DDR4 Haswell-E Scaling Review: 2133 to 3200 with G.Skill, Corsair, ADATA and Crucial
by Ian Cutress on February 5, 2015 10:10 AM ESTProfessional Performance: Windows
Agisoft Photoscan – 2D to 3D Image Manipulation: link
Agisoft Photoscan creates 3D models from 2D images, a process which is very computationally expensive. The algorithm is split into four distinct phases, and different phases of the model reconstruction require either fast memory, fast IPC, more cores, or even OpenCL compute devices to hand. Agisoft supplied us with a special version of the software to script the process, where we take 50 images of a stately home and convert it into a medium quality model. This benchmark typically takes around 15-20 minutes on a high end PC on the CPU alone, with GPUs reducing the time.
Photoscan, on paper, would offer more possibilities for faster memory to make a difference. However it would seem that the most memory dependent stage (stage 3) is actually a small part of the overall calculation and was absorbed by the natural variation in the larger stages, giving at most a 1.1% difference between times.
Cinebench R15
Cinebench is historically CPU dependent, giving a 2% difference from JEDEC to peak results.
3D Particle Movement
3DPM is a self-penned benchmark, taking basic 3D movement algorithms used in Brownian Motion simulations and testing them for speed. High floating point performance, MHz and IPC wins in the single thread version, whereas the multithread version has to handle the threads and loves more cores.
3DPM is also relatively memory agnostic for DDR4 on Haswell-E, showing that DDR4-2133 is good enough.
Professional Performance: Linux
Built around several freely available benchmarks for Linux, Linux-Bench is a project spearheaded by Patrick at ServeTheHome to streamline about a dozen of these tests in a single neat package run via a set of three commands using an Ubuntu 14.04 LiveCD. These tests include fluid dynamics used by NASA, ray-tracing, molecular modeling, and a scalable data structure server for web deployments. We run Linux-Bench and have chosen to report a select few of the tests that rely on CPU and DRAM speed.
C-Ray: link
C-Ray is a simple ray-tracing program that focuses almost exclusively on processor performance rather than DRAM access. The test in Linux-Bench renders a heavy complex scene offering a large scalable scenario.
Natural variation gives a 4% difference, although the faster and more dense memory gave slower times.
NAMD, Scalable Molecular Dynamics: link
Developed by the Theoretical and Computational Biophysics Group at the University of Illinois at Urbana-Champaign, NAMD is a set of parallel molecular dynamics codes for extreme parallelization up to and beyond 200,000 cores. The reference paper detailing NAMD has over 4000 citations, and our testing runs a small simulation where the calculation steps per unit time is the output vector.
NAMD showed little difference between our memory kits, peaking at 0.7% above JEDEC.
NPB, Fluid Dynamics: link
Aside from LINPACK, there are many other ways to benchmark supercomputers in terms of how effective they are for various types of mathematical processes. The NAS Parallel Benchmarks (NPB) are a set of small programs originally designed for NASA to test their supercomputers in terms of fluid dynamics simulations, useful for airflow reactions and design.
Despite the 4x8 GB results going south of the border, the faster memory does give a slight difference in NPB, peaking at 4.3% increased performance for the 3000+ memory kits.
Redis: link
Many of the online applications rely on key-value caches and data structure servers to operate. Redis is an open-source, scalable web technology with a b developer base, but also relies heavily on memory bandwidth as well as CPU performance.
When tackling a high number of users, Redis performs up to 17% better using 2800+ memory, indicating our best benchmark result.
Facebook
Twitter
RSS
120 Comments
View All Comments
Dasa2 - Thursday, February 5, 2015 - link
To back up some of what i said here is a few linksI3 2100 matching 2500k@4ghz in dirt 3
http://www.tomshardware.com/reviews/gaming-fx-pent...
Arma a cpu bottlnecked game where a 2600k@4.3ghz with 2133c9 ram is faster than at 4.9ghz with 1600c11
http://forums.bistudio.com/showthread.php?166512-A...
Thief CPU|RAM performance
http://forums.atomicmpc.com.au/index.php/topic/557...
Bf4 1600c9=60fps 2400c10=70fps
http://www.team-greatbritain.com/call-of-duty-ghos...
Xbit ddr3 review looks a bit different to yours...
http://www.xbitlabs.com/articles/memory/display/ha...
Margalus - Friday, February 6, 2015 - link
And not one of those is using ddr4...Dasa2 - Friday, February 6, 2015 - link
Nope hence why I would like a decent review site like anandtech to do a proper job of there ddr4 reviewIm not expecting a big of a difference from higher speeds quad channel ddr4 by comparison to what can be seen in dual channel ddr3 but even there haswell ddr3 tests showed jack all due to the same problem with there tests so how can we know for sure
FlushedBubblyJock - Sunday, February 15, 2015 - link
You're correct, you made your points, so of course someone without many watts currently on display there said something silly, as usual being stupid pays off and those not dumbed down to base below average levels suffer the frustrating beyond belief consequences.mrcaffeinex - Friday, February 6, 2015 - link
The problem is that we currently do not have a non-enthusiast platform available that supports DDR4. The new X99 platform is also running quad-channel, so the best comparison to a prior platform would have to be using X79 (attempting to keep as close to apples to apples as possible). The point that can be taken from this article as it is right now, is that you can skip buying insanely-priced DDR4-3000+ memory because your X99 rig will probably not perform noticeably different with DDR4-2133.As the process matures and more systems adopt DDR4, then you'll be able to do a better comparison across multiple performance levels, but as it is right now, if you're buying into X99, you're buying a high-end CPU. I look forward to the extensive comparative tests that you have mentioned, but I do not see them happening until either the mainstream platform (LGA 115x) is running DDR4 or AMD has any offering that supports DDR4.
Dasa2 - Friday, February 6, 2015 - link
Unfortunately you cant take that from this article as the gaming tests wouldnt show if there was any gain from faster ram even if it did boost cpu performance by 15%These tests were worse than a complete waist of time from a gaming perspective as they could be very misleading
At a guess i would expect to see somewhere between 3-7% difference going from ddr4 2133 to ddr4 3200 at the same timings although most of that gain will probably be between 2133 and 2666 happy to be proven wrong though
Sushisamurai - Friday, February 6, 2015 - link
although I agree it would be nice to see the impact DDR4 timings and speeds on CPU bound games, I unfortunately don't see the real world application to it. With DDR4, we're working on Haswell-E, which already has a lot of compute power - if we were to run into any CPU bottlenecks, wouldn't it make more sense to spend more of the budget into the CPU instead of RAM? Unless, you had enough money to buy top CPU and top RAM, then the point becomes quite moot no?Dasa2 - Friday, February 6, 2015 - link
Depends how big the gain is from faster ram doesnt it and we wont know that until its tested properly with the ram speed compared cpu speeds toTesting cpu or ram performance with gpu bottleneck games is a waist of time unless your AMD trying to sell fx8150...
The only cpu limited games at this stage on Haswell-E will be the ones with bad multithreading support so spending a heap more on the cpu for extra cores from the 5960x wont help
What will help is spending extra for a better overclock and maybe faster ram but how far do you go
tim851 - Friday, February 6, 2015 - link
> The games you chose to review are so badly GPU bottlenecked its sad.That's why they were running these games at reduced resolutions and IQ settings, Einstein.
What game should Anandtech benchmark that is NOT GPU LIMITED - Quake 3 Arena?
Dasa2 - Friday, February 6, 2015 - link
They shouldnt reduce detail settings just no aa and resolution to 1080p while running a gtx980 or two (r9-290\gtx970\gtx780oc minimum)But with the likes of dirt 3 even if they do reduce detail settings its still gpu bottlnecked
Arma\Dayz are some of the only games that can be cpu bottlnecked with a single gtx770
Dying Light is very demanding on both cpu and gpu
http://translate.googleusercontent.com/translate_c...
There is a lot of games that can be a bit of a blend of cpu\gpu limitation with enough gpu power although most of these will run 60fps fine on a 5820k a fair few of them wont do 120-144fps
http://translate.google.com/translate?depth=6&...
As they are a blend there limitation can vary from one part of the game to the next for example testing BF4 SP although easier to get consistent results will be far more gpu limited than MP some levels will also be more gpu limited than others
This is why i suggest putting different models and clocks speeds of cpu in against ram speed results so that people can see where the limit really is and where money is best spent