Re: Asus X99-R5E and Xeon?
My impatience (and fear that the offer would be revoked) has ruled the day!
I now have this SuperMicro-branded OEM Xeon E5-1680-3 (in fact this is where my IT contact originally purchased it 2 months ago). Not sure if SuperMicro is good, bad, or ugly - but it still has fully comply with Intel part spec and compatibility, yes?
Now I worry whether the R5E mobo (which isn't engineered for the server market) implements all the BIOS workarounds/fixes/errata identified in the Intel Xeon E5-3 Processor Specification Update. It seems that buggy TSX instructions in my new proc are disabled in microcode, even though they worked (improperly?) in my old proc - damn! - I want my multithreading madness! None of the other listed errata (aside perhaps from minor thermal throttling under certain conditions) really applies to my hardware or concerns me much.
I suppose I could still swap it back if advised how terribly foolish I have been. "You retrograded to a Xeon??? WTF were you thinking, man?"
Re: Asus X99-R5E and Xeon?
To address your original question, this link will show you a good breakdown on the technical differences between the four CPUs you originally mentioned (I love ARK).
http://ark.intel.com/compare/81909,83361,82767,82930
Quote:
Originally Posted by
Konrad
I notice that these two 8-core Xeons appear superior to this 8-core i7 (indeed, I suspect they might even be different bins/packaging of otherwise identical parts). Unless I'm missing something important?
IIRC, the first two digits determine the die; the second two determine the bin. The first digit determines the scalability factor (1 == 1S, 2 == 2S, 4 == 4S).
Quote:
Originally Posted by
Konrad
The E5-16xx part is a "1S Uni-processor", the E5-26xx parts are "2S Dual-/Multi-processor" - does this mean anything at all on my single-socket single-processor X99 mobo?
1S == Only one CPU per MBB.
2S == Up to two CPUs per MBB.
4S == Up to four CPUs per MBB.
Quote:
Originally Posted by
Konrad
I'm guessing QPI specs won't have any impact on compatibility/performance with my X99 chipset (which links to the proc through 5GT/s DMI2.0)?
I am not intimately familiar with QPI, but the below should give you a good starting point. At a bare minimum, I would not expect performance to be any worse than with the i7.
http://en.wikipedia.org/wiki/Intel_Q...h_Interconnect
Quote:
Originally Posted by
Konrad
I obviously won't be using more than 64GB, the maximum my mobo supports (although there are a few X99 mobos out there which will support 128GB or 256GB, when 16GB and 32GB DDR4 UDIMMs become available). But I suspect the generally beefier capabilities of the integrated memory controllers in the costly Xeon procs might tolerate larger and faster memory configurations.
As you mention, the memory controllers are on-die for all of these CPUs. As such, I would fully expect the MBB to work just fine with higher density DIMMs. In fact, I would expect that you should even be able to run RDIMMs with the E5's, since the MBB should not know the difference (actually, yes, the Asus QVL report even mentions 16GB RDIMMs). All three E5's you mention can support up to 768GB RAM natively.
Quote:
Originally Posted by
Konrad
If Xeons can overclock well (and I suspect they generally can't)
I cannot really speak to this other than to say that I have very rarely heard of Xeons being overclocked. This is likely more due to their intended market than anything else.
Quote:
Originally Posted by
Konrad
I don't expect any compatibility issues with my CPU cooler, chassis, PSU, etc. I don't expect any legal complications. (My IT guys work with a mix of Dell, HP, IBM/Lenovo, and Supermicro equipment, if that matters.)
Agreed, I would not expect any compatibility issues. I don't know what kind of legal complications you are thinking of, but there will be none of those either, unless you are somehow in violation of some special support contract you have to only use consumer-grade equipment. :P
Quote:
Originally Posted by
Konrad
Somebody informed me that Intel locks out Xeon CPUs in microcode as the platforms age ... I have no idea what this means, but if true then it sounds worrisome.
I would ask them for more clarification on what they mean. If they mean that Intel disables Xeons beyond a certain age, that is absolutely and definitively false. Keep in mind, some of the customers using Xeon CPUs have decade+ system lifetimes with downtime measured in human lives. There are a lot of dumb things that consumer electronics manufacturers pull that just does not fly in the enterprise space.
Quote:
Originally Posted by
Konrad
Is this a good trade or bad trade? Am I just asking for a ton of Xeon problems?
Back of the envelope guestimate, going to the E5-1680V3, I would expect that you would get a slight performance boost and a generally more stable product.
Quote:
Originally Posted by
Konrad
Not sure if SuperMicro is good, bad, or ugly - but it still has fully comply with Intel part spec and compatibility, yes?
All Intel CPUs are manufactured by Intel and all members of a model line go through (and pass) exactly the same QA. The only thing that the SuperMicro branding means is either a) absolutely nothing, or b) SuperMicro may have tested some samples of that model in some of their server models. I suspect the former as the page does not mention a specific SuperMicro server or MBB model.
Quote:
Originally Posted by
Konrad
Now I worry whether the R5E mobo (which isn't engineered for the server market) implements all the BIOS workarounds/fixes/errata identified in the
Intel Xeon E5-3 Processor Specification Update. It seems that buggy TSX instructions in my new proc are disabled in microcode, even though they worked (improperly?) in my old proc - damn! - I want my multithreading madness! None of the other listed errata (aside perhaps from minor thermal throttling under certain conditions) really applies to my hardware or concerns me much.
This I cannot really comment on beyond saying that if you need a detailed answer, probably the only way you will get it is to either test it yourself or ask Asus.
Re: Asus X99-R5E and Xeon?
Thanx for the detailed answers - pretty much as I expected, but I was apprehensive about expecting something unexpected, lol.
I am now fairly convinced my best-of-the-i7's part was second-bin for this E5 part, lol.
Interesting that TCase for the E5-1680-3 is 66.26C and for the i7-5960X is 66.8C ... still running my traditional self-imposed 24-hour stress test, looking good so far, this E5 proc peaked at 73C while the i7 peaked at 79C (on best mounting) with the same cooler.
I am infinitely pleased that the iMC on this proc is spectacularly robust. My 8x8GB DDR4-3000 gave me hours or finicky troubleshooting and tweaking to run at speed before, but now they just drop in with out-of-the-box XMP settings and zero-tweaked full performance awesomeness.
Quote:
Originally Posted by
x88x
... I have very rarely heard of Xeons being overclocked. This is likely more due to their intended market than anything else.
I haven't played at overclocking this proc yet, although I am confident it might do well. Perhaps unlikely as well as my above-the-curve i7 part, but hard to say. VID/VCache/etc defaults are all a lot lower on this processor, which looks promising since these are what ultimately determine the overclocking bottlenecks on Haswell-E parts.
Quote:
Originally Posted by x88x
I would ask them for more clarification on what they mean. If they mean that Intel disables Xeons beyond a certain age, that is absolutely and definitively false. Keep in mind, some of the customers using Xeon CPUs have decade+ system lifetimes with downtime measured in human lives. There are a lot of dumb things that consumer electronics manufacturers pull that just does not fly in the enterprise space ...
...
... if you need a detailed answer, probably the only way you will get it is to either test it yourself or ask Asus.
It is the support expert from Asus (ROG Forums) who mentioned this, though I still don't seem to understand exactly what he was saying about this (after multiple readings). He is informative, helpful, and professional but the omens suggest the two of us just aren't destined to get along, lol.
No reply yet on my question about BIOS errata, yeah I might end up having to personally test whether the particular issues which concern me have been addressed. I am under the impression that any problems which exist in these Haswell-EP procs are also present (but hardly relevant and/or never mentioned) in their Haswell-E counterparts.
Re: Asus X99-R5E and Xeon?
Couple notes here:
Xeons are usually locked when it comes to overclocking. That is not to say you can't change things like voltage and multipliers if they are motherboard based settings.
I have been working with the Xeon platforms for many many years and have never heard of a company locking out Xeon models in the microcode. If the board supports that model of CPU it will always support that model.
There really is no such thing as a Supermicro branded CPU. Supermicro is a partner of Intel's and buys the CPU's at an incredibly cheap price since they move so many of them. One caveat: They only buy and sell "tray" cpu's that only come with a one year warranty, whereas the "boxed" CPU's come with a 3 year warranty.
My company is a Supermicro partner and reseller and I could almost hit their main building with a well bounced super ball...
Most motherboards that support the i7's will also support the E3 versions of the Xeon CPU's. It is unique for a motherboard to support i7 and Xeon E5 CPU's, but ASUS has worked with Intel many years to add this sort of unique support.
If you have any specific questions when it comes to overclocking on that board with the Xeon, let me know and I will run it past my Intel rep.
BTW: We are also an Intel Platinum Partner which has many benefits including free engineering support for these exact situations. Free samples, yearly conferences (in Dallas this year) and lots of swag...
