Xeon or i7 for Inventor and AutoCAD? (2017-2018)

[DRoam]

I've seen some older suggestions regarding this, but the PC hardware world is ever-changing. As of now, all other things being equal, what are Inventor and AutoCAD going to be "happiest" with: a Xeon with sufficient specs and complementing hardware, or an i7 with sufficient sufficient specs and complementing hardware? This is assuming we can fork out the money for whichever of those is more expensive...

Basically the question is, assuming we have the budget to put into a "really good" Xeon or a "really good" i7 (maybe not top of the line, but this definitely isn't a "budget PC"), is there a clear better choice? And for either choice, what are the ideal specs (# of cores/threads, clock speed, cache size, etc.) and complementing hardware (memory type/specs, graphics card type/specs, etc.)?

Is that a straightforward enough question to answer?

@Neil_Cross, in this video, all of the top PC's (in both the visual- and CPU-intensive tests) were i7's. But in this video, you recommend a Xeon. Is that due to budget considerations? Would an i7 be a better choice with a sufficient budget?

My understanding is that Xeons, at least initially, were designed to be server-type processors--"workhorses", tuned for running constantly for long-term processsing applications where multi-threading and parallel processing are utilized heavily--applications like servers, rendering, FEA, etc. ... and they are NOT optimized for the short, quick bursts of linear processing that Inventor primarily thrives on (except in the drawing environment... which we do care about, too).

Has that changed? Are there now some Xeons that are better suited than i7's for Inventor's (and AutoCAD's) needs, like the Xeon E5-1630v4 that Neil recommended? Or is an i7 still going to be better suited?

Any input and wisdom is appreciated.

[Ray_Feiler]

Because Inventor is primarily single threaded I believe this chart is still the best reference to use when shopping for a new CPU.

https://www.cpubenchmark.net/singleThread.html

[pball]

Going from my experience and the general consensus of the How fast is your Inventor PC really? thread I would recommend an i7 processor. Inventor in general performs better with a higher clock speed processor compared to lower clock speed, even with more cores/threads. The i7 7700k is the processor in our new CAD systems at work and I'm happy with the performance. Below is a link to my inventor benchmark in the previously mentioned thread.

As for going with a xeon processor there are plenty of reasons to do so, usually workload related. Rendering, FEA, and simulations are the main ones I can think of since they take advantage of the more core/threads xeons traditionally have.

My work PC Specs

i7-7700k

32gb ram

1060 graphics card

500gb ssd

[Neil_Cross · Accepted solution]

The i7 is still the better CPU for Inventor if you're only interested in performance, if you want Inventor to run as fast as possible, to the point where the bottleneck is the software itself, the i7-7700K is the best CPU on the market for that.

Xeons these days are practically identical to i7's, the argument will never be settled.  But it is a fact that they're made of the same silicone, for most i7's there's an identical Xeon cut from the same cloth.  However the i7's have their clock frequency tuned slightly higher, and most are unlocked to allow the user to overclock them for even better performance, and Inventor laps that up.  You can't do that with a Xeon.

Look at this, the i7-7700K spec

Then look at this, the Xeon E3-1275v6

Both of those CPUs are absolutely identical, they are literally the same chip, but the i7 has been tuned to a higher clock speed so it will perform better for Inventor.

With the Xeon, they've enabled a few features on the chip that they left disabled on the i7, like ECC memory support.  That workstation ecosystem you can put the Xeon into may still perform slightly behind the i7, but what you lose in performance you might gain in reliability and system longevity.  

So that's why I always recommend Xeons into an enterprise environment, being able to open an assembly 5 seconds quicker with the i7 is great and all, but reliability and enterprise support is key.  You're not going to get premium next day on-site support with that custom rig you put together yourself from parts off the internet.

CPU age is even more important than clock speed too.  An 8 year old CPU at 3.7GHz will be magnitudes slower than a current generation CPU at 3.7GHz.  An 8 year old motherboard and chipset won't have anywhere near as many modern ports & features as a board from today.

It's a complicated answer to an easy question.  But if you value reliability and support, go with a current gen Xeon clocked over 3.7ghz and you'll be close to the i7 (the modern Xeons turbo themselves anyway).  

[dgorsman]

Add another opinion in here: as long as the supporting hardware is sufficient, it won't matter.  You could have an AMD processor too.  Inventor won't care one whit.

There are a couple of reasons for going with a workstation processor like a Xeon.  The first has been mentioned - they are available with more cores, and the motherboards support multiple processors (2, 4, or more), so FEA and CPU-renderers work faster.  However, most users won't really see the benefit of that for the extra cost.  The other is if you have an Enterprise account with one of the Big Box suppliers like Dell or HP, which only supply Xeon-based workstations.

[DRoam]

Thanks to all for the great responses. Neil, you answered my question. I think the Xeons will be the better choice for us if the performance loss is negligible, as Enterprise support is important for us (and of course so is longevity).

Thanks all for the input.

I'm shifting focus to graphics card now. I started another thread for that, so if you all wouldn't mind heading over there and pitching in, that would be very appreciated: Graphics card for Inventor 2017-2018, using 3+ displays @ 2560 x 1440.

Thanks again!

[Neil_Cross]

I know it's only my word and work to support it, but I've thoroughly and extensively tested every module within Inventor that they claim supports multi-core utilisation, and in every test the core count is all but irrelevant.  The software never seemed optimised well enough, every test I performed with real world models the software preferred and performed best with a higher clock speed.  Even graphics, I could have the exact same graphics card installed into 5 different systems and get 5 totally massively different results... with the system clocked the highest on the CPU pumping out the most FPS.

Most of my results are published but here's an example of clock speed being king:

 

A 6 year old i5 2500K overclocked to 5.1GHz meshing a complex model faster than a 12 core Xeon, an 8 core Ryzen 7 1700X, and if you can see (not sure how clear this pic will come out) a 2 core Pentium at 4.2GHz solving the mesh in some cases faster than a 6 core Xeon and the 8 core Ryzen.

I did 23 tests like this testing literally every single apparently multi-threaded module of Inventor, the conclusion is ultimately that the software needs better optimisation for modern hardware.

Purely in my own opinion, based on my own thorough testing.

If anyone is interested though, I'm currently building a new Ryzen 7 based mini ITX workstation with Radeon Pro WX graphics and will be benching it on these tests now that the platform has matured a fair bit.  That'll be on YouTube in the next week or so hopefully.

[DRoam]

Thanks a million for your diligent research, Neil. That's really valuable information to have. Really surprising that operations that utilize multiple threads don't seem to do so more efficiently.

I'd be curious to know if the other top CAD packages (SolidWorks, SolidEdge, Creo, CATIA) have the same multi-thread struggles as well, or if they utilize multi-core better. In other words, is this just a limitation of 3D CAD in general right now? Is it maybe due to the fact that all of these CAD systems are built from their core on single-thread architecture, and building multi-core use onto that (efficiently) isn't easy/feasible? Or is it due to the nature of 3D CAD itself, and multithreading will just never be useful for 3D CAD? (Aside from simulation, which unquestionably should benefit from it if implemented properly). Don't know if you know the answers to those questions or not, just wondering out loud.

[DRoam]

Alright, so we've switched focus to the i7's at our IT guy's request. And apparently Dell, for whatever crazy reason, isn't able to provide the i7-7700K... can't imagine why.

So, right now it looks like it's between the i7-4790K and the i7-6700K.

Both of those are older than the 7700 (especially the 4790), but both have same base clock and seem to perform really well in benchmarks. The 4790 actually seems to do a little better than the 6700 even though it's older.

Anyway, any thoughts on those CPUs (or others)?

EDIT: The i7-7740X is the top CPU on the single-thread benchmark list, and isn't much more expensive. It's also very new. Any thoughts on that one as well?

[Neil_Cross]

Dell do provide the 7700K, it'll just be limited to certain models:

Under no circumstances get the 4790K, it's a great CPU but it's 3 or 4 generations old now and therefore so is the architecture it's based on, it's limited to DDR3 RAM, the chipset it supports isn't as modern as the newer ones, it makes no sense to buy something that old when people are considering coming off that now to upgrade.  That doesn't mean its a bad CPU though, I use one in my main rig but I'm also considering upgrading.

Don't pay too much attention to the scores at the top end of the Inventor Bench thing, some of us began getting ridiculous just to try and max out the score i.e. me overclocking my 4790K to 4.8GHz which isn't sustainable, but it pulls a better score than a stock 6700K, but the 6700K is factually a better CPU in every way.

The 7740X is a complete waste of a lot of money.  It's almost exactly identical to the 7700K, but needs a X299 based motherboard which are hella-expensive and designed for much higher end CPUs, i.e. a X299 motherboard has 8 RAM slots but if you put in a 7740X you can only use 4 of them.  But KabyLake-X (the platform name for the 7740X) is a very new platform, Dell are usually light years behind the times so I wouldn't expect to see those in Precision Workstations any time soon if ever... the X represents enthusiast grade, I don't recall ever seeing the X range of CPUs in a Dell Precision.

If you're buying Dell Precision, I personally wouldn't go for an i7 but they've got a quality build going for £1900

Precision T3620

i7-7700K

Windows 10

NVIDIA Quadro P4000

32GB DDR4 RAM (2x16GB)

512GB PCIe M.2 NVMe SSD

1TB secondary drive

Absolute WINNING deal that, I doubt Dell will allow you to overclock the CPU in their BIOS but still those specs are killer for £1900.

In fact, you don't even strictly need the P4000, if you swapped that out with a Radeon Pro WX3100 you can have all that for £1500 which is frankly one of the best deals I've seen in a long time.