Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I am in the process of specifying a low-mid range server (which we will purchase multiples of). The application is essentially a software router, so it is very NIC-based I/O intensive. Processor selection is proving challenging.

In comparing the Intel E5-1410 (4C, 2.8GHz) to a price-comparable E5-24xx proc like the E5-2407 (4C, 2.2GHz), one notable difference is that the 2407 has a QPI bus, while the 1410 does not. Details are sparse, although it is clear that the 1410 can only be run in a single die configuration.

I understand that the QPI bus is used to share IO across processors/cores. It appears that the 1410 has integrated IO in lieu of QPI, but it seems impossible to find even a rough metric with regards to I/O performance between the two processor variants.

How might I go about doing price/performance evaluation of the two processor variants? I'm sure that I'll need to test for my particular use case, but I'm having a challenging time finding even general information on relative performance.

EDIT:

The 14xx series does not support memory sparing and memory mirroring, which is supported by the 24xx series. Still looking for details WRT IO capability.

share|improve this question
    
    
@TheCleaner: Thanks. That was the one post that I found that I didn't link to in my question. I cannot reconcile the claim that "QPI doesn't apply since it's only used between processors" with all of the other information that I've found regarding QPI with suggests that it is an IO fabric, albeit one that is optimized for multiprocessor systems. –  HikeOnPast Jul 25 '12 at 4:34

2 Answers 2

up vote 3 down vote accepted

The E5-2407 has a single 6.4GT/s QPI connection good for 204.8 Gbit/s.

The E5-1410 has a single DMI 2.0 connection good for only 20 Gbit/s.

If you care about IO you NEED the 2407 ok.

share|improve this answer
    
The E5 platform has an on-die PCI-E 3.0 controller, one of the big changes of the new E5 platform. A great advantage is that PCI-E traffic doesn't have to traverse QPI links anymore (as was the case with Nehalem and Westmere). From a technical POV you should get all the IO goodness you'll need with any of the E5s. –  pfo Jul 25 '12 at 19:18
    
In the case of IO QPI/DMI is for hooking up a PCI-E root complex to the cores/memory. DMI 2.0 equals (it essentially is) PCI-E 2.0 4x. If any of the components you care about needs more bandwidth then you won't be able to get more than that. –  pfo Jul 25 '12 at 19:22
    
@Chopper3, valid, but peripherals can't use QPI directly; they have to go through an IOH. While I agree with your answer, it seems to be comparing apples to oranges. (qdpma.com/SystemArchitecture/SystemArchitecture_QPI.html) –  HikeOnPast Jul 26 '12 at 17:39
    
@pfo, so the E5-1410 has an on-die PCIe 3.0 controller, but it's disabled in favor of the DMI 2.0 bus? –  HikeOnPast Jul 26 '12 at 17:41

Buy one of each and test them. I think that's about the only way you can be sure how your particular application will perform.

I didn't find anything more in a quick Google search than you did.

share|improve this answer
    
Agreed, but my question is targeted at structural/spec differences between the two processor families rather than "what will run my app the fastest?" –  HikeOnPast Jul 25 '12 at 18:48

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.