Time to retire the Dell M610 blade? What’s next?

Westmere X5762 vs Nehalem X5570

After analysis of Westmere vs Nehalem;

The gains aren’t small, but neither are they substantial enough to motivate upgrading from Nehalem to Westmere – only in select cases will the gains be in the 40-50% range. Similarly, the power efficiency improvements are nice, but not profound compared to the prior generation.

The clock-rate increase for the 95W TDP chips (including turbo boost) that can be used in the M610 is 3.2GHz to 3.6GHz (I lock my chips to 3.2GHz permanently by disabling all power management in the BIOS and I also disable hyperthreading. YMMV). You’re looking at a mathematical difference of 11%. The extra cost vs the gains support the original conclusion that the gains are not substantial enough to motivate the upgrade.

The Nehalem architecture was kick-ass, and remembering back to my benchmarks in 2008/2009, I saw a thread for thread 30% increase in capability.

If you’re looking at a multi-threaded application, you could see 20-40% increase in performance with the extra 2 cores. Perhaps virtualization would see the most benefit here. However, when you look at the price tag that STILL exists against the X5672 chips, it really looks less appealing than it could, if you were looking for a quick interim upgrade without replacing all your systems.

Again, your mileage may vary. For me, this isn’t a large enough gain to even warrant picking up a couple of chips to test them.

The Nehalem X5570

Interestingly,  from performance tuning my BIOS, my benchmark of the CPU resulted in interesting results compared to the one on CPU Benchmark’s website:

My result was an aggregate score of: 10,232 for CPU Benchmark.
The 2 results recorded on CPU Benchmark’s website: 6,025.

I cannot delve deeper into the tests submitted to the website, but I can only assume that something was wrong with their setup.
If I look at the Single Thread results for my CPU, it is completely (expectantly) destroyed by the Intel Ivy Bridge i7-3770k @ 3.5GHz. 70.1% faster. This is pretty impressive, and I think it warrants a further look into the modern CPUs and building a prototype system for further testing.

Faster, Faster, Faster

My initial research is showing that the best thread-for-thread CPU for applications that aren’t so multi-threaded seems to be the Intel® Xeon® Processor E3-1290V2 (8M Cache, 3.70 GHz) with turbo boost of 4.10GHz. It’s a 4 core chip with 8 threads (HT) on the 22nm lithography and only 87W TDP (wow!). Unfortunately, it looks like it will not accept a multi-CPU configuration system.  This might not be a problem for me.

Since I’m in the mix, I’m also looking at the Core i7-3770K 3.5GHz (3.9GHz turbo) chip. I bet this will run stable at 4.10GHz without much trouble. I understand it’s a desktop CPU, but the price alone make it worth a second look – especially if I’m thinking about self building. Clock for clock, thread for thread, this looks like a great chip. It holds its own against the E3-1290V2 at a fraction of the price. I have a 2600k at home stable for over a year at 4.5GHz.

A side by side of these two chips. They’re pretty similar.

Intel Core i7-3770K Intel Xeon E3-1290 v2
Intel Core i7-3770K Picture
is not
Specifications differences
Market segment Desktop Server
Manufacturer Intel
Family Intel Core i7 Intel Xeon
Model number i7-3770K E3-1290 v2
CPU part number CM8063701211700 CM8063701099101
Box part number BX80637I73770K
Core name Ivy Bridge Ivy Bridge-H2
Platform name Carlow
Microarchitecture Ivy Bridge
Technology (micron) 0.022
Socket Socket 1155
Frequency (MHz) 3500 3700
Turbo Frequency (MHz) 3900 / 3900 / 3800 / 3700 4100 / 4000 / 3900 / 3800
Clock Multiplier 35 37
L1 cache 128 KB (code) / 128 KB (data)
L2 cache (KB) 1024
L3 cache (KB) 8192
TDP (Watt) 77 87
Cores 4
Multiprocessing 1