Nehalem EP Benchmarks

[Lone Rangir]

I'd like to purchase a pair of e5520 CPUs, but is it worth it to spend extra money on a faster model?

E5520 - $385
E5530 - $540
E5540 - $780
X5550 - $1000

I'm having a hard time finding benchmarks to compare the models against each other. Can anyone point me in the right direction?

Is it money well spent to up the frequency, or is it diminishing returns?

 

[mozartrules]

Diminishing return. I am testing an X5560 at work and it is great, but you get more for your money with the 5520. The only reason for the X5560 is that nothing slower meets my benchmark needs and when you replace existing iron in the 100k range nobody asks questions about a couple of thousand extra.

 

[lopri]

Diminishing return without a doubt.

 

[Idontcare]

Originally posted by: Lone Rangir
Is it money well spent to up the frequency, or is it diminishing returns?

You can't evaluate price/performance of the computer systems without taking into account the price of the entire system.

You listed cpu prices, but what is that in addition to for the rest of the rig.

If your system price is $10k without cpu's, then your question is whether or not $385 versus $1000 enables a price/performance delta to justify your system costing $10,385 versus $11,000...

In other words the $1000 cpu need not performance 3x faster than the $385 cpu just to make the price/performance worthwhile to buy the $1k cpu. You are comparing the performance of the entire investment in your computing rig.

Now if you were talking about comparing price/performance of upgrades for a pre-existing system then your comparison of just cpu prices is valid. But I am assuming you don't already own a Nehalem-based xeon socketed system.

 

[Nemesis 1]

As normal your spot on. I would like to know only 1 thing . What is cost of AMDs chipsets Vs intels cost. What ever the differance is . Its not alot. But Intel MB s cost more and are in higher demand so many more are produced. This has the effect of making intel M/B cheaper to produced . Yet the low volumn AMD M/Bs sell for less. In economics this makes zero sense. So I do believe the M/b makers should be held accountable for the hugh price differances betweeen the 2 . Without Intels high volumn sales and pricies M/B makers would lose $$$ on AMD products . This should be investagated. This makes zero sense.

I do believe you guys will see in news soon about a class action lawsuite thats going to go after M/B makers for these reasons . Proping up Intel M/B pricies while allowing low volumn un profittable AMD cheap M/Bs . Once we have the chipset pricies . econmomics of volumn will show that this is illeagel .

 

[Lone Rangir]

The total cost of the hardware is only around $3200.
$500 for the MB
$1200 for RAM ( 48GB )
$800 for the two E5520
Case, PSU, Optical, yada yada.... round out the rest.

We are RAM bound, in our current platform, which is 2 to 3 year old Opteron-based platform( 16GB ) . The new machine has to be Nehalem based, this was more or less an edict from management.

I'm not sure if I can justify asking for $800 more to go from E5520s to E5540s (or even faster) . I'm trying to find some hard data to show how much performance will increase. I understand it will be faster, but how much faster? 5% or 25%...? I have to have a data driven argument.

 

[Lone Rangir]

In general, how much of an increase do you see when you go up one speed bin?

 

[lopri]

AT did a virtualization test with X5570 and L5520 in it. L5520's performance should be identical to E5520.

http://it.anandtech.com/IT/showdoc.aspx?i=3606&p=7

 

[Lone Rangir]

Wow, $1000 more each for 30% better performance.... that's going to be a tough sell.