CPU Performance: System Tests

Our System Test section focuses significantly on real-world testing, user experience, with a slight nod to throughput. In this section we cover application loading time, image processing, simple scientific physics, emulation, neural simulation, optimized compute, and 3D model development, with a combination of readily available and custom software. For some of these tests, the bigger suites such as PCMark do cover them (we publish those values in our office section), although multiple perspectives is always beneficial. In all our tests we will explain in-depth what is being tested, and how we are testing.

All of our benchmark results can also be found in our benchmark engine, Bench.

For our graphs, some of them have two values: a regular value in orange, and one in red called 'Intel Spec'. ASUS offers the option to 'open up' the power and current limits of the chip, so the CPU is still running at the same frequency but is not throttled. Despite Intel saying that they recommend 'Intel Spec', the system they sent to us to test was actually set up with the power limits opened up, and the results they provided for us to compare to internally also correlated with that setting. As a result, we're providing both sets results for our CPU tests.

Application Load: GIMP 2.10.4

One of the most important aspects about user experience and workflow is how fast does a system respond. A good test of this is to see how long it takes for an application to load. Most applications these days, when on an SSD, load fairly instantly, however some office tools require asset pre-loading before being available. Most operating systems employ caching as well, so when certain software is loaded repeatedly (web browser, office tools), then can be initialized much quicker.

In our last suite, we tested how long it took to load a large PDF in Adobe Acrobat. Unfortunately this test was a nightmare to program for, and didn’t transfer over to Win10 RS3 easily. In the meantime we discovered an application that can automate this test, and we put it up against GIMP, a popular free open-source online photo editing tool, and the major alternative to Adobe Photoshop. We set it to load a large 50MB design template, and perform the load 10 times with 10 seconds in-between each. Due to caching, the first 3-5 results are often slower than the rest, and time to cache can be inconsistent, we take the average of the last five results to show CPU processing on cached loading.

AppTimer: GIMP 2.10.4

.

FCAT: Image Processing

The FCAT software was developed to help detect microstuttering, dropped frames, and run frames in graphics benchmarks when two accelerators were paired together to render a scene. Due to game engines and graphics drivers, not all GPU combinations performed ideally, which led to this software fixing colors to each rendered frame and dynamic raw recording of the data using a video capture device.

The FCAT software takes that recorded video, which in our case is 90 seconds of a 1440p run of Rise of the Tomb Raider, and processes that color data into frame time data so the system can plot an ‘observed’ frame rate, and correlate that to the power consumption of the accelerators. This test, by virtue of how quickly it was put together, is single threaded. We run the process and report the time to completion.

FCAT Processing ROTR 1440p GTX980Ti Data

.

3D Particle Movement v2.1: Brownian Motion

Our 3DPM test is a custom built benchmark designed to simulate six different particle movement algorithms of points in a 3D space. The algorithms were developed as part of my PhD., and while ultimately perform best on a GPU, provide a good idea on how instruction streams are interpreted by different microarchitectures.

A key part of the algorithms is the random number generation – we use relatively fast generation which ends up implementing dependency chains in the code. The upgrade over the naïve first version of this code solved for false sharing in the caches, a major bottleneck. We are also looking at AVX2 and AVX512 versions of this benchmark for future reviews.

For this test, we run a stock particle set over the six algorithms for 20 seconds apiece, with 10 second pauses, and report the total rate of particle movement, in millions of operations (movements) per second. We have a non-AVX version and an AVX version, with the latter implementing AVX512 and AVX2 where possible.

3DPM v2.1 can be downloaded from our server: 3DPMv2.1.rar (13.0 MB)

3D Particle Movement v2.1

.

3D Particle Movement v2.1 (with AVX)

.

Dolphin 5.0: Console Emulation

One of the popular requested tests in our suite is to do with console emulation. Being able to pick up a game from an older system and run it as expected depends on the overhead of the emulator: it takes a significantly more powerful x86 system to be able to accurately emulate an older non-x86 console, especially if code for that console was made to abuse certain physical bugs in the hardware.

For our test, we use the popular Dolphin emulation software, and run a compute project through it to determine how close to a standard console system our processors can emulate. In this test, a Nintendo Wii would take around 1050 seconds.

The latest version of Dolphin can be downloaded from https://dolphin-emu.org/

Dolphin 5.0 Render Test

.

DigiCortex 1.20: Sea Slug Brain Simulation

This benchmark was originally designed for simulation and visualization of neuron and synapse activity, as is commonly found in the brain. The software comes with a variety of benchmark modes, and we take the small benchmark which runs a 32k neuron / 1.8B synapse simulation, equivalent to a Sea Slug.

Example of a 2.1B neuron simulation

We report the results as the ability to simulate the data as a fraction of real-time, so anything above a ‘one’ is suitable for real-time work. Out of the two modes, a ‘non-firing’ mode which is DRAM heavy and a ‘firing’ mode which has CPU work, we choose the latter. Despite this, the benchmark is still affected by DRAM speed a fair amount.

DigiCortex can be downloaded from http://www.digicortex.net/

DigiCortex 1.20 (32k Neuron, 1.8B Synapse)

.

y-Cruncher v0.7.6: Microarchitecture Optimized Compute

I’ve known about y-Cruncher for a while, as a tool to help compute various mathematical constants, but it wasn’t until I began talking with its developer, Alex Yee, a researcher from NWU and now software optimization developer, that I realized that he has optimized the software like crazy to get the best performance. Naturally, any simulation that can take 20+ days can benefit from a 1% performance increase! Alex started y-cruncher as a high-school project, but it is now at a state where Alex is keeping it up to date to take advantage of the latest instruction sets before they are even made available in hardware.

For our test we run y-cruncher v0.7.6 through all the different optimized variants of the binary, single threaded and multi-threaded, including the AVX-512 optimized binaries. The test is to calculate 250m digits of Pi, and we use the single threaded and multi-threaded versions of this test.

Users can download y-cruncher from Alex’s website: http://www.numberworld.org/y-cruncher/

y-Cruncher 0.7.6 Single Thread, 250m Digitsy-Cruncher 0.7.6 Multi-Thread, 250m Digits

.

Agisoft Photoscan 1.3.3: 2D Image to 3D Model Conversion

One of the ISVs that we have worked with for a number of years is Agisoft, who develop software called PhotoScan that transforms a number of 2D images into a 3D model. This is an important tool in model development and archiving, and relies on a number of single threaded and multi-threaded algorithms to go from one side of the computation to the other.

In our test, we take v1.3.3 of the software with a good sized data set of 84 x 18 megapixel photos and push it through a reasonably fast variant of the algorithms, but is still more stringent than our 2017 test. We report the total time to complete the process.

Agisoft’s Photoscan website can be found here: http://www.agisoft.com/

Agisoft Photoscan 1.3.3, Complex Test

.

Our New Testing Suite for 2018 and 2019 CPU Performance: Rendering Tests
Comments Locked

136 Comments

View All Comments

  • SaturnusDK - Wednesday, January 30, 2019 - link

    The price is the only big surprise here. At $3000 for the CPU alone and three times that in system price it's actually pretty decently priced. The performance is as expected but it will soon be eclipsed. The only question is what price AMD will change for it's coming Zen2 based processors in the same performance bracket, we won't know until then if the W3175X is a worthwhile investment.
  • HStewart - Wednesday, January 30, 2019 - link

    I thought the rumors were that this chip was going to be $8000. I am curious what Covey version of this chip will perform and when it comes out.

    But lets be honest, unless you are extremely rich or crazy, buying any processor with large amount of cores is crazy - to me it seems like high end gaming market is being taking for ride with all this core war - buy high end core now just to say you have highest performance and then next year purchase a new one. Of course there is all the ridicules process stuff. It just interesting to find a 28 core beats a AMD 32 core with Skylake and 14nm on Intel.

    As for Server side, I would think it more cost effective to blade multiple lower core units than less higher core units.
  • jakmak - Wednesday, January 30, 2019 - link

    Its not really surprising to see an 28 Intel beating an 32Core AMD. After all, it is not a hidden mystery that the Intel chips not only have a small IPC advantage, but also are able to run with a higher clockrate (nevertheless the power wattage). In this case, the Xeon-W excells where these 2 advantages combined are working 28x, so the 2 more cores on AMD side wont cut it.
    It is also obvious that the massive advantage works mostly in those cases where clock rate is the most important part.
  • MattZN - Wednesday, January 30, 2019 - link

    Well, it depends on whether you care about power consumption or not, jakmak. Traditionally the consumer space hasn't cared so much, but its a bit of a different story when whole-system power consumption starts reaching for the sky. And its definitely reaching for sky with this part.

    The stock intel part burns 312W on the Blender benchmark while the stock threadripper 2990WX burns 190W. The OC'd Intel part burns 672W (that's right, 672W without a GPU) while the OCd 2990WX burns 432W.

    Now I don't know about you guys, but that kind of power dissipation in such a small area is not something I'm willing to put inside my house unless I'm physically there watching over it the whole time. Hell, I don't even trust my TR system's 330W consumption (at the wall) for continuous operation when some of the batches take several days to run. I run it capped at 250W.

    And... I pay for the electricity I use. Its not cheap to run machines far away from their maximally efficient point on the curve. Commercial machines have lower clocks for good reason.

    -Matt
  • joelypolly - Wednesday, January 30, 2019 - link

    Do you not have a hair dryer or vacuum or oil heater? They can all push up to 1800W or more
  • evolucion8 - Wednesday, January 30, 2019 - link

    That is a terrible example if you ask me.
  • ddelrio - Wednesday, January 30, 2019 - link

    lol How long do you keep your hair dryer going for?
  • philehidiot - Thursday, January 31, 2019 - link

    Anything up to one hour. I need to look pretty for my processor.
  • MattZN - Wednesday, January 30, 2019 - link

    Heh. That's is a pretty bad example. People don't leave their hair dryers turned on 24x7, nor floor heaters (I suppose, unless its winter). Big, big difference.

    Regardless, a home user is not likely to see a large bill unless they are doing something really stupid like crypto-mining. There is a fairly large distinction between the typical home-use of a computer vs a beefy server like the one being reviewed here, let alone a big difference between a home user, a small business environment (such as popular youtube tech channels), and a commercial setting.

    If we just use an average electricity cost of around $0.20/kWh (actual cost depends on where you live and the time of day and can range from $0.08/kWh to $0.40/kWh or so)... but lets just $0.20/kWh.

    For a gamer who is spending 4 hours a day burning 300W the cost of operation winds up being around $7/month. Not too bad. Your average gamer isn't going to break the bank, so to speak. Mom and Dad probably won't even notice the additional cost. If you live in cold environment, your floor heater will indeed cost more money to operate.

    If you are a solo content creator you might be spending 8 to 12 hours a day in front of the computer. For the sake of argument, running blender or encoding jobs in the background. 12 hours of computer use a day @ 300W costs around $22/month.

    If you are GN or Linus or some other popular YouTube site and you are running half a dozen servers 24x7 plus workstations for employees plus running numerous batch encoding jobs on top of that, the cost will begin to become very noticable. Now you are burning, say, 2000W 24x7 (pie in the sky rough average), costing around $290/month ($3480/year). That content needs to be making you money.

    A small business or commercial setting can wind up spending a lot of money on energy if no care at all is taken with regards to power consumption. There are numerous knock-on costs, such as A/C in the summer which has to take away all the equipment heat on top of everything else. If A/C is needed (in addition to human A/C needs), the cost is doubled. If you are renting colocation space then energy is the #1 cost and network bandwidth is the #2 cost. If you are using the cloud then everything has bloated costs (cpu, network, storage, and power).

    In anycase, this runs the gamut. You start to notice these things when you are the one paying the bills. So, yes, Intel is kinda playing with fire here trying to promote this monster. Gaming rigs that aren't used 24x7 can get away with high burns but once you are no longer a kid in a room playing a game these costs can start to matter. As machine requirements grow then running the machines closer to their maximum point of efficiency (which is at far lower frequencies) begins to trump other considerations.

    If that weren't enough, there is also the lifespan of the equipment to consider. A $7000 machine that remains relevant for only one year and has as $3000/year electricity bill is a big cost compared to a $3000 machine that is almost as fast and only has $1500/year electricity bill. Or a $2000 machine. Or a $1000 machine. One has to weigh convenience of use against the total cost of ownership.

    When a person is cognizant of the costs then there is much less of an incentive to O.C. the machines, or even run them at stock. One starts to run them like real servers... at lower frequencies to hit the maximum efficiency sweet spot. Once a person begins to think in these terms, buying something like this Xeon is an obvious and egregious waste of money.

    -Matt
  • 808Hilo - Thursday, January 31, 2019 - link

    Most servers run at idle speed. That is a sad fact. The sadder fact is that they have no discernible effect on business processes because they are in fact projected and run by people in a corp that have a negative cost to benefit ratio. Most important apps still run on legacy mainframe or mini computers. You know the one that keep the electricity flowing, planes up, ticketing, aisles restocked, powerplants from exploding, ICBM tracking. Only social constructivists need an overclocked server. Porn, youtubers, traders, datacollectors comes to mind. Not making much sense.

Log in

Don't have an account? Sign up now