probably makes a solid, stable, space-heating workstation at a reasonable price, though you're stuck with awful FB-DIMM memory.

Very close to it. If you're looking for a toy... They're certainly capable. The one I'm using has E5410 quad core Xeon's in it. Use it as a developer workstation running VM's and a local SQL server where I can test code. One box. Multiple personalities.

ok jumping back in the middle here I have a question about xeon processors relative to consumer processors. Are they not functionally under-clocked or lower TDP binned twins to consumer CPUs. I'm sure they have a little of this and a little of that of difference but generally speaking they push for lower thermals/power demands so that they can power on at 100 percent utilization in a multi CPU config without exploding or starting on fire. A prior poster sighted you can't OC xeons without sacrificing stability... I understand that might be a relative measurement. Basically if I can OC a xeon on a dual socked SR-2 config to say 3.8 or 4 ghz and only use it in short bursts with Superior cooling to what it would receive in a server stack wouldn't that not really effect the perceived stability of the product since it wouldn't reach the thermals to shorted its life etc??? Basically I'm saying can you change the task and work load, clock accordingly and see benefits? I see there is a whole community built up around the sr-2 and ocing the cheapest dual socket processors to 4.0 ghz and doing little spurts of folding or gaming or photoshop(a extreme variable use model)

Mr Bill wrote:

Part of the problem is that those Dual MP motherboards needed something like 40A in the 5V + 3.3V but all the newer PSU's only put out their power at 12V.

The Tiger MPX was a 2nd gen Dual Socket A board. It had a 4-pin ATX12V connector like modern mobos, and a Molex (right next to the ATX12V) in case you had an "old school" PSU without the ATX12V connector. From the manual:

(Also note the 64-bit PCI slots... for its day, this was a powerhouse mobo!)

kamikaziechameleon wrote:

ok jumping back in the middle here I have a question about xeon processors relative to consumer processors. Are they not functionally under-clocked or lower TDP binned twins to consumer CPUs. I'm sure they have a little of this and a little of that of difference but generally speaking they push for lower thermals/power demands so that they can power on at 100 percent utilization in a multi CPU config without exploding or starting on fire. A prior poster sighted you can't OC xeons without sacrificing stability... I understand that might be a relative measurement. Basically if I can OC a xeon on a dual socked SR-2 config to say 3.8 or 4 ghz and only use it in short bursts with Superior cooling to what it would receive in a server stack wouldn't that not really effect the perceived stability of the product since it wouldn't reach the thermals to shorted its life etc??? Basically I'm saying can you change the task and work load, clock accordingly and see benefits? I see there is a whole community built up around the sr-2 and ocing the cheapest dual socket processors to 4.0 ghz and doing little spurts of folding or gaming or photoshop(a extreme variable use model)

I believe they undergo additional testing, and are binned more conservatively.

The OC issue isn't that they *won't* OC... it is that most motherboards which are intended to be used for Xeon-based systems have no OC options.

kamikaziechameleon wrote:

ok jumping back in the middle here I have a question about xeon processors relative to consumer processors.

As has already been pointed out they're more conservatively clocked and tested. They may have larger cache (12MB vs 4MB). Memory controller is set up to work with larger memory space, higher DIMM count. Additional links for processor interconnect (QPI, HT, etc).

In the x86 world the desktop and server (i7 vs Xeon, Phenom vs Opteron, etc.) chips may be the same die with capability disabled in the desktop version. Not always the case though.

just brew it! wrote:

kamikaziechameleon wrote:

ok jumping back in the middle here I have a question about xeon processors relative to consumer processors. Are they not functionally under-clocked or lower TDP binned twins to consumer CPUs. I'm sure they have a little of this and a little of that of difference but generally speaking they push for lower thermals/power demands so that they can power on at 100 percent utilization in a multi CPU config without exploding or starting on fire. A prior poster sighted you can't OC xeons without sacrificing stability... I understand that might be a relative measurement. Basically if I can OC a xeon on a dual socked SR-2 config to say 3.8 or 4 ghz and only use it in short bursts with Superior cooling to what it would receive in a server stack wouldn't that not really effect the perceived stability of the product since it wouldn't reach the thermals to shorted its life etc??? Basically I'm saying can you change the task and work load, clock accordingly and see benefits? I see there is a whole community built up around the sr-2 and ocing the cheapest dual socket processors to 4.0 ghz and doing little spurts of folding or gaming or photoshop(a extreme variable use model)

I believe they undergo additional testing, and are binned more conservatively.

The OC issue isn't that they *won't* OC... it is that most motherboards which are intended to be used for Xeon-based systems have no OC options.

Ok so this is why the SR-2 is such big noise.

just brew it! wrote:

The Tiger MPX was a 2nd gen Dual Socket A board. It had a 4-pin ATX12V connector like modern mobos, and a Molex (right next to the ATX12V) in case you had an "old school" PSU without the ATX12V connector. From the manual:
(Also note the 64-bit PCI slots... for its day, this was a powerhouse mobo!)

I also see the AGP slot. I seem to recall that most dual socket motherboards of the era didn't come with an AGP slot. Who needs fancy 3D graphics cards in the server being the - at the time - reasonable logic.

cheesyking wrote:

Are there still problems with software licensing on multi socket? I seem to remember all kinds of funny schemes when multi core came along but is that still a concern today?

Ditto with JBI.

For MS, you have to get Windows 7 Professional or Ultimate in order to use two sockets. For anything beyond that, you need to get some version of Window Server (Latest one is 2008 R2). Typically, the basic version is limited to 4 sockets and Advanced version is at 16 sockets. The Datacenter edition can support a silly amount of sockets which would only exist within a big iron/computing cluster.

Krogoth wrote:

The Datacenter edition can support a silly amount of sockets which would only exist within a big iron/computing cluster.

Don't most systems like that run some variant of *NIX these days?

Datacenter Edition has the best licensing terms for Hyper-V. So the "silly" amount of socket support has some advantages there.

As for if Windows is used in what we might class a big iron environment, I suppose someone must as the Datacenter Edition SKU is pretty old. I believe it debuted with Windows 2000.

You see hosting sites offering the Microsoft stack (and sites running asp[x] pages) -- those may be hosted within a VM from *nix, but even then the most cost-effective approach (for large sites) is probably the DataCenter license.
(Naturally, Microsoft would be happy to tell you about some of them)

Ryu Connor wrote:

just brew it! wrote:

I also see the AGP slot. I seem to recall that most dual socket motherboards of the era didn't come with an AGP slot. Who needs fancy 3D graphics cards in the server being the - at the time - reasonable logic.

It's still reasonable logic for workstations/servers that aren't going to get used for GPGPU or 3D/video tasks. The "developer workstation running VM's and a local SQL server" mentioned earlier being a classic example.

Matrox made some PCI-X (not e) video cards for those with long white slots to burn.

just brew it! wrote:

Mr Bill wrote:

Part of the problem is that those Dual MP motherboards needed something like 40A in the 5V + 3.3V but all the newer PSU's only put out their power at 12V.

The Tiger MPX was a 2nd gen Dual Socket A board. It had a 4-pin ATX12V connector like modern mobos, and a Molex (right next to the ATX12V) in case you had an "old school" PSU without the ATX12V connector. From the manual:

(Also note the 64-bit PCI slots... for its day, this was a powerhouse mobo!)

I have a build on that board also. But it was a refurb and always seemed a little flakey in the memory. My MSI board was a review board and it worked with every CPU I put on it. Unfortunately it really needed those 40A on the 5V and 3.3V leads. It still works but I turned it off when I built the system in my sig.

Yeah, considering $259 is around the price of high end consumer boards... It isn't going to cost around $3k unless you buy the high end or mid range server segment stuff. It's entirely possible to make a cheaper dual socket config.

just brew it! wrote:

Well, given the vintage of the equipment I am assuming this was back in the days before 700W+ PSUs were readily available in the DIY market.

I couldn't if it was that, or "I'm bored, and I have a pile of hardware."

Back on topic, if ARM ever gets going in the consumer space, we could see a return of boards with multiple CPUs, socket or otherwise. Getting wide makes more sense for them then getting fast.

I can't bear to look at those socket tabs without an involuntary shudder.

Captain Ned wrote:

I can't bear to look at those socket tabs without an involuntary shudder.

At least the Thermalright HSFs I used on that build attached to all 6 tabs. By then I'd also gotten pretty good at installing/removing Socket A HSF clips, so it was no biggie.

just brew it! wrote:

I used to run a dual Barton MP rig (on a Tyan Tiger MPX) as my main desktop too. And yes, it was quite a space heater!

I still use my Asus A7M266-D board, although I am still running ye olde 1.2 pally MP's in mine...never was able to find reasonably priced barton MP's.

Starfalcon wrote:

just brew it! wrote:

I used to run a dual Barton MP rig (on a Tyan Tiger MPX) as my main desktop too. And yes, it was quite a space heater!

I still use my Asus A7M266-D board, although I am still running ye olde 1.2 pally MP's in mine...never was able to find reasonably priced barton MP's.

I ran MP-unlocked Athlon XPs in my Tyan for quite some time before getting real Barton MPs.

When I got the board, barton MP's were insanely expensive...and now they are nearly impossible to find. The joy of upgrading obsolete hardware...

My dual MP board was an engineering review sample. It ran every generation of MP's without issue. Really makes you wonder how much they cut corners on the production boards...