ES Xeon Discussion

[britinpdx]

The topic of ES Xeons has been mentioned in a few threads recently, and I'm somewhat ignorant and it's time to educate myself (and others I'm sure).

Intel do have a support page that gives some basic information on Engineering Samples (and do point out that they are property of Intel)

"Nathan_P" over at [H]ard|Forum has posted a list of Sandy/Ivy/Haswell ES processors

@Patriot had mentioned in a recent thread ..

If anyone has questions on ES chips and stepping etc feel free to ask...

So I'll get us started.

1) If ES processors are pre-production units provided to OEMs and ODMs for testing prior to release, why would they be of interest ?
2) Should ES processors only be considered for use with "non critical" data (anyone remember the Pentium FDIV bug) ?
3) How do I determine if ES processors are missing capabilities ?
4) Any specific ES processors that I should avoid (or look for) ?
5) It appears that the S-Spec is the best way to identify ES processors, but there also appears some processors that have "released to production" S-Specs but are in fact ES processors .. what's up with that ?

 

[Patriot]

I helped Nathan with that list.

Why a stepping is passed over for production is not always known.
Sometimes it is not passing longevity tests is more leaky then they would like (consumes more power) or other non-patchable problems.

1) Cheap, because they are a mostly unknown quantity

2) Depends on what the use case would be. And what stepping you are talking about. I wouldn't put a high availability database on certain ES chips

3) Google and people who have dealt with them.

4) That is a mighty long list... Sigh brain dump commence....
Lets just start with 1366...
Nehalem/45nm /55xx Anything before C stepping is bad... like... they changed the whole electrical layout in B2.
D stepping uses less power.

1156/ w/e
Um... just throw it away. This socket should never have happened.

Westmere/32nm/56xx
A0 stepping worked... Have to hand it to intel on this one... not often is the very first silicon stable.
That said... have found quite a few of them die young... I think B2 is retail.

SB/1155
C stepping same as big silicon

SB-EN/32nm/E5 1356/2011
B0/B1 QAxx Turbo is borked on B steppings and not always stable. Most motherboard vendors have locked them out of later bioses to deter use. I would stay away.
C0 QBxx works, vt-d working is unknown
C1 QBxx retail stepping with borked vt-d
C2 QBxx retail stepping with working vt-d

IB/22nm/E3 v2 /1155
turbo and bigger on ES and you find odd chips.
aim for QExx
I have a spectacular little guy. 2.2ghz 4c/8t act 3ghz system pull load running cinebench was 45w with GT430gfx card in. No, you may not have it.


IB-EN/22nm/E5 v2 1356/2011
I have met very few Ivy chips that did not work... Just work at getting later steppings.
That said I don't think I have played with any QCxx IB-E
Go for QDxx/QExx

HW/22nm/E3 v3 /1150
B steppings are just unstable... C stepping or go home

HW-EN/22nm/E5 v3
You want C0 or later equivalent stepping Due to 3 distinct dies there are odd steppings like M0 and stuff. Same as Ivybridge.
QGxx QFxx might work... Like Sandybridge motherboard vendors are locking out A and B stepping chips.

Not that I expect much amd love here... and their ES chips are rare.
ZS on the front of the string denotes it... You want G45 or G54


5) If an ES chip has the same stepping as a retail chip it just means it was the first of the batch... they checked it and sent it out and didn't find anything wrong... no reason to rev.

You will often find chips that do not fit any retail chip guidelines... either by clock or core count. They may suite your needs better than a retail chip... but Intel chose to delineate the market differently.

There is plenty more in my head... that is just all I could think of at the moment...
Feel free to ask more. You can also show me a picture of the underside of a chip and I can tell you what it is...

 

[britinpdx]

Brilliant !

that gives me some good guidelines as to what to look for.

If an ES chip has the same stepping as a retail chip it just means it was the first of the batch... they checked it and sent it out and didn't find anything wrong... no reason to rev.

I just received an LGA2011 processor that is not marked externally as ES or confidential, and fact it is marked as "E5-2620v2" and "SR1AN". CPU-Z identifies it as ES as does the Intel processor diagnostic tool (which apparently won't run once it detects an ES processor). Interestingly (now that you pointed it out) the SMT passives on the bottom of the package do have a different layout.

Which somewhat leads into the next issue .... Motherboard and BIOS support for ES processors...

I dropped the processor into an Intel S2600CP2 and it wouldn't boot. However, the same processor in a Supermicro X9DRI boots without issue.

Intel not supporting their own processors past a certain rev doesn't surprise me. My only other Intel board is an DX79SR, which will support an E5-2603, but won't support an E5-1620v2. I didn't even bother testing the ES processor in this board.

Any particular guidelines for Motherboard manufacturers ... my limited testing shows Intel poor, Supermicro good.

 

[TType85]

SR1AN is production version of the E5-2620 V2. That's not an ES processor, it should work in just about any board that supports the v2 processors. The ones I had worked find in a S2600CP2 with the latest bios updates.

 

[andrewbedia]

Hey, I like my X3440 "workstation" (actually use it as a spare gaming machine with a Radeon HD 6990 in it).

1156/ w/e
Um... just throw it away. This socket should never have happened.

 

[Patriot]

Hey, I like my X3440 "workstation" (actually use it as a spare gaming machine with a Radeon HD 6990 in it).

a 1 year socket... I stand by my statement lol.

 

[andrewbedia]

It was like $150 for a complete system. Dumb or not, it worked out great for me.

 

[ATS]

1) If ES processors are pre-production units provided to OEMs and ODMs for testing prior to release, why would they be of interest ?
2) Should ES processors only be considered for use with "non critical" data (anyone remember the Pentium FDIV bug) ?
3) How do I determine if ES processors are missing capabilities ?
4) Any specific ES processors that I should avoid (or look for) ?
5) It appears that the S-Spec is the best way to identify ES processors, but there also appears some processors that have "released to production" S-Specs but are in fact ES processors .. what's up with that ?

1) Generally cheap and free. In general they shouldn't be of interest unless they are closer to the free end of the spectrum

2) You absolutely should not use and ES for anything that you care about at all. Esp if you don't know the exact province and full internal stepping information vs PRQ for the ES and the specific fuse configuration vs PRQ parts. The ES could have hardware bugs and/or fuse setting significantly different from production parts. Also ES parts could of been subject to accelerated testing which could significantly lower its life span.

3) In general, without inside info, it is extremely difficult.

4) Consider any ES product at best a plaything without inside knowledge. And its unlikely that even most insiders really know.

5) They can be later generation ES parts. AKA, the ES parts were shipped and nothing was found to be a significant issue between ES shipping and PRQ. There can easily be multiple generations of ES parts shipped to partners over the course of post silicon.

 

[acmcool]

How is QFGH for haswell-EP?

 

[skynet]

HW-EN/22nm/E5 v3
You want C0 or later equivalent stepping Due to 3 distinct dies there are odd steppings like M0 and stuff. Same as Ivybridge.
QGxx QFxx might work... Like Sandybridge motherboard vendors are locking out A and B stepping chips.

Not that I expect much amd love here... and their ES chips are rare.
ZS on the front of the string denotes it... You want G45 or G54

I'm considering buying a e5 Xeon 2690 v3 (or 2695) from the following listing:

2690: Intel Xeon E5-2690 v3 ES LGA2011-3 12C Compatible with X99 i7-5820K 5930K 5960X

Stepping Info from page:

S-spec numbers

ES/QS processors Production processors
Part number QFRX QGNY SR1XN
BX80644E52690V3 +
CM8064401439416 + + +

Architecture / Microarchitecture

Microarchitecture Haswell
Platform Grantley-EP
Processor core Haswell-EP
Core steppings M0 (QFRX)
M1 (QGNY, SR1XN)

2695: Intel Xeon E5-2695 v3 ES LGA2011-3 14C Compatible with X99 i7-5820K 5930K 5960X

the 2695 doesn't include stepping info on the page except for a cpu-z capture that shows Stepping: 1

----

So question is, how do these processors look?

The 2690 seems to be a stepping you described as odd (M0)

Planning to use them in a dual socket setup for general use and running simulation software while I sleep (comparable to fritz)

Thanks for any advice you or anyone else can provide.

 

[skynet]

I got in touch with the seller for more information:

2690 v3 QEYJ
2695 v3 EQY6

 

[Patriot]

Min for V3 and V4
QGxx for 14c+ preferred for all... QF can work for 12c down on certain chips... use cpuworld to see if its the latest stepping or not.
QKxx for v4

 

[skynet]

Thanks for the reply Mr Patriot.

Do you have any specific reason to believe this wont work 2690 v3 (12c) QEYJ?

This is currently the most heavily sold stepping on ebay. I've read through the "buyers feedback" to see if anyone has had any problems and so far so good. I googled around a little bit and the number of people posting about using them is pretty low but so far I can't find a good reason to not buy a pair.

 

[gridrunner]

Im waiting on a 2690 v3 QEYJ should be here in next 2 days. I will report back here what i find. I am hoping vt-d will be supported as i will be using it for running some virtual machines with gpu passthough.

 

[Patriot]

Its like buying B0 sandybridge... it might work for a while in your particular board... but you are doomed to board compat issues and potentially short chip life.

 

[skynet]

Thanks Patriot. Will take that into consideration.

gridrunner have you got your 2690 up and running yet?

 

[mouse]

I'm interested in this 2630L v4 ... QHVK..any thoughts?
Thanks

 

[RadeoNite]

I actually bought one of those from that seller 2 weeks ago, but I don't remember the sSpec number being mentioned back then (I could be wrong though). Anyway, it arrived fine and I could test it out in a X99 motherboard briefly.

Mine doesn't have the "Intel confidential" etching on it though, beside that the sSpec number seems to be SR12G which is not mentioned on cpu-world or anywhere else. For now I'm still waiting for my other parts to arrive before being able to play around with it again.

 

[skynet]

I've made a point of going through "seller's feedback" on pages of the main guys who sell these high end ES chips and I haven't seen any bad reviews. So the chips are seem to be working well out of the box, if any problems do arise they'll happen further down the line.

It also helps that most of the top sellers confirm what motherboard you use and that the chip they are selling will work with your motherboard before they sell it. We are also protected by ebay's guarantee. In cases like this where a seller has ~20 of the same stepping chips and has 10 sold with seller's feedback on that chip, I think it's perfectly fine to go ahead and buy one.

Personally, I am just allowing for a somewhat high possibility that the chip doesn't last very long. If windows suddenly doesnt boot in 1 year I won't be surprised (and I'll have a backup chip), if it still boots in 3 years, I'll consider myself lucky.

 

[Nathan_P]

I've got a pair of early B0 Sandy bridge E5-2665's, they've been running in a folding rig for 3 years 24/7 and haven't missed a beat however YMMV and as Patriot wisely pointed out I can not flash the bios to a newer one as it breaks compatability with the chips.