• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

smilingcrow said:
I did wonder!
If anything CPU usage might go up as it has data coming in faster it can process more so usage goes up.
Click to expand...

No because the actual data been processed is the same given its a fixed 60fps, not uncapped framerate. If a cpu (or gpu) is waiting for data it has a i/o wait chewing cycles.

I have measured 2 games vesperia and division 2, both fixed 60fps.

If I am going to be accused of BS next time I just wont share my data, thats what I call a toxic community to just accuse someone of BS especially when someone has done nothing themselves just likely dialled in XMP and left it at that so dont have their own data points.
 
Maybe you could prove a bit better with the built in tester in Ryzen DramCalculator.....

v1.5.0 (May 7th, 2019)
Main Changes

  • Added feature to compare current system timings with recommended ones. Turns on with the button "Compare timings".
  • Added the ability to select the topology of the motherboard. Sheet box with the name "Topology". The meaning of this undertaking is a more accurate prediction of procODT / RTT settings. Of course, the function is not perfect and with each update of the calculator the prediction accuracy will increase.
  • Partially cleaned the "Main" tab of information that is not used and clutters the user interface. This is a small advance towards the ease of use of this application.
  • SOC voltage prediction blocks have undergone significant changes for each generation of Ryzen processors.
  • Another recommendation is temporarily added to the CAD_BUS block.
  • Recycling of most presets for each type of memory. The reason is simple - new bios make new adjustments.
  • Block "Misc items" received new prediction algorithms. In particular, the GDM prediction.
  • Improved support for system configurations consisting of 4 DIMMs.
  • Improved algorithm for calculating profiles "Debug".
  • Corrections of sudden crash programs in the calculations.
  • Fixed a problem when during the import of the html profile the memory type was automatically switched to Samsung b-die mode.
  • Added button link to system setup guide using Ryzen DRAM Calculator. This button is located in the "Help" tab.
  • Updated information in the "About" tab and added feedback to me via Twitter.
Key features of MEMbench

  • Support for processors with 1 to 32 threads (temporary restrictions) .
  • Error within 0.5%.
  • Automatic determination of the number of threads in the system.
  • Unified rating for desktop and HEDT processors (AMD and Intel).
  • Qualitatively implemented multi-threading.
  • Thanks to the HCI 6 kernel (freeware), this product is free for the end user. I also want to note that this version is not modified (does not contain hacks).
  • Complete absence of the influence of the processor frequency on the benchmark result.
  • The result is affected by each timing (picture with themes, see below).
  • Ability to work in the mode of ordinary memory check with a fixed delayed stop at a certain stage and this stage is determined by you.
  • Four presets for benching "Easy mode", "Default mode", "Custom mode" and "Memtest".
  • Easy mode is designed specifically for systems with a small amount of installed RAM.
  • Easy to use benchmark, you need to select only MEMbench mode and click "Run".
  • Provision of full-fledged system information for the Ryzen processors (for Intel, the timing-determining block is not yet connected).
  • Ability to save two own results for further comparison.
  • Display information in the chart.
  • A window-table with information about the current state of benching or memory checking.
  • "Screenshot" Button
  • All possible protection against situations of low memory in the system.
  • Does not use swap file.
  • Loyal stress CPU test in "Default mode".
 
chrcoluk said:
No because the actual data been processed is the same given its a fixed 60fps, not uncapped framerate. If a cpu (or gpu) is waiting for data it has a i/o wait chewing cycles.

I have measured 2 games vesperia and division 2, both fixed 60fps.

If I am going to be accused of BS next time I just wont share my data, thats what I call a toxic community to just accuse someone of BS especially when someone has done nothing themselves just likely dialled in XMP and left it at that so dont have their own data points.
Click to expand...
I really wouldn't bother about what a lot of peeps say about ram, ram throughput, utilsation, timings or tweaking on here. 99% only ever set xmp because they have no idea at all how to clock ram. It takes too long and a lot of thinking to do it properly. Most have not got the time, patience or intelligence to do it.................they never have done in the past and as time goes on it becomes even more obvious newer generations are even more lazy where ram is concerned.[/QUOTE]
 
chrcoluk said:
No because the actual data been processed is the same given its a fixed 60fps, not uncapped framerate. If a cpu (or gpu) is waiting for data it has a i/o wait chewing cycles.
I have measured 2 games vesperia and division 2, both fixed 60fps.
Click to expand...
If it's processing the same amount of data in the same time frame then not sure why real-world utilisation is 50% less?
It sounds unlikely to me but without further info I can't comment in a meaningful way beyond a gut reaction.
 
smilingcrow said:
If it's processing the same amount of data in the same time frame then not sure why real-world utilisation is 50% less?
It sounds unlikely to me but without further info I can't comment in a meaningful way beyond a gut reaction.
Click to expand...

this is fine its a polite opinion, the issue I had is with someone just coming out calling me a BS'er.

If something is done faster then utilisation will be lower as its measured by idle time, something done faster means more idle time.
 
kitfit1 said:
I really wouldn't bother about what a lot of peeps say about ram, ram throughput, utilsation, timings or tweaking on here. 99% only ever set xmp because they have no idea at all how to clock ram. It takes too long and a lot of thinking to do it properly. Most have not got the time, patience or intelligence to do it.................they never have done in the past and as time goes on it becomes even more obvious newer generations are even more lazy where ram is concerned.
Click to expand...

I know, ram is something that has been massively undertested vs other components, but what I do know is many assumptions get made based on synthetic benchmark tests. In recent years we have had some people do game testing with no visible difference but then other people do game testing with significant difference, tests contradicting each other, those test themselves are not great howver as they just using same timings on different clock speeds so the tests improve both latency and throughput so its unknown which is providing the benefit, but they are useful to know is "some" benefit.

I did my own personal testing and got the results I got, I decided to share them, thats all. As you said generally people plug in the ram, either leave at Jedec, or enable XMP and then leave it after. I expect the vast majority of people unlike cpu and gpu o/c dont do any performance testing of before and after.
 
chrcoluk said:
I know, ram is something that has been massively undertested vs other components, but what I do know is many assumptions get made based on synthetic benchmark tests. In recent years we have had some people do game testing with no visible difference but then other people do game testing with significant difference, tests contradicting each other, those test themselves are not great howver as they just using same timings on different clock speeds so the tests improve both latency and throughput so its unknown which is providing the benefit, but they are useful to know is "some" benefit.

I did my own personal testing and got the results I got, I decided to share them, thats all. As you said generally people plug in the ram, either leave at Jedec, or enable XMP and then leave it after. I expect the vast majority of people unlike cpu and gpu o/c dont do any performance testing of before and after.
Click to expand...

I've done my fair share of ram overclocking to have an informed opinion on this. I was the only person on here to have got 3466 Cl14 from a 1700. I also out benched 8pack on his ram benchmarks on ryzen but I digress. You do not half your cpu utilisation by simply adjusting timings. I think your testing is flawed or you had a placebo moment.
 
There may well be something else having an affect I of course accept that, although you cant really placebo readings, placebo is when something "feels faster", whilst I am actually taking measurements as reported to me by software.

Really ultimately what needs doing is more people doing these types of testing, so there is more data out there to compare to.

Also I have only measured 2 games, I have not tested baseline windows, browser, encoding, and other things, just 2 games.

Also important to note I am not claiming everything I do has reduced cpu usage. The type of data been processed will have an affect on how it may be affected.

Dont really want this to drag on as I am not going to claim this is scientifically 100% correct, as there is variables involved that could have impacted things, but I will stand by that I didnt just come on here to BS people. I will accept tho that there is possible flaws in the claims made. As division 2 been the game it is I wouldnt be measuring the exact same scenes, gameplay etc.

Feel free to open another thread tho if you want to discuss more about your own testing (which you havent detailed what you did, and if you want me to run any specific tests, you just stated you achieved high overclocks which by itself doesnt mean a whole lot, if you never tested the affect of that), but by your post you seem to be suggesting you feel memory throughput is more important than latency, whilst I consider the opposite is more important.
 
Last edited:
Benchmarking is tough to get right as you have to match conditions, I did some benchmarking of my ram timings to see difference but it was not lab conditions and done over different days, one of my results on superposition is something I struggle to replicate as I came in one morning it was very cold my water was nice and chilled so my system boosted clocks more than ever, really I needed to warm up the system and room but meh, I was only doing out of interest prior to chip upgrade.

my memory benchmark numbers here manual timings vs XMP at same clockspeed red and green bars are xmp vs tuned in same config
 
Last edited:
gavinh87 said:
I've done my fair share of ram overclocking to have an informed opinion on this. I was the only person on here to have got 3466 Cl14 from a 1700. I also out benched 8pack on his ram benchmarks on ryzen but I digress. You do not half your cpu utilisation by simply adjusting timings. I think your testing is flawed or you had a placebo moment.
Click to expand...

Are you this fellow?

https://www.youtube.com/watch?v=VEslh6hGpYQ
 
gavinh87 said:
I've done my fair share of ram overclocking to have an informed opinion on this. I was the only person on here to have got 3466 Cl14 from a 1700. I also out benched 8pack on his ram benchmarks on ryzen but I digress.
Click to expand...

I had 3550 working on the first AGESA. It comes down to alignment issues. Some CPUs would have been easier to work with at the time than others, too.
 
All the RAM talk goes right over my head

Never been into ram overclocking and miss the days where it didn’t matter too much or maybe it did and I was naive, either way I was one of those people who believed ram speed have almost no affect on game performance but it does. I just set xmp profile on mine for now, maybe I should try pushing it it is Samsung b die but I’m too lazy
 
You must log in or register to reply here.
Back
Top Bottom