[NEW MERCENARIES] Please note that all new forum users have to be approved before posting. This process can take up to 24 hours, and we appreciate your patience.

RYZEN 2nd Gen Performance + RAM Test

CloakshireCloakshire
Vindictus Vertreter: 4,580
Posts: 646
Mitglied
edited Dezember 21, 2018 in Tech Support
Here's a quick video showing how Ryzen Second Gen cpu's perform in Vindictus while hosting with multiprocessing enabled.
This video contains two ~30 second clips. The first is at stock ram frequency and the second shows the performance gained from a RAM overclock.



Game Settings Used:
USNUbrP.png


The important computer specs:
Processor: Ryzen 7 2700X (Stock Clock and Cooler)
Graphics: GTX 1070 Ti Founder's Edition
Primary Drive: Samsung 960 EVO M.2 250Gb SSD
Secondary Drive: Seagate 2TB 7200RPM HDD
Motherboard: MSI x470 Gaming Plus
RAM: G.SKILL Ripjaw V 3200Mhz 2x8 16Gb


Capture Information:
Recording Program used: Nvidia Shadowplay
Vindictus is stored on the SSD while videos are written to the HDD.


Thoughts: While there is a performance increase when overclocking the ram to 2933mhz, there is a cost. In both clips the CPU temp hovered around 53-54°C, but the CPU fan was spinning significantly faster with the RAM set to 2933mhz. This is due to the fact that the CPU's memory controller speeds up in tandem with RAM frequency thus causing more heat. It looks like the stock cooler was able to keep up and maintain the same temperature after overclocking the RAM, but the downside was that the CPU fan was spinning fast enough to hear. At stock RAM speeds (2133mhz), the CPU fan is nearly silent.

UPDATE 12/20/2018
Astera Raid Benchmarks at 2133, 2933, and 3200Mhz RAM speed configurations while hosting. Let me give a disclaimer first that the average fps number does not represent the max. In all tests, I did hit 60fps but it wasn't stable and would drop under certain conditions, but the drops never made the game hitch or become unplayable. For example, when Elchulus was all up in my face, my fps would drop but while he was in the air, I would hit ~60. I opted to not record video clips as it might hinder performance. But without further adieu, here's the results.

Individual Configuration Test Results

https://i.imgur.com/S5eDywv.jpg
https://i.imgur.com/Fx9kf6E.jpg
https://i.imgur.com/PDvXD80.jpg


Combined Test Results

BbyBYGO.png


Thoughts
Got some surprising results when testing out all three frequencies. As you can see in all the comparison chart, increasing the RAM frequency helped boost the fps in all tests to create a upward staircase effect.

Claire saw a minimal increase on the average since the slowest speed pulled 57fps leaving only ~3fps to gain and this was easily made up after the first step to 2933Mhz. On the minimum though, increased RAM speed gave a nice improvement. Going from stock to 2933Mhz yielded a ~10% gain while boosting it further to 3200Mhz yielded a ~19% gain over stock.

Elchulus is a demanding raid going by the results. My only guess is that it's partially due to the fact that he's a large boss plus his arena has a far draw distance. Anyways, going from stock, I saw an 8% gain on 2933Mhz and a 10% gain on 3200Mhz when comparing the average fps. The minimums tell a similar story with only a ~5% and ~10% gain at 2933 & 3200mhz respectively. Note that out of all the tests, Elchulus bolstered the lowest score at 3200Mhz.

Macha Phase 1 is surprisingly demanding for it being such a small arena to the point of it having the lowest results of the stock 2133 and 2933Mhz tests. However, this raid saw the biggest gains overall when increasing the RAM speed. Going from stock I saw an ~11% gain at 2933mhz, and a ~24% gain at 3200mhz when comparing the averages. The minimums however are strange as the first jump wasn't too big at only 11%, but the second one was huge at a 41% increase over stock. There's something about the first part of Macha that loves fast RAM.

Macha Phase 2 falls into a very similar situation that Claire did. The average framerate was again already near the top leaving little to no room for improvement. Although the minimums have room to grow because increasing the RAM speed pulled me out of the 40's and into the 50's. Only a 1fps gain after bumping the speed up once, but a 6fps or ~12% increase in minimum framerate over stock.

When I get around to Neam and Balor, I will only be doing stock 2133 testing. The reason behind this is similar to what I explained earlier in this post. Increasing the RAM speed causes the memory controller in the CPU to run hotter, thus making the fan spin faster. This is done automatically at the BIOS level as tweaking the RAM above 2750Mhz causes the fan profile to be automatically "optimized." After extended testing at 2933 & 3200Mhz speeds, my fan profile somehow got stuck. After changing my RAM speed back to stock, the fan would still be running absurdly fast as if the RAM speed was still on an OC'd frequency. I could not override and change it back so my only option was to do a CMOS reset which did fix the issue. Due to the possibility of this happening again, and the fact that I don't wanna potentially reset my BIOS again, I'm electing to only post 2133Mhz results. I will however utilize the above benchmarks in order to create educated guesses on how RAM speed will affect hosting Redeemer Raids.


UPDATE 12/21/2018

Season 3 Raid Performance Results at 2133Mhz RAM Speed

FtVR7N4.jpg


Will rework Astera tests to include all RAM frequencies, average fps, minimum framerates, and percentage gains based on RAM speed increases. Will not be doing video but rather lining up all three screenshots of afterburner's tracked results.

Will do a average/min comparison for all Season 3 raids.

Will update with 8-man raid hosting in Balor/Neam in the future. Will do differing RAM speeds (2133, 2933, and 3200).

Will add Nifl raids

Will condense individual Astera Raid result charts into links and replace with links.

Will add Nifl, Redeemer, and Astera raid results in the form of a chart similar to Season 3 Raid chart.

Will do a repeat of "all" tests when not hosting to show full performance.
kls9BigAssTTs

Kommentare

  • Shadowcity2Shadowcity2
    Vindictus Vertreter: 760
    Posts: 21
    Mitglied
    Thanks for the vid. What's your exact RAM kit and are you able to get it to 3200 MHz on your platform?
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    edited Juni 5, 2018
    Thanks for the vid. What's your exact RAM kit and are you able to get it to 3200 MHz on your platform?

    This is my exact RAM kit: https://www.newegg.com/Product/Product.aspx?Item=N82E16820231941&cm_re=g.skill_ripjaws_v_series_16gb-_-20-231-941-_-Product

    It may be advertised as "Intel Z170 Platform / Intel X99 Platform" memory, but it works with Ryzen.

    This was the motherboard I used: https://www.newegg.com/Product/Product.aspx?Item=N82E16813144177&cm_re=msi_x470_gaming_plys-_-13-144-177-_-Product
  • Shadowcity2Shadowcity2
    Vindictus Vertreter: 760
    Posts: 21
    Mitglied
    Are you able to run your RAM at full speed at 3200 MHz CL16 or are you limited to 2933 MHz? If you can run it at 3200 MHz CL16 is it stable? By stable I mean rock solid stable.
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    Are you able to run your RAM at full speed at 3200 MHz CL16 or are you limited to 2933 MHz? If you can run it at 3200 MHz CL16 is it stable? By stable I mean rock solid stable.

    I can run at 3200mhz stable, however I'm choosing to leave it at 2133mhz for the reason I listed at the bottom of the original post. I'll keep it at stock until the computer "needs" the extra boost in power which probably won't be near the end of its life. I tend to hold off on overclocking anything till the end so I can squeeze another year or two out of it.
  • Shadowcity2Shadowcity2
    Vindictus Vertreter: 760
    Posts: 21
    Mitglied
    Thank you for the confirmation.
  • adyrocksadyrocks
    Vindictus Vertreter: 500
    Posts: 17
    Mitglied
    i obtained almost same results with core 2 duo e8400 and a gtx9600 (10 years old PC)
    game is not even near optimized for multiple cores and dx11(12) video cards
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    edited Juni 7, 2018
    adyrocks wrote: »
    i obtained almost same results with core 2 duo e8400 and a gtx9600 (10 years old PC)
    game is not even near optimized for multiple cores and dx11(12) video cards

    Yeah, I'm going to go ahead and call BS on that claim of similar results with a core 2 duo e8400.

    Note the performance of the core 2 duo e7500.
    An e8400 is only ~5% better than the e7500, so yeah.... nice troll post.

  • adyrocksadyrocks
    Vindictus Vertreter: 500
    Posts: 17
    Mitglied
    awww , to bad i dont have all components to make that old PC again ....
    btw with i5 8600k and 3000MHz memory i have steady 60 fps :P
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    adyrocks wrote: »
    awww , to bad i dont have all components to make that old PC again ....
    btw with i5 8600k and 3000MHz memory i have steady 60 fps :P

    Noice. I was debating on the 8700K because it offers better performance in gaming, but seeing as I only have a 60Hz monitor and both processors can cap my refresh rate, Ryzen was the choice due to the multi-threaded performance gains. Comes in handy when rendering video.
  • EninyaEninya
    Vindictus Vertreter: 870
    Posts: 25
    Mitglied
    Cloakshire wrote: »
    adyrocks wrote: »
    awww , to bad i dont have all components to make that old PC again ....
    btw with i5 8600k and 3000MHz memory i have steady 60 fps :P

    Noice. I was debating on the 8700K because it offers better performance in gaming, but seeing as I only have a 60Hz monitor and both processors can cap my refresh rate, Ryzen was the choice due to the multi-threaded performance gains. Comes in handy when rendering video.

    A bit of necro, but I plan to get an 8700k, despite running 1080p/60 FPS on my current monitor. The reason is that I prefer modifying game graphics using Nvidia Inspector, SweetFX, or configs (which affect this game) to amplify their graphics. With VSnobs gone, I never did repost the graphics configs I made elsewhere.
  • RizzleRizzle
    Vindictus Vertreter: 2,260
    Posts: 199
    Mitglied
    Eninya wrote: »
    Cloakshire wrote: »
    adyrocks wrote: »
    awww , to bad i dont have all components to make that old PC again ....
    btw with i5 8600k and 3000MHz memory i have steady 60 fps :P

    Noice. I was debating on the 8700K because it offers better performance in gaming, but seeing as I only have a 60Hz monitor and both processors can cap my refresh rate, Ryzen was the choice due to the multi-threaded performance gains. Comes in handy when rendering video.

    A bit of necro, but I plan to get an 8700k, despite running 1080p/60 FPS on my current monitor. The reason is that I prefer modifying game graphics using Nvidia Inspector, SweetFX, or configs (which affect this game) to amplify their graphics. With VSnobs gone, I never did repost the graphics configs I made elsewhere.

    Please - repost them!!
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    Updated Original Post with Astera Raids
    kls9
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    edited Dezember 20, 2018
    Updated 12/20/2018
    Changes
    • Removed old 2133Mhz Astera video and test results.
    • Added a complete comparison of Astera raids at all three frequencies.
    • Added thoughts on each raid's performace.
    • Added mini announcement at the bottom regarding plans for host testing of Redeemer Raids.
  • CloakshireCloakshire
    Vindictus Vertreter: 4,580
    Posts: 646
    Mitglied
    edited Januar 9, 2019
    IMPORTANT UPDATE: It seems that the way Ryzen processes Vindictus's particle effects is the root cause for many frame drops. Going off of my Season 3 Hosting Chart found above, I noted that I got this weird 26 fps drop in Lugh whenever he performed the teleporting spin attack. My old rig (FX8150/Radeon 6950) never experienced this problem, but it ran with lower graphic settings. Through some trial and error, I managed isolate the fps drops as being tied to particle effects setting. I tested low particle effects when hosting and the ~26 fps drop didn't happen, and instead only went as low as 57fps during the spinning attack.

    This video shows 3 short clips of Lugh doing the attack in question. The only difference between each of the clips is that the particle effects are being changed from high, to medium, to low.
    High - framerate drops to 42
    Medium - framerate drops to 55
    Low - framerate drops to 59 (not impacted).




    I even cranked up other graphics settings to further emphasize the impact of particle effects.
    Shader Detail: High --> Very High
    Antialiasing Mode: None --> 8x
    Texture Filtering Mode: Trilinear --> Anisotrophic 16x

    joppHKI.png


    The oddity here is that Lugh's framedrop does NOT occur when he is encountered within Onslaught even with particles on High. There are also other bosses that cause similar framedrops such as Neam's short dash attack that leaves afterimages. I tested low particle effects on her as well and the fps drop issue disappeared. I'll be reworking the original post with particle effects set to low as that seems to be Ryzen's Achilles heel for vindictus.


    *Tested Elchulus and Macha as those were low performing raids. Current 2133mhz results with above graphics settings are matching and/or beating old 3200mhz tests using the original graphics setting. With this in mind, I have concluded that all of my framerate lows are due to particles not being optimized for Ryzen.
  • VernandiVernandi
    Vindictus Vertreter: 1,165
    Posts: 27
    Mitglied
    That FPS drops with high particle settings not only ryzen problem. Before upgrade to ryzen i had i5 4690k (with 4.8ghz overclock) and had same fps drops on lugh\neam\similar effects bosses.

    One tip for Arisha players (in the main): DOF effect in vindictus very unoptimized too. So if you wanna maximum smooth framerate while playing arisha, better use just high shader detail, not very high. Because every warp hole on very high shader detail you will have dof effect and drop, depend scene, fps 5-8 .
    That thing also same on Ryzen config (r7 1700 4ghz; 3,2ghz CL 16 ram) and intel config (i5 4690k 4,8ghz; 2.1ghz cl11 ram)