Will an Opteron 1218 work in a motherboard that supports the 1210?
January 23, 2010 12:58 PM Subscribe
Will an Opteron 1218 work in a motherboard that supports the Opteron 1210?
I'm building a ZFS-based backup and storage system for my father, who is a photographer and publisher. I have an AMD Opteron 1218 processor, and had my sights on the Gigabyte GA-MA790X-UD4P motherboard. I picked that motherboard as it supports the Opteron 1210 processor, and I assumed that it would accept the 1218 processor.
However, my vendor asked Gigabyte whether the board supported the 1218. The short reply was that "their desktop boards do not support server processors", which in my opinion is counteracted somewhat by their desktop board indeed supporting the 1210 "server processor".
As far as I can tell, the only difference between the 1210 and 1218 models is the clockspeed multiplier. The thermal profile is identical, and both processors are from the same family and same revision. The CPUID string will even be very similar, with only the field that signifies clockspeed being different.
I am hopeful that the 1210 support will allow 1218 support, and that Gigabyte's denial stems from them simply not allocating manpower to evaluating the 1218 in the board.
In short, would you think 1210 motherboard support is enough to allow the 1218 to work?
I'm building a ZFS-based backup and storage system for my father, who is a photographer and publisher. I have an AMD Opteron 1218 processor, and had my sights on the Gigabyte GA-MA790X-UD4P motherboard. I picked that motherboard as it supports the Opteron 1210 processor, and I assumed that it would accept the 1218 processor.
However, my vendor asked Gigabyte whether the board supported the 1218. The short reply was that "their desktop boards do not support server processors", which in my opinion is counteracted somewhat by their desktop board indeed supporting the 1210 "server processor".
As far as I can tell, the only difference between the 1210 and 1218 models is the clockspeed multiplier. The thermal profile is identical, and both processors are from the same family and same revision. The CPUID string will even be very similar, with only the field that signifies clockspeed being different.
I am hopeful that the 1210 support will allow 1218 support, and that Gigabyte's denial stems from them simply not allocating manpower to evaluating the 1218 in the board.
In short, would you think 1210 motherboard support is enough to allow the 1218 to work?
Response by poster: :) Thanks for the answer.
I have to work with a local Icelandic vendor, and unfortunately they can't take the board back unless faulty. Lesser men than I would resort to vendor manipulation :) ... but we're on such friendly terms with our vendor that it's not worth it. For which I am actually very grateful, I realize now.
A knowledgeable acquaintance suggest I find someone with a board and simply test it. I suspect that some free, cold beer will solve this problem.
posted by krilli at 1:15 PM on January 23, 2010
I have to work with a local Icelandic vendor, and unfortunately they can't take the board back unless faulty. Lesser men than I would resort to vendor manipulation :) ... but we're on such friendly terms with our vendor that it's not worth it. For which I am actually very grateful, I realize now.
A knowledgeable acquaintance suggest I find someone with a board and simply test it. I suspect that some free, cold beer will solve this problem.
posted by krilli at 1:15 PM on January 23, 2010
Okay, I'd say give it a try, and if it fails, buy a new CPU. While it's convenient to use an existing chip, the 3-year old cpu doesn't have that much value. If it doesn't work, you won't have much trouble finding a faster replacement for less than half the cost of the mainboard. I very much recommend against settling for a motherboard you don't like as much just for the sake of being able to use and old dusty CPU.
Also, I know you guys are up to your eyeballs in cheap geothermal electricity, but any replacement chip will almost certainly use less power (or be a ton faster), which means heat, so that'll make the system quieter and/or longer lasting.
posted by aubilenon at 2:42 PM on January 23, 2010
Also, I know you guys are up to your eyeballs in cheap geothermal electricity, but any replacement chip will almost certainly use less power (or be a ton faster), which means heat, so that'll make the system quieter and/or longer lasting.
posted by aubilenon at 2:42 PM on January 23, 2010
Response by poster: True, aubilenon, very true.
What I'm after in the Opteron is that it's cheap, and I hear it has ECC cache, which is very reassuring in a storage tank.
But very good points ... I'm actually thinking that the other motherboard, a Tyan n3400b, would probably be a smidge more stable and have some features I *might* put to use. So I'm in a pretty good position overall.
posted by krilli at 3:00 PM on January 23, 2010
What I'm after in the Opteron is that it's cheap, and I hear it has ECC cache, which is very reassuring in a storage tank.
But very good points ... I'm actually thinking that the other motherboard, a Tyan n3400b, would probably be a smidge more stable and have some features I *might* put to use. So I'm in a pretty good position overall.
posted by krilli at 3:00 PM on January 23, 2010
Wow. For a publisher -- as in, an actual business? Think twice about using an unsupported combination even if it can be compelled to work.
[I would not even consider ZFS for this application on any platform other than a reliable (read: tech support at the ready) server box (with opensolaris, even).]
posted by rr at 7:33 PM on January 23, 2010
[I would not even consider ZFS for this application on any platform other than a reliable (read: tech support at the ready) server box (with opensolaris, even).]
posted by rr at 7:33 PM on January 23, 2010
Response by poster: Fortune favors the bold :)
Seriously though, good point. And indeed I am thinking twice.
I'll be the tech support, at the ready. I have a degree in computer science, long experience with Mac OS X, consider myself to have enough experience with Linux-based SAN-building, will be building a backup strategy, and allowing for complete failure of any individual component or machine. Also, Stanford's Jim Little has a very good blog illustrating how they use commodity-hardware-based ZFS boxes for a similar purpose.
It's a small operation. Mirrored drives in the G5 tower. 5-6 drive RAIDZ2 chunks in the pool. Secondary off-site RAIDZ2 backup unit, connected via secure tunnel over commodity fiberoptics. ECC memory on all machines, ECC cache on ZFS boxes. Redundancy via ubiquity in ZFS machines, using hardware that is known to work.
Do you have or know any concrete bad experiences with ZFS, rr? I'm totally open to the idea being shot down.
posted by krilli at 3:40 AM on January 24, 2010
Seriously though, good point. And indeed I am thinking twice.
I'll be the tech support, at the ready. I have a degree in computer science, long experience with Mac OS X, consider myself to have enough experience with Linux-based SAN-building, will be building a backup strategy, and allowing for complete failure of any individual component or machine. Also, Stanford's Jim Little has a very good blog illustrating how they use commodity-hardware-based ZFS boxes for a similar purpose.
It's a small operation. Mirrored drives in the G5 tower. 5-6 drive RAIDZ2 chunks in the pool. Secondary off-site RAIDZ2 backup unit, connected via secure tunnel over commodity fiberoptics. ECC memory on all machines, ECC cache on ZFS boxes. Redundancy via ubiquity in ZFS machines, using hardware that is known to work.
Do you have or know any concrete bad experiences with ZFS, rr? I'm totally open to the idea being shot down.
posted by krilli at 3:40 AM on January 24, 2010
Response by poster: Let me rephrase: Fault tolerance via ubiquitous commodity parts availability in the ZFS machines.
posted by krilli at 10:29 AM on January 24, 2010
posted by krilli at 10:29 AM on January 24, 2010
Do you have or know any concrete bad experiences with ZFS, rr?
I see no reason to bet on ZFS until it has a few more years of reliability -- Joyent's experience really kind of shook the "ZFS is the perfect filesystem" stuff. Not so much the failure as the recovery issues.
You are planning on running Net/FreeBSD on the G5? Or what?
posted by rr at 10:59 AM on January 24, 2010
I see no reason to bet on ZFS until it has a few more years of reliability -- Joyent's experience really kind of shook the "ZFS is the perfect filesystem" stuff. Not so much the failure as the recovery issues.
You are planning on running Net/FreeBSD on the G5? Or what?
posted by rr at 10:59 AM on January 24, 2010
Aother way to phrase my response: if this is ad hoc volunteer tech support and someone has a business riding on it, the correct approach is to keep it simple and well known.
ZFS is new. It is not simple.
posted by rr at 11:00 AM on January 24, 2010
ZFS is new. It is not simple.
posted by rr at 11:00 AM on January 24, 2010
Response by poster: Conclusion: It works. The "unsupported" Opteron 1218 works perfectly fine in the Gigabyte GA-MA790X-UD4P motherboard. Of course it does!
And I'm going to be using ZFS. The reason I will allow myself to use ZFS is that I'm taking backups, too!
posted by krilli at 1:03 PM on March 7, 2010
And I'm going to be using ZFS. The reason I will allow myself to use ZFS is that I'm taking backups, too!
posted by krilli at 1:03 PM on March 7, 2010
Response by poster: Here's some good reading on practical use of ZFS: http://breden.org.uk/2008/03/02/a-home-fileserver-using-zfs/
posted by krilli at 1:07 PM on March 7, 2010
posted by krilli at 1:07 PM on March 7, 2010
This thread is closed to new comments.
It seems to me fairly likely that it would work, but *shrug*. What's their return policy like?
posted by aubilenon at 1:08 PM on January 23, 2010