The Intel Kaby Lake-X i7 7740X and i5 7640X Review: The New Single-Threaded Champion, OC to 5GHz
by Ian Cutress on July 24, 2017 8:30 AM EST- Posted in
- CPUs
- Intel
- Kaby Lake
- X299
- Basin Falls
- Kaby Lake-X
- i7-7740X
- i5-7640X
Benchmarking Performance: CPU System Tests
Our first set of tests is our general system tests. These set of tests are meant to emulate more about what people usually do on a system, like opening large files or processing small stacks of data. This is a bit different to our office testing, which uses more industry standard benchmarks, and a few of the benchmarks here are relatively new and different.
All of our benchmark results can also be found in our benchmark engine, Bench.
PDF Opening
First up is a self-penned test using a monstrous PDF we once received in advance of attending an event. While the PDF was only a single page, it had so many high-quality layers embedded it was taking north of 15 seconds to open and to gain control on the mid-range notebook I was using at the time. This put it as a great candidate for our 'let's open an obnoxious PDF' test. Here we use Adobe Reader DC, and disable all the update functionality within. The benchmark sets the screen to 1080p, opens the PDF to in fit-to-screen mode, and measures the time from sending the command to open the PDF until it is fully displayed and the user can take control of the software again. The test is repeated ten times, and the average time taken. Results are in milliseconds.
PDF opening is all about single thread frequency and IPC, giving the win to the new KBL-X chips.
FCAT Processing: link
One of the more interesting workloads that has crossed our desks in recent quarters is FCAT - the tool we use to measure stuttering in gaming due to dropped or runt frames. The FCAT process requires enabling a color-based overlay onto a game, recording the gameplay, and then parsing the video file through the analysis software. The software is mostly single-threaded, however because the video is basically in a raw format, the file size is large and requires moving a lot of data around. For our test, we take a 90-second clip of the Rise of the Tomb Raider benchmark running on a GTX 980 Ti at 1440p, which comes in around 21 GB, and measure the time it takes to process through the visual analysis tool.
FCAT similarly favors frequency and IPC. For this sort of workload, the Core i7 is the best chip to get.
3D Movement Algorithm Test v2.1: link
This is the latest version of the self-penned 3DPM benchmark. The goal of 3DPM is to simulate semi-optimized scientific algorithms taken directly from my doctorate thesis. Version 2.1 improves over 2.0 by passing the main particle structs by reference rather than by value, and decreasing the amount of double->float->double recasts the compiler was adding in. It affords a ~25% speed-up over v2.0, which means new data.
As 3DPM expands into several threads, the new quad-core parts will easily get trounced here by AMD's 8-cores for the same price. The Core i7-7800X puts on a good showing, as per core Intel's chips give a higher score.
DigiCortex v1.20: link
Despite being a couple of years old, the DigiCortex software is a pet project for the visualization of neuron and synapse activity 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. The results on the output are given as a fraction of whether the system can simulate in real-time, so anything above a value of one is suitable for real-time work. The benchmark offers a 'no firing synapse' mode, which in essence detects DRAM and bus speed, however we take the firing mode which adds CPU work with every firing.
DigiCortex likes a bit of everything: cores, threads, IPC, frequency, uncore frequency, and memory frequency. The Core i7 parts roughly double the Core i5s due to the thread count, and also the AMD Ryzen parts skip ahead as well due to having double the threads to the Core i7.
Agisoft Photoscan 1.0: link
Photoscan stays in our benchmark suite from the previous version, however now we are running on Windows 10 so features such as Speed Shift on the latest processors come into play. The concept of Photoscan is translating many 2D images into a 3D model - so the more detailed the images, and the more you have, the better the model. The algorithm has four stages, some single threaded and some multi-threaded, along with some cache/memory dependency in there as well. For some of the more variable threaded workload, features such as Speed Shift and XFR will be able to take advantage of CPU stalls or downtime, giving sizeable speedups on newer microarchitectures.
Agisoft is like a Formula 1 race circuit: the long fast straights and techical corners make it a nightmare to have the technology to be the best at both, and Photoscan has enough serial code for high single thread performance to take advantage but also massively parallel sections where having 12-18 threads makes a difference. Despite having half the threads, the single core performance of the Core i7-7740X makes it pull ahead of the Ryzen 7 chips, but when comparing the four threads of the Core i5-7640X to the twelve threads of the Ryzen 5 processors, having 12 threads wins.
Facebook
Twitter
RSS
176 Comments
View All Comments
mapesdhs - Monday, July 24, 2017 - link
Ok, you get a billion points for knowing Commodore BASIC. 8)IanHagen - Monday, July 24, 2017 - link
Dr. Ian, I would like to apologize for my poor choice of words. Reading it again, it sounds like I accused you of something which is not the case.I'm merely puzzled by how Ryzen performs poorly using msvc compared to other compilers. To be honest, your finds are very relevant to anyone using Visual Studio. But again, I find Microsoft's VS compilar to be a bit of an oddball.
A few weeks ago I was running my own tests to determine wether my Core i5 4690K was up to my compiling tasks. Since most of my professional job sits on top of programming languages with either short compile times or no compilation needed at all, I never bothered much about it. But recently I've been using C++ more and more during my game development hobby and compile times started to bother me. What I found puzzling is that after running a few test I couldn't manage to get any gains through parallelism, even after verifying that msvc was indeed spanning all 4 threads to compile files. Than I tried disabling two cores and clocking the thing higher and... it was faster! Not by a lot, but faster still. How could it be faster with a 50% decrease in the number of active cores and consequently threads doing compile jobs? I'm fully aware that linking is single threaded, but at least a few seconds should be gained with two extra cores, at least in theory. Today I had the chance to compile the same project on a Core i7 7700HQ and it was substantially slower than my Core i5 4690K even with clocks capped to 3.2 GHz. In fact, it was 33% slower than my Core i5 at stock speeds.
Anyhow… Dr. Ian’s findings are a very good to point out to those compiling C++ using msvc that Skylake-X is probably worth it over Ryzen. For my particular case, it would appear that Kaby Lake-X with the Core i7 7740X could even be the best choice, since my project somehow only scales nicely with clocks.
I just would like to see the wording pointing out that Skylake-X isn’t a better compiling core. It’s a better compiling core using msvc at this particular workload. On the GCC side of things, Ryzen is very competitive to it and a much better value in my humble opinion.
As for the suggestion, I’d say that since Windows is a requirement trying to script something to benchmark compile times using GCC would be daunting and unrealistic. Not a lot of people are using GCC to work on the Windows side of things. If Linux could be thrown into the equation, I’d suggest a project based on CMake. That would make it somewhat easy to write a simple script to setup, create a makefile and compile the project. Unfortunately, I can not readily think of any big name projects such as Chromium that fulfill that requirement without having to meddle with eventual dependency problems as the time goes by.
Kevin G - Monday, July 24, 2017 - link
These chips edge out their LGA 1151 counter parts at stock with overclocking also carrying a slight razor edge over LGA 1151 overclocks. There are gains but ultimately these really don't seem worth it, especially in light of the fragmentation that this causes the X299 platform. Hard to place real figures on this but I'd wager that the platform confusion is going to cost Intel more than what they will gain with these chips. Intel should have kept these in the lab until they could offer something a bit more substantial.mapesdhs - Monday, July 24, 2017 - link
I wonder if it would have been at least a tad better received if they hadn't cripplied the on-die gfx, etc.DanNeely - Tuesday, July 25, 2017 - link
LGA2066 doesn't have video out pins because it was originally designed only for the bigger dies that don't include them; and even if Intel had some 'spare' pins it could use adding video out would only make already expensive mobos with a wide set of features that vary based on the CPU model even more expensive and more confusing. Unless they add a GPU to either future CPUs in the family (or IMO a bit more likely) a very basic one to a chipset variant (to remove the crappy one some server boards add for KVM support) keeping the IGP fully off in mainstream dies on the platform is the right call IMO.DrKlahn - Monday, July 24, 2017 - link
Great article, but the conclusion feels off:"The benefits in the benchmarks are clear against the nearest competition: these are the fastest CPUs to open a complex PDF, at the top for office work, and at the top for most web interactions by a noticeable amount."
In most cases you're talking about a second or less between the Intel and AMD systems. That will not be noticeable to the average office worker. You're much more likely to run into scenarios where the extra cores or threads will make an impact. I know in my own user base shaving a couple of seconds off opening a large PDF will pale in comparison to running complex reports with 2 (4 threads) extra cores for less money. I have nothing against Intel, but I struggle to see anything in here that makes their product worth the premium for an Office environment. The conclusion seems a stretch to me.
mapesdhs - Monday, July 24, 2017 - link
Indeed, and for those dealing with office work it makes more sense to emphasise investment where it makes the biggest difference to productivity, which for PCs is having an SSD (ie. don't buy a cheap grunge box for office work), but more generally dear god just make sure employees have a damn good chair to sit on and a decent IPS display that'll be kind to their eyes. Plus/minus 1s opening a PDF is a nothingburger compared to good ergonomics for office productivity.DrKlahn - Tuesday, July 25, 2017 - link
Yeah an SSD is by far the best bang for the buck. From a CPU standpoint there are more use cases for Ryzen 1600 than there is the i5/i7 options we have from HP/Dell. Even the Ryzen 1500 series would probably be sufficient and allow even more per unit savings to put into other areas that would benefit folks more.JimmiG - Monday, July 24, 2017 - link
The 7740X runs at a just over 2% higher clock speed than the 7700X. It can overclock maybe 4% higher than the 7700X. You'd really have to be a special kind of stupid to pay hundreds more for an X299 mobo just for those gains that are nearly within the margin of error.It doesn't make sense as a "stepping stone" onto HEDT either, because you're much better off simply buying a real HEDT right away. You'll pay a lot more in total if you first get the 7740X and then the 7820X for example.
mapesdhs - Monday, July 24, 2017 - link
Intel seems to think there's a market for people who buy a HEDT platform but can't afford a relevant CPU, but would upgrade later. Highly unlikely such a market exists. By the time such a theoretical user would be in a position to upgrade, more than likely they'd want a better platform anyway, given how fast the tech is changing.