VDI in UCS

[dani_bosch]

Hello,

I'm having a brainstorm with some colleagues for a VDI project in UCS, and we were wondering the differences between the several Xeon 5600 processors (besides the obvious #cores difference) applied to a VDI implementation.

For instance, is there any important difference between Xeon E5620 and E5640 for such a solution (besides the 0.2 Ghz difference of the clocking which I don't perceive so important)?? I see one is twice the price compared to the other...that's why I have many doubts at this respect.

Any inputs will be appreciated.

Thanks,

[rmadhana]

Hi Dani,

While comparing x5620 with x5640 only the clock speed varies. But otherwise they are significant difference within the same family processors which will affect your VDI implementation. such as

1. CPU Frequency

2. Turbo ON/OFF

3. Hyperthreading

4. No of Cores

5. Power Utilization

6. Memory Bus Speed

7. QPI Link Speed.

Regards,

Ranga Madhanagopal

[dani_bosch]

But if for instance you focus on these two chips aforementioned, that is E5620 and E5640, when you compare them at www.intel.com you realize there are only 4 differences:

a) clock speed (2.4 GHz vs 2.66 GHz)

b) max turbo freq (2.66 GHz vs 2.93 GHz)

c) bus/core ratio (18 vs 20)

d) embedded options (YES vs NO)

The price is almost TWICE for one compared to the other...

I'm not an expert in CPU's, that's why I'm asking...what is the BIG difference between one and the other?

Would you position a Xeon E5620 for a VDI implementation? (I've only seen high-end CPUs when reviewing VDI designs and white papers out there).

Thanks a lot,

[rmadhana]

Dani,

Below is one other response i got from one of our experts in the team.

Yes, there is little difference except the bus/core ratio is different (which determines other diffs).   The higher price of E5640 comes mainly from this higher ratio because  E5640 dies (semiconductor term on the chip) must come from the smaller and higher quality portion of the silicon prints.   For example, from one 300mm wafer, probably only 15% of the dies would be qualified for E5640 with higher yield failures while 85% are left for E5620 or slower processors with much lower yield failure ratio.   Hence Intel has to charge higher price for E5640.   

Looks like there is not much difference technically.

Regards,

Ranga

[dani_bosch]

So, if I understood it correctly, the main difference comes from MTBF of the chip, rather than from the technical features...is that right?

Thanks,

[lwatta · Accepted solution]

Not MTBF but the fact that it is harder for Intel to actually build the chip.