Overclock.net - An Overclocking Community

Overclock.net - An Overclocking Community (https://www.overclock.net/forum/)
-   AMD CPUs (https://www.overclock.net/forum/10-amd-cpus/)
-   -   Updated! Ryzen R5 Unleashed - CPU & RAM OC tests (https://www.overclock.net/forum/10-amd-cpus/1721340-updated-ryzen-r5-unleashed-cpu-ram-oc-tests.html)

B4rr3L Rid3R 02-25-2019 08:39 AM

Updated! Ryzen R5 Unleashed - CPU & RAM OC tests
 



Ryzen performance comparison in different scenarios:

Dual-Channel vs Single-Channel (16GB sticks)
Manual memory OC 2933Mhz CL14 vs Default Auto Bios setting
Default Bios CPU clock vs 3.7GHz CPU OC
Games & Applications side by side
Results table and graphic

- 7-Zip
- Blender BMW 2019
- Strange Brigade DX12
- Dirt Rally DX11
- Metro Exodus DX12
- Handbrake
- Aida 64
- Forza 4 horizon
- War Thunder
more...

Conclusions:

1 - After 2866Mhz CL14 and forward the difference between Single-channel and Dual is minimal except for 7-Zip.
2 - For most games Memory clock and timing have more impact than dual channel against single.
3 - You can unleash up to 30% performance gains by tweaking your PC properly, that could be achieved while gaming with extra FPS (30 FPS increase in Dirt Rally!) and also while heavy working profissionally.
4 - better timing do help with performance but not as much as memory clocks.

Undervolter 02-28-2019 12:46 PM

Thank you for testing applications too and especially 2 of my favourites, 7zip and Handbrake! I get tired of seeing people worrying only about games.

rdr09 03-01-2019 12:39 AM

1 Attachment(s)
+Rep. Nice results, especially the increase in fps. Imagine if you can achieve a stable 3466 Cl14 at least for games, which I did and got a 61 in Aida using G.Skill FlareX. Currently using a cheaper G.Skill Ripjaws on my 1600. Without monitoring the fps, I don't see much any difference between the two rams, though. Just in benchmarks.

cssorkinman 03-01-2019 06:49 AM

Quote:

Originally Posted by rdr09 (Post 27873918)
+Rep. Nice results, especially the increase in fps. Imagine if you can achieve a stable 3466 Cl14 at least for games, which I did and got a 61 in Aida using G.Skill FlareX. Currently using a cheaper G.Skill Ripjaws on my 1600. Without monitoring the fps, I don't see much any difference between the two rams, though. Just in benchmarks.

Did you happen to compare draw calls using futuremark's api feature test?

rdr09 03-01-2019 10:58 AM

Quote:

Originally Posted by cssorkinman (Post 27874138)
Did you happen to compare draw calls using futuremark's api feature test?

I just did but idk how to interpret them.

CPU stock RAM at 3200 Cl14 DOCP

https://www.3dmark.com/3dm/34093435?

CPU stock RAM at 3466 Cl14 using Calculator.

https://www.3dmark.com/3dm/34093811?

The DX11 MT actually went down. Maybe my 3466 setting is not fully stable.

Alpi 03-01-2019 11:20 AM

4 Attachment(s)
Absolutely ! Ryzen needs good memory to reach better performance. Actually I've never seen anything like this stuff ! ;) Performance gains from memory speeds is awesome ! A few days earlier made some test, here are some FC5 benchmark results. Nothing changed but mem speeds. 2133, 2800, 3466 mhz. Timing was 14-14-14-28 1T. I could manage this 3466c14 for 24/7 stable quite easy fortunately. :)

cssorkinman 03-01-2019 05:04 PM

Quote:

Originally Posted by rdr09 (Post 27874464)
I just did but idk how to interpret them.

CPU stock RAM at 3200 Cl14 DOCP

https://www.3dmark.com/3dm/34093435?

CPU stock RAM at 3466 Cl14 using Calculator.

https://www.3dmark.com/3dm/34093811?

The DX11 MT actually went down. Maybe my 3466 setting is not fully stable.

Certainly possible it wasn't completely stable , it's also a good idea to run it multiple times and average the results or kick out the high and low.

With first gen Ryzen I'd realize most of the gains by 2933 mhz - allowing the motherboard to set the timings for speeds of comparisons of 2133mhz and up.

Quote:

Originally Posted by Alpi (Post 27874502)
Absolutely ! Ryzen needs good memory to reach better performance. Actually I've never seen anything like this stuff ! ;) Performance gains from memory speeds is awesome ! A few days earlier made some test, here are some FC5 benchmark results. Nothing changed but mem speeds. 2133, 2800, 3466 mhz. Timing was 14-14-14-28 1T. I could manage this 3466c14 for 24/7 stable quite easy fortunately. :)

Nice aida scores - FX rigs would make similar gains going from cl6 1600 to cl 10 2400 + , depending on the app and what was actually limiting fps.

AlphaC 03-01-2019 08:45 PM

I would run the same test on R7 1700X with 3466C14 but I'm on Linux and I don't have a RX 580.

API overhead with 3200C16 on my R7 2700X isn't that impressive though with AMD pro drivers.

DX 11 single thread over 2 Million is decent

https://www.anandtech.com/show/11223...k-api-overhead
https://www.tomshardware.com/reviews...ng,4977-3.html
https://www.tomshardware.com/reviews...pu,5025-3.html
https://www.tomshardware.com/reviews...pu,5107-3.html

https://www.expreview.com/56597.html
i7-7700K + RX Vega 56 ~ 1.9Mil


------------


Random browsing the database:
[email protected] + TITAN V , 2.8Million DX11 single thread and ~4.55million multi https://www.3dmark.com/aot/239401
[email protected] + GTX 1080 ,3.2 million DX11 single thread and ~5million multi https://www.3dmark.com/aot/234573
[email protected] + gtx 1080 , 3.5million DX 11 single thread and ~ 5.3 million multi https://www.3dmark.com/aot/256712
[email protected] + VEGA56 , 2.6million Dx11 single thread and ~2.3 million multi https://www.3dmark.com/aot/259545
[email protected] + GTX 1070 , 2.56million DX11 single thread and ~2.64 million multi https://www.3dmark.com/aot/200368?locale=en_GB
[email protected] + GTX 1080 , 2.5million DX11 single thread and ~3.7million multi https://www.3dmark.com/aot/207801?locale=en_GB
[email protected] + GTX 1070 , 2.9million Dx11 single thread and ~4.3million multi https://www.3dmark.com/aot/220187
[email protected] + RX 470, 2.3 Million Dx11 single thread and ~2.3 million multi https://www.3dmark.com/aot/202653
[email protected] + GTX 1080 , 2.15 Million DX11 single thread and 2.75 million multi https://www.3dmark.com/aot/181927?locale=en_GB
i7-7900X @4.3GHz + GTX 1080 , 2.2 Million DX11 single thread and ~ 4.1 million multi https://www.3dmark.com/aot/221086
i7-5960X @4.4GHz + GTX 1080 , 2.2 Million DX11 single thread and ~2.5million multi https://www.3dmark.com/aot/201305
[email protected] + GTX1080 , 1.88million Dx11 single thread and ~2 million multi https://www.3dmark.com/aot/205294
[email protected] + RX 470 , 2.25 million DX11 single thread and ~ 2.1 million multi https://www.3dmark.com/aot/199856


(Ryzen)

R7 2700X @4.5GHz + GTX 1080 ti, 2.64 million single thread and ~4.47million multi https://www.3dmark.com/aot/284455
R7 [email protected] + GTX 1080 ti , 2.2Million Dx 11 single thread and ~3.9 million multi https://www.3dmark.com/aot/251084
R7 1700 @3.8GHz + GTX 1070 , ~2.08Million Dx11 single thread and ~3.5 million multi https://www.3dmark.com/aot/229943
R7 [email protected] + GTX 1080 , 2million DX11 single thread , ~2.2 million multi https://www.3dmark.com/aot/199978
R7 [email protected] + Vega 64 , ~1.65 million DX11 single thread and ~1.65 million multi https://www.3dmark.com/aot/293630
R7 [email protected] + GTX 1080 ti , ~2.4 million DX11 single thread and ~4.1million multi https://www.3dmark.com/aot/222809
R7 1800X + GTX 1070 , 1.97 million DX11 single thread and ~2 million multi https://www.3dmark.com/aot/199230?locale=en_GB
TR [email protected] + Titan XP , ~1.87 million DX11 single thread and ~2.4 million multi https://www.3dmark.com/aot/253817
TR 1920X @4GHz + VEGA 64 ~1.36Million single thread and ~1.36Million multi https://www.3dmark.com/aot/286707

rdr09 03-01-2019 11:32 PM

Quote:

Originally Posted by cssorkinman (Post 27874872)
Certainly possible it wasn't completely stable , it's also a good idea to run it multiple times and average the results or kick out the high and low.

With first gen Ryzen I'd realize most of the gains by 2933 mhz - allowing the motherboard to set the timings for speeds of comparisons of 2133mhz and up.

Something wrong with 3DMark on my 2700 system. I ran same test on my R5 1600 and got 2.7M in DX 11 MT. Thing is, even in Firestrike, the 2700 used to get a physics score of 17K at stock, now just 12K. It does not ffect my games so no big deal.

Problem with the R5 1600, it won't run the Vulkan test. Just blinks. Could it be my GTX 1060?

AlphaC 03-01-2019 11:55 PM

Are you running your drivers on multithread (threaded optimization) with shader cache?


All times are GMT -7. The time now is 09:46 PM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.

User Alert System provided by Advanced User Tagging (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
vBulletin Security provided by vBSecurity (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.