AMD RyZen CPU Architecture for 2017

The question is why the 720 has lower FPS than the 1080. Isnt the point of using 720 to get much higher FPS in first place?
 
Barclays apparently has 51 million worth of AMD stock options or some such due this Friday, so they need to get the stock down as much as possible to prevent loss
(I have no clue how this stuff actually works, just reporting what I heard from a fellow)
If they have stock and wants to buy one they will try to lower the price, if they want to sell they will try to rise it. At this point I wish I could have some money to spend to buy AMD stock now that is low, although with AMD you never know but with Lisa the company seems to be a clear north, leadership and products to get there. But its still risky or not as sure as others "opportunities" but I think theres a good chance AMD avoided become the next HTC.
 
The question is why the 720 has lower FPS than the 1080. Isnt the point of using 720 to get much higher FPS in first place?

This is a percentage rating, whatever their top result was, it counted as 100%
 
Testing 2666 MT/s with auto timings versus 3200 MT/s only changing primary timings versus my setup at 3466 MT/s changing both primary and secondary timings and I'm getting some very interesting results.

The change from 2666 to 3200 is smaller than 3200 to 3466

Will be posting the results as soon as I'm done.
 
For now I'll just post this, later on I'll post the rest (timings and game benchmark for RotTR and Far Cry Primal):

2666 MT/s with Auto timings from the BIOS: http://www.3dmark.com/3dm/21128455

2666_apirnxbo.png


3200 MT/s only changing primary timings to 14-14-14-34: http://www.3dmark.com/3dm/21128358

3200_apif2zy4.png


3466 MT/s changing primary timings to 14-14-14-28 and optimized sub-timings: http://www.3dmark.com/3dm/21129135

3466c14_apiwtlbd.png

The 1700 was at 4.0GHz for all of these tests, only changing memory speeds and latency.
 
Last edited:
Could you run Euler3D CFD benchmark with the same settings? This test is quite sensitive to memory and cache performance, since it process shared data objects on multiple threads.

It's a simple console application:
http://www.caselab.okstate.edu/research/euler3dbenchmark.html

Sure, this is the result with 3466C14 + tweaked sub-timings at 4.0GHz:

3466c14_euler3dopaui.png

At 3200 only changing primary timings:

3200_euler3d33yzn.png

2666 everything auto:

2666_euler3dlfacm.png
 
Last edited:
Thanks!
Zen is really craving mem/cache latency and bandwidth. Performance scaling shows it right there.
Hopefully AMD can reduce the memory latency a bit in Zen 2. They are slightly behind Intel there. These benchmarks show clearly that reducing memory latency would significantly help closing the gap between Intel in desktop workloads. The Zen core itself seems very good, they just need some "uncore" improvements. AMD said that there's lots of low hanging fruit left, so things are looking pretty good for them in the future.
 
Hopefully AMD can reduce the memory latency a bit in Zen 2. They are slightly behind Intel there. These benchmarks show clearly that reducing memory latency would significantly help closing the gap between Intel in desktop workloads. The Zen core itself seems very good, they just need some "uncore" improvements. AMD said that there's lots of low hanging fruit left, so things are looking pretty good for them in the future.

Latency and core frequency, if they manage to hit both of these then Zen should be on par with Kaby/Coffee in gaming tasks and ahead in MT tasks. Memory timings make a huge difference with Ryzen, and for whatever reason the auto timings set by AMD or board vendors are actually quite terrible :mad:

I'll post everything a bit later, my timings, the voltage that I'm using and so on.
 
The game results:

Rise of the Tomb Raider Dx12 (1080p Ultra preset):
rottrmountainpeakucyut.jpg
rottrsyriahoxlu.jpg
rottrgeothermalvalleyw6zkr.jpg


Far Cry Primal Benchmark(1080p Ultra + HD texture pack):
farcryprimalo3am5.jpg


Timings used on the memory:
2666 Auto:
2666_timingsjyx73.png


3200C14 only primary:
3200_timingsstbsy.png


3466C14 primary + sub-timings tweaked:
3466c14_timingslblxh.png

Voltage used with the Crosshair 6 Hero:

Dram: 1.45v + 1.45v boot dram voltage
Soc: 1.175v w/ LLC2
 
Last edited:
I did test Hitman Dx12 as well:

hitman1kmldj.jpg
hitman2bdafu.jpg


Final edit: It should be fine now, used the frame data the game provided in text form. The average when counting every frame is slightly higher than the game average function, possibly because they don't count outliers above a certain range. I used the excel min/max/average functions for those charts above across 12000+ frames for each run excluding the second frame which always was the lowest by far.
 
Last edited:
I did test Hitman Dx12 as well:

hitman1kmldj.jpg
hitman2bdafu.jpg


Final edit: It should be fine now, used the frame data the game provided in text form. The average when counting every frame is slightly higher than the game average function, possibly because they don't count outliers above a certain range. I used the excel min/max/average functions for those charts above across 12000+ frames for each run excluding the second frame which always was the lowest by far.
Can you do the same but imitating an 1500? I mean 4C8T. I'd like to know how it improves and how big of a difference it has.

33 fps more its a huge improvements...How long did it take you to set those timings correctly?
 
Can you do the same but imitating an 1500? I mean 4C8T. I'd like to know how it improves and how big of a difference it has.

33 fps more its a huge improvements...How long did it take you to set those timings correctly?

Quite a bit because I haven't done proper memory overclocking previously, I followed advice from The Stilt and his timings and tuned my memory a bit further. When I have time I'll do more tests.
 
Back
Top