Xeon E3 v4

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Deslok

Well-Known Member
Jul 15, 2015
1,122
125
63
34
deslok.dyndns.org
they would be interesting for small VDI boxes if you can pass the IGP through, it's enough to help a lot of things without needing a full GPU and could be interesting in SMB enviroments
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
They will not be around long enough to make a dent in the market TBH.
 
  • Like
Reactions: eva2000

EffrafaxOfWug

Radioactive Member
Feb 12, 2015
1,394
511
113
I'm inclined to agree, although it looks like E3's with graphics might be the new standard. Boring for the graphics but I'd be interested to see what sort of a difference the L4 cache makes to some typical server loads.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
I'm inclined to agree, although it looks like E3's with graphics might be the new standard. Boring for the graphics but I'd be interested to see what sort of a difference the L4 cache makes to some typical server loads.
Well the "issue" Intel has is that the E5 is better at the high end (you can virtualize and get better power/ performance and economics) and the Xeon D/ C2000 series are good for the dedicated low end looking at compute only. So the E3 is basically now relegated to applications that need high clocks, low expand-ability/ networking, low core counts and low memory.

The E3 V4 has 1/4 the memory capacity of the Xeon D, a PCH + the need for a 10Gb adapter (figure ~20w). The E3 has fewer PCIe 3.0 lanes too. The E3 has higher single-threaded performance but lower multi-threaded.

If you are looking for pure low power, the upcoming 25w TDP Xeon D becomes attractive as does the C2000 and the successor we will have (finally) coming.
 
  • Like
Reactions: hjfr

SDLeary

Member
Aug 4, 2015
48
9
8
56
Well the "issue" Intel has is that the E5 is better at the high end (you can virtualize and get better power/ performance and economics) and the Xeon D/ C2000 series are good for the dedicated low end looking at compute only. So the E3 is basically now relegated to applications that need high clocks, low expand-ability/ networking, low core counts and low memory.

The E3 V4 has 1/4 the memory capacity of the Xeon D, a PCH + the need for a 10Gb adapter (figure ~20w). The E3 has fewer PCIe 3.0 lanes too. The E3 has higher single-threaded performance but lower multi-threaded.

If you are looking for pure low power, the upcoming 25w TDP Xeon D becomes attractive as does the C2000 and the successor we will have (finally) coming.
What could the onboard graphics on the v4 (and presumably v5) E3s provide in terms of OpenCL? Could this of benefit to the Workstation crowd?

SDLeary
 

Jeggs101

Well-Known Member
Dec 29, 2010
1,529
241
63
What could the onboard graphics on the v4 (and presumably v5) E3s provide in terms of OpenCL? Could this of benefit to the Workstation crowd?

SDLeary
I think for mobile workstations or very dense microservers you get good benefit. For a mITX or bigger workstation you'll go PCIe every time for real performance.
 

SDLeary

Member
Aug 4, 2015
48
9
8
56
I think for mobile workstations or very dense microservers you get good benefit. For a mITX or bigger workstation you'll go PCIe every time for real performance.
I should have been clearer. :)

I was in fact thinking of a Workstation built on mITX and a discrete card. CPU, main fictionality; GPU, graphics; iGPU, OpenCL compute (or perhaps reverse the GPU and iGPU). I'm thinking in terms of an architecture similar to the MacPro, where you have CPU and dual GPUs, one of which is assigned as a compute engine.

SDLeary
 

Patriot

Moderator
Apr 18, 2011
1,450
789
113
I should have been clearer. :)

I was in fact thinking of a Workstation built on mITX and a discrete card. CPU, main fictionality; GPU, graphics; iGPU, OpenCL compute (or perhaps reverse the GPU and iGPU). I'm thinking in terms of an architecture similar to the MacPro, where you have CPU and dual GPUs, one of which is assigned as a compute engine.

SDLeary
Does it also need to overheat when you use the gfx like the macpro?
Also without a riser that can split the x16 to dual x8's no dual gpu off mitx.
 

SDLeary

Member
Aug 4, 2015
48
9
8
56
Does it also need to overheat when you use the gfx like the macpro?
Also without a riser that can split the x16 to dual x8's no dual gpu off mitx.
Reread what I posted. :D

It wouldn't need a dual GPU setup, and thus no riser. CPU, iGPU, and GPU, a beefier version of the setup that many larger laptops with discrete GPUs have. In the case of the possible Workstation that I'm describing, the iGPU would be assigned OpenCL duties.

SDLeary
 

MiniKnight

Well-Known Member
Mar 30, 2012
3,072
973
113
NYC
I find that interesting, and I bet you could do it. I would likely fall into the category of using onboard graphics for 4K 2D and then a PCIe card for OpenCL if that was my primary concern. But that's for my use case.
 

Patriot

Moderator
Apr 18, 2011
1,450
789
113
Reread what I posted. :D

It wouldn't need a dual GPU setup, and thus no riser. CPU, iGPU, and GPU, a beefier version of the setup that many larger laptops with discrete GPUs have. In the case of the possible Workstation that I'm describing, the iGPU would be assigned OpenCL duties.

SDLeary
If your discrete is not better at OpenCL than the igp... what is the point of having it?
Intel has had quicksync for some time...while faster than cpu accel, not sure if it would keep up with the descrete.
 

SDLeary

Member
Aug 4, 2015
48
9
8
56
If your discrete is not better at OpenCL than the igp... what is the point of having it?
Intel has had quicksync for some time...while faster than cpu accel, not sure if it would keep up with the descrete.
I believe that certain Adobe apps (Premiere Pro?) have the ability to assign a gpu to compute duties while the other performs normal GPU functions.

Also I found this on Apples discussion forums:

Reply Helpfulby Grant Bennet-Alder on Aug 6, 2014 6:23 PM

so what is the second card (or the one on slot 1) actually doing?

It turns out that the stuff a graphics card has been doing all this time is almost exactly what we used to call an Array Transform processor. It can, in parallel, execute the same set of mathematical instructions (like multiply and accumulate) in lock-step, on thousands of numbers at a time. There are some really interesting problems that could use that kind of compute power.

So.. the second card is doing ... computation.

The trouble with having a graphics card doing massively parallel computations is that it gets interrupted to re-draw the screen 60 to 100 times a second, and it can't just pick up where it left off. So the second card is deliberately NOT connected to any displays, so it does not get interrupted by trivial stuff.
So I'm assuming that such a configuration should be good in other situations where high speed computation and simultaneous rendering are needed. Admittedly an edge case, especially in the configuration that I posited.

SDLeary
 

hjfr

Member
Nov 21, 2013
79
20
8
France
The E3 V4 has 1/4 the memory capacity of the Xeon D, a PCH + the need for a 10Gb adapter (figure ~20w). The E3 has fewer PCIe 3.0 lanes too. The E3 has higher single-threaded performance but lower multi-threaded.

If you are looking for pure low power, the upcoming 25w TDP Xeon D becomes attractive as does the C2000 and the successor we will have (finally) coming.
Intel has a wide range of (low power) server products now. Xeon e3 1200 v4/v5, e3-1505, xeon D, C2000 (too old now I think: ~2013). But the availibility of these products are really bad for home/lab users, specially in Europe. :-/