Thinking about those new E5-2600-v4 benchmarks from Patrick...

[joek]

...and thinking about which of the new or old CPU's would be appropriate... the new Intel E5-v4 CPU's have something to say for themselves, but when put against the competition (E5 V3 CPU's), which is preferable?

Taking a cue from Patrick's color-coded ranking, I tried to do the same. NOTE: Prices in the table are from Intel ark, which are not necessarily representative. Also, Intel changes specifications and does not always update all documentation concurrently. The attached table is based on best information available at time of construction.

And to anyone listening or who knows, I would really like to incorporate the Xeon-D turb-bins. But AFAICT Intel has not published them.

Corrections-suggestions welcome. If you want the source spreadsheet PM me, be happy to provide.

edit: update with more recent (corrected) sheet.

 

[Jerry Renwick]

I prefer to E5 V3 CPU.

 

[joek]

I prefer to E5 V3 CPU.

Why?

 

[T_Minus]

@joek if you are ok with spicy then the E5-2011 v3 line up for price / performance is very hard to pass up. Just make sure you're getting something made right before production and not right after introduction of said chip I believe @Patriot or someone else posted some guides on how to know when the chip was manufactured based on the letters and previous generations.

If you're buying for work/production obviously I'd go production chip

 

[Davewolfs]

These chips appear to be incredibly expensive. You better have a good use case for them.

 

[joek]

@Davewolfs -- Agree some of them are way out of my home budget or use cases. $500-700 is about the limit for a CPU for me, and that demands it be capable of consolidating/eliminating 2-3 other system in my stable (which is ancient). The 2630 v4 looks very good, especially when looking at how far it can stretch in turbo vs. the v3. I looked specifically at turbo-bins and speeds across CPUs in another thread (that chart is also easier to read than the above).

 

[Joseph Nunn]

I really like this chart, it shows a number of brilliant ways to understand the SKUs Intel is putting out for Broadwell EP and how they relate to the Haswell EP SKUs.

But, is it me or does the last column look completely wrong?

The column label says GHz/Core at 4 Cores, which I believe means when using 4 cores what will be the maximum turbo boost GHz frequency of the 4 active cores. If so then according to the turbo bins shown on
List of Intel Xeon microprocessors - Wikipedia, the free encyclopedia
there is a mistake in the calculations.

For example on the top line, the 2699 v4 can run at 3.3 GHz when using 4 cores, not 2.35 GHz. I don't even understand where the number in the chart could come from, even when running all 22 cores the 2699 v4 can still do that at 2.8 GHz for each of those 22 cores!

Am I not understanding the final column correctly?

@joek I also looked at the other chart you provided a link to, the listed turbo bins also do not jive with what is on the wiki link above. For example you show the 2699 V4 as operating at a turbo frequency of 4 GHz with 2 cores active, but checking Intel Ark for that processor shows a maximum turbo frequency of 3.6, which does agree with the wiki link listed turbo bins I provided a link for.

If you don't mind me asking, how did you come up with the numbers? Am I not understanding how turbo speeds are computed?

Joseph

 

[joek]

@ joseph Nunn Sorry, thought I had updated that with corrected numbers; the discussion was continued in thread
Xeon E5-2600 v3-v4 base and turbo speeds
which also lists the sources.

 

[Joseph Nunn]

@ joseph Nunn Sorry, thought I had updated that with corrected numbers; the discussion was continued in thread
Xeon E5-2600 v3-v4 base and turbo speeds
which also lists the sources.

Hello joek, I mention in my post that I visited that page and the turbo speeds listed are wrong there as well. I mentioned 2699 v4 as an example, which I believe should have the exact same turbo speeds across all core counts as the 2699 v3. As I mention in my original post, a quick look at Intel Ark will tell that the 2699 v4 cannot hit 4 GHz at all as its max turbo speed is 3.6 GHz, and yet you use that as a source so I just don't know how you got the turbo numbers you did. Checking another source you mention, hexus.net, also gives a max of 3.6 GHz, although they do it with a lower clock rate and higher turbo bin, however we know the base clock rate mentioned there of 1.8 GHz is wrong so the turbo bins must wrong be as well.

Ultimately, I believe the link I provided to the processor list on Wikipedia has the accurate turbo bin values, as they agree with the maximum turbo speeds listed on Intel Ark, and the base clock frequencies are also correct there.

Joseph

 

[joek]

Thanks @Joseph Nunn. I now understand the error. (Apology for the tardy replies but I'm swamped at the moment.) Looks like there was an error in the original Intel launch documents.... e.g., they showed Ef-2699-v4 with a base frequency of 1.8GHz and a max turbo boost of +18... which would yield the max 3.6GHz you mention. But some other documents specified a base frequency of 2.2GHz, which is how I got to 4.0GHz = 2.2 + 18. Oh well. Note to self: Do not take early marketing docs at face value, and double-check the numbers--if they look too good to be true, they probably are. Thanks again for checking the numbers.

p.s. I try to use primary sources if at all possible; unfortunately Wiki doesn't provide a cite for the turo-bin numbers. For v3 Intel provides numbers in their specification documents (which is always my first choice, even over ark), but I have not seen such for v4 (only marketing materials). If you have any pointers to primary sources, I'd greatly appreciate it. Thanks again.

 

[Joseph Nunn]

Personally I would consider Wikipedia more reliable than a primary source. It is effectively peer reviewed, expert moderated, and has been shown in studies to have remarkably few errors compared to more traditional information sources. If there was an error in a primary source such as from Intel, the Wikipedia community would in all probability be able to identify it and promote the correct information.

Joseph