Memory Speed Scaling and Windows 10 Power Modes
AMD has been pushing the use of higher speed memory with the Ryzen 5 reviews in order to compensate for some of the performance impact and deltas seen in the Ryzen 7 reviews. I have already noted the potential conflict in regard to cost, but how does increased memory speed improve an array of our testing applications?
Memory speed improvements are negligible from 2400 MHz to 3200 MHz in Cinebench single thread and multi-thread iterations, Handbrake H.264 encoding, Blender rendering, and one game, Ghost Recon: Wildlands. Geekbench shows reasonable performance increases in both single threaded and multi-threaded results, indicating that some workloads (compression, speech recognition) can still improve with faster fabric/memory.
The gaming results show 12-14% jumps when running at 3200 MHz compared to 2133 MHz. That is clearly a big enough gap to warrant the purchase of 3200 MHz memory and to push the adoption of those higher speeds for general usage. Intel’s architecture doesn’t require higher speed memory and sees very little uptick from the higher speed, higher cost memory.
Windows 10 Power Mode Testing
I did a considerable amount of testing to compare the Balanced and High Performance modes in Windows 10 with Ryzen 5, in this case we are showing results from the 1500X. There were only three data points of interest – WebXPRT (+14%), Cinebench ST (+4.7%) and Rise of the Tomb Raider (+3.6%). In all other instances the difference between the two settings was well within the margin of error and I would view the positive results as the outliers. The WebXPRT result is the MOST interesting to me as it is indicative of an inherent advantage in responsiveness without the core parking concerns of the Windows 10 Balanced profile.
OK – I’ll go full defense of
OK – I’ll go full defense of PCPer. I’m not sure why there’s mega trolling here. This review is a very good technical review because:
– Only site with core ping times, giving detail to the architecture
– Memory scaling and Power mode in simple, useful format
– They listened – charts are now higher contrast, easier to see
– Good mix of useful benchmarks (audio, video, games)
– Sale links to all products used is always good
– A Video summary adds a lot to an article like this imo
Could use work:
– Please show 12/16 thread scaling in 7-Zip in future
– How does this compare to the Geekbench estimate from September? 🙂
– A bit too many synthetics (Cinebench, SiSoft, Geekbench) that i’m not sure actually mean anything
– Where’s the $180 i3-7350K and $60 Pentium G4560? 🙂
I look forward to the commentary on the podcast.
Thanks for the feedback and
Thanks for the feedback and positive input. 🙂
tom hardware also layed out
tom hardware also layed out the latency incase you didn’t notice, just not a graph but a table chart. but imo tom is a bit biased towards intel hardware, testing benchmark at 5ghz but when comes to power consumption, they test it at default clock instead of 5ghz, biased and misleading.
Toms showed OC results but
Toms showed OC results but they also included base clock speed as well on their charts. They overclocked both sides as well. Seemed legit to me.
Looks pretty good overall.
Looks pretty good overall. I’m looking forward to what tweaks they do on Zen+ to iron out some of the downsides. These chips would be unbelievable if they were a monolithic chip.
Good to know that, even with
Good to know that, even with the knowledge that RotR DX12 is an anomaly in that Nvidia has shit DX12 drivers and it tanks the Ryzen CPU performance, PCPer is still trucking along only using Nvidia GPUs in their test bed and are using a benchmark that have been confirmed to be unreliable at best.
Sad really. I expected better.
The NV GPU used is still far
The NV GPU used is still far faster than any available AMD part. When testing CPUs, ensure the GPU is not the bottleneck (or the least amount possible). Spinning up a different testbed config for a single game is not something we have enough hours in the day for.
It might be interesting to
It might be interesting to see how many people are running Nvidia cards with their Ryzen processors. You are, to some extent, testing Nvidia’s drivers and they are not going to be optimized for AMD processors. I am not saying that it isn’t a valid test, since there certainly will be people running Nvidia cards. I would like to see some numbers with lower end cards though, especially with the more mid-range to low end parts to see if those differences actually will be noticeable. With this high end card, the differences, even when extreme are often a difference between something like 80 and 100 fps. While this is noticeable, it isn’t really much of an issue. I would expect the gaps to be smaller for lower performance video cards, but without testing, we don’t know how much smaller.
Also, if anyone has time and a test setup it would be interesting to run the AotS benchmark before and after the patches with different memory speeds. I am wondering if the performance boost from higher memory clock disappears with code more optimized for Ryzen’s cache set-up.
Could you change the your
Could you change the your latency chart look something closer to a correlation matrix? I don’t know if that’s the correct term, but presenting the data in that format would better show how local pings are faster than cross-CCX pings.
I would expect the diagonal to be close to 0, the upper left and bottom right quarters to be fast, and the lower left and upper-right quarters to be slow
You could format it something like this but I expect the data to look like this
I like that style, let me
I like that style, let me look into it!
@Ryan I have spotted a small
@Ryan I have spotted a small typo in the legend of graph ‘Ryzen 5 Memory Speed Tesing’ on page Memory-Speed-Scaling-and-Windows-10-P.
RedBrown is labeled DDR4-200 (but should probably be DDR4-2400)
Shit…you’re correct!
Shit…you're correct!
Great processor.
Great processor.
Also, I think that it will be
Also, I think that it will be a new word in the performance speed run. Now this build show great possibilities, but it is only a start. Ryzen architecture was developed to prove that you can get high performance for the considerable price. I ordered a research paper help about next microsoft configurations. Results were pretty impressive, but as we can see now, AMD is getting higher and higher in performance, but lowering price of components. In today’s market it is the only smart way to make business.
We can’t comment anymore
We can’t comment anymore without logging in. That really really sucks.
Really, I think it’s great so
Really, I think it's great so far. 🙂
Yay no more anonymous
Yay no more anonymous AMD-heads
Yes. You’re right. It was
Yes. You’re right. It was *only* AMD fans causing problems in the comments. Anonymous Nvidia fans *never ever ever* randomly abused people here.
*sigh*
I was talking about this
I was talking about this thread but yeah sigh away
Seriously, truly, from the
Seriously, truly, from the bottom of my heart, thank you guys for finally making this happen. 🙂
I love it. It’s like we can
I love it. It’s like we can actually have discussions in the comments now, instead of having to plod through 250 post of abusive fanboy nonsense in the hopes of finding a comment of substance.
It took way too long, but I’m so glad they finally did it that I might start coming back to the site again. (Today is my first visit in months.)
You and the trolls are the
You and the trolls are the only people on that side of the argument 😀
What a sh*t show the comments
What a sh*t show the comments turned into.
Not sure why all the AMD fans feel the need to nitpick every last little bit to find something that invalidates the results. The conspiracy theories run so deep its hilarious. Tests take time, money and effort. The reviewers aren’t perfect, they might not be able to do every last test between receiving a sample and embargo lifting… thats a fact of life, not some deep-seated bias.
Bottom line, when I look at PcPer’s results for the Ryzen 5 I see the same thing I’ve seen at every other review site: a compelling product for AMD that doesn’t win every test. Finally, you can buy an AMD chip instead of the i5 and not feel like you were “taking one for the team”. The fact that it has weaknesses is objective fact, its up to the reader to decide whether or not those weaknesses are an issue for them.
Thank you sir.
For someone
Thank you sir.
For someone with AMD in their username, I appreciate the input.
Ah, don’t read too much into
Ah, don’t read too much into that 😉 Been a forum member under that nick since 2003 when this place was AMDMB and figured I’d keep it on this side of the fence for the sake of continuity
Well then all we can say
Well then all we can say there is thank you for sticking around for so long! 🙂
It’s not just
It’s not just “AMD-Heads”…PC v Console, FPS v RPG, Pinky v The World, and so on. Removing the anonymity on technical sites removes zealots w/o perspective…which is smart. The most interesting thing to watch for here isn’t how much tweaking AMD does over the next few months…it’s how much does Intel respond to and how they respond. Lowering prices will not get them back to a dominating lead they grew complacent with.
I would agree and I would not
I would agree and I would not expect them to lower prices.
To differentiate your review
To differentiate your review you could do this.
Graph the CPU / GPU usage during your gaming test, along side frame time.
Because nobody test CPU/GPU scaling anymore buyer need to guess if the CPU is good enough for the GPU, (or vice versa) and have no clue how well the CPU will hold up in the future.
For example if you see “i5 get 85FPS, r5 gets 80fps”
You would choose the i5, right ?
But what if the i5 was 100% pegged to get 85fps, but the r5 only 40% ?
What makes me angry is no one
What makes me angry is no one is testing the differences with a Radeon card compared to the Nvidia drivers.
It’s a known issue, they ran the same benchmarks and game test’s on other sites with a Radeon card and showed better improvements on ryzen.
The issue is all your test’s are with nvidia cards who have drivers that react shitty with Ryzen.
http://i.imgur.com/OjSIcgM.jpg
https://youtu.be/QBf2lvfKkxA
https://youtu.be/0tfTZjugDeg
Ryan: Thank you for a fair
Ryan: Thank you for a fair review of the 1600x and 1500x. I have a 5960x with a GTX 1080 and a 6700K with a GTX980TI and just recently built a 1800x with RX480s in CF. I’m using an Asus Crosshair VI mb and Gskill FlareX DDR4-3200 ram so I’m able to get great performance out of the Ryzen. Tomorrow I will begin a 1600x build coupled with a GTX 1050ti.
Where AMD has finally shown it mettle and relevance is producing cpus that legitimately can stand toe to toe with Intel and have a chance. Quite an accomplishment.
You should probably mention
You should probably mention that the Ryzens have lower idle power consumption than the Intels in your last graph. That’s awesome. And it probably matters more (in probably most cases) than load consumption, thinking solely electricity bill, due to how much time is spent in either state.
Is it just me or did you list
Is it just me or did you list out the test bed components?
I have a friend looking to build a Ryzen 7 1700x with a B350 board and it would beneficial to know what DIMM’s were used to obtain the 2400 MHz – 3200 MHz speeds. As some DIMM cannot go above 2400 with the current BIOS out there from some vendors.