dev workstation: i7 vs Xeon, does it make a difference ?

I recently found a great source of heavily discounted refurbished Xeon processors, this set me thinking to build a dual Xeon system instead of defaulting on a i7-4790K.
Ofcourse it has to be able to accomodate a good modern gfx card, so PCI-e 3.0 x16 slots are a must, and this narrows it down, but still there are possibilities here.
But before i go out and dump hard earned cash into something crazy, it seems a good idea if any of you folks here have experience with performance gains from such a choice.

As far as i am aware especially the building of lighting is very cpu intensive. On the other hand, if it saves 20 minutes on something you do twice a week… well…

Opinions anyone ?

It depends which Xeons you’re getting. YOu have to go pretty far up before there’s a speed increase by getting Xeons vs. desktop processors. Xeons are built for multitasking and reliability, not so much for speed. Getting a faster processor is only going to improve lighting build times along with shader and code compiling. The rest of the program that uses the CPU can only go so far.

That does make a lot of sense, thank you for your reply.

Well I just updated from a 6 year old computer to an off the rack i7 and although there was a bump in performance the bump is enough that it’s a comfortable experience that I would feel it a wast to want more. For me fast is fast enough and if anything if I wanted to spend more money I would spend it on a really comfortable office chair with a built in seat warmer and ice chest.

Or perhaps a good 4k screen.

The biggest advantage of the Xeons are that they can support more memory and memory bandwidth than the i7s. That bandwidth difference can be massive when performing operations like building lighting on a large, complex scene, since a typical i7 has two memory channels and around 25GB/s of bandwidth, compare to a typical Xeon which has four or more memory channels and at least double the bandwidth as a result. Xeons are also generally better at handling multi-threaded applications, so you may additionally gain some mileage there.

The Xeon’ are also much more stable / reliable, but this is less of an issue since we’re probablt not really doing anything like protracted hardcore data analysis.

There is a caveat though - if you want to make use of the benefits provided by the Xeon, you need to build the system to support it. You’d want plenty of fast memory and a motherboard that supports it, with plenty of bandwidth. This is an additional cost over the top of an already expensive processor that cannot be avoided, since you’ll not be getting any benefit from it.

Personally, I don’t feel like the potential extra performance boost here and there is really worth the colossal extra investment. Heck, for things like lightmass, you can just setup up one or two cheap additional boxes to act as part of a swarm anyway :slight_smile:

Have you got SSDs yet?

The Haswell Xeon CPUs are very cheap, and cheaper as the next better i7 (~50€ difference as of posting this, cheapest Xeon 1240v3 was ~250€ and the cheapest i7 4770K was ~300€). It’s kind of a “secret” tip really if you have to make compromises money-wise. It’s definitively more interesting as an i5 if you do more as just games.

With some effort (custom bios, specific boards), even a littel OC would be possible, but not as much as i5 or i7 models.

If I was you I’d wait for 6 months to upgrade completely my current PC.

The 1240v3 compares quite directly to the 4770K indeed, same socket, 3.4 GHz at €254,- each, compared to €299 for the 4770K.

However, over here the i7-4790K is €310,-
For such a small difference you go from 3.5 GHz to 4.0 GHz, so that is an easy step to make.

Not all that sure if a Xeon is going to be such a great improvement over the 4790K.

@errvald: i dont have a current PC, working on a laptop.

I have a 6-core (12 thread) Xeon at work. (E5-2620), and a 3-core (8-thread) i7-Extreme at home. (i7-3820K). It’s kind of hard to test performance because my GPU in work is a bloody Quadro K4000 (don’t ask), whereas at home I have a GTX 980. GPU performance is (as expected) utterly mental at home, but sucks in work. I also use Quad-Channel 1600Mhz Memory at home.

My Work PC certainly throws data around faster than my home PC, it can certainly load textures into memory much quicker, but for actual processing I’d say my i7 is faster. I could do a CPU test in-engine I suppose.

My Work PC tends to throw data around much quicker than my home PC, but my home PC actually processes it faster. In real terms, I honestly don’t think the Xeon out performs the i7 at all. In fact my home PC is actually faster for rendering applications too, like Maya and Max, but that could be a number of factors contributing to that. Given a choice, I’d take the i7, unless you’re going to overclock the Xeon to similar speeds. (Good luck getting 6 cores stable at 3.3-4.0 Ghz though :p).

Honestly, just save money where you can on CPU and go for a monster GPU instead. GTX 980’s are notoriously cheap compared to their predecessors :wink:

You don’t want to waste money over spec’ing your new CAD workstation, but at the same time you shouldn’t under spec it either or aim for minimum requirements as you will have to live with low levels of productivity until you replace the workstation.

Perhaps sparing out €215 by buying a gtx970 instead of a gtx980 could be poured into a bigger SSD and more RAM instead.
As for the Xeons, i guess that the logical thing to do is go for the i7-4790K at the moment, and if performance is ever going to be a problem then considder going for a heavyduty solution, perhaps even with a dual processor xeon.

I had done a comparison a while back, from the Xeons Vs. an i7 3930k, and based of CPU benchmarks it would require 2 $500 Xeons to get the same speed as the single i7

You guys may want to do your research by visiting credible sites and reading up on individual’s real world experiences with these things. For example, even if you’re on an old X58 platform, the Xeon X5650 (available for $75 off of ebay) can easily match, and out-perform, even more modern Haswell-based i5s and i7s.

Also, as for the “good luck getting a stable OC” comment above, these things are known to hit 4.1Ghz+ without issue, and normally with a much lower vCore than your typical i7 setup on the same X58 platform.

In any case, check out the below threads from 2 very credible forums dedicated to Hardware enthusiasts who know their stuff and don’t mess around when it comes to performance.