Is there such a thing as a manufacturer-specific CPU?

[MichaelD]

The answer has got to be "NO" but thought I'd throw it out there. I'm looking to buy some new S771 Xeons for my workstation. Found a great price on some E5450's (3.0GHz/1333FSB) but they are advertised as "IBM Xeon." See page here: http://www.compuvest.com/Desc.jsp?iid=1807056

They also have the same model Xeon listed with a different Part# as "Dell Xeon." I'd imagine that just means that these are pulls from IBM or Dell servers/workstations, but you never know...maybe some weird OEM thing they had Intel build into the CPU? When searching that same site for the same Xeon, through their "CPUs" link, the price is much more. I mean, a CPU is a CPU, right? Thoughts?

 

[Ferzerp]

Potentially has the IBM specific heatsink with it.

Or is in IBM branded packaging, etc.

Intended for IBM servers. (When you buy a 2nd proc from IBM, Dell, etc it doesn't come in an Intel box)

 

[pelov]

Wasn't there an article about this just a few weeks ago? One that specified that some customers get specific chips made by Intel/AMD, or at least a little extra attention?

 

[MichaelD]

I know that GPU manuf's definitely "bin" the good chips for sale to their major partners, e.g. NVidia and EVGA. But not sure how that works with CPUs, especially "Enterprise" parts like Xeons.

The comment about possibly coming with a system-specific heatsink makes sense, though the webpage doesn't say that.

Thanks for your thoughts.

 

[Ferzerp]

Typically, on server parts, you're also buying the maintenance, support, etc. You avoid finger pointing by sourcing all the processors from the server vendor. It's the same with memory.

You can save money by going elsewhere, but most of us stick with the vendor because the extra that we are paying keeps us with one neck to choke should things go poorly, and downtime is far, far, far more costly than any piece of hardware that we purchase. I will say that we're more willing to go elsewhere for memory, etc now that we've decoupled all our workloads from any single piece of hardware though. Even if there is a delay of 1 week in replacing a proc of a dimm, who cares? We're built out for the loss of an entire blade chassis, and even with 16 extra servers, we still spend far less on hardware than we did when every system had a dedicated piece of hardware.

 

[MichaelD]

I agree, Ferzerp. The "one neck to choke" is typically the way we go at work as well. This is for my workstation at home though (I should've specified in my OP) so the only thing I'm worried about is matching the stepping on the CPUs.