Benchmark Result Discussion

Fresh VaM installation

5950X + RTX 4080
1.png


5800X3D + RTX 4080
2.png
 
Last edited:
Moved to a 5600, a fair bit of improvement on baseline 3 for just a 100 upgrade, but frame averages slightly dropped everywhere else, tested before and after I made the swap back in february the results are still the same, I think this is its limit. Mem is only 2400, but I have not noticed this drop on more traditional benchmarks.

Benchmark-20230211-040030.png


Both have 3060ti at mem 7700 core 1980
Benchmark-20230511-043604.png
 
I feel like the baseline 3 score should be a few ticks higher, ie physics time. Any suggestions?
 

Attachments

  • Benchmark-20230513-174416.png
    Benchmark-20230513-174416.png
    856 KB · Views: 0
It could potentially be RAM, I assume you're running at the stock 4800 spd. I am facing a similar issue with my Asus G18 as I am trying to find a way to upgrade reliably to 5200 or 5600 ram spd. AMD's processors can see significant gains with faster Ram.

Have you considered an undervolt? I've had reasonable gains undervolting by -125mv on my system, and an improvement in the Baseline 3 Physics Time.


Vanilla Install on Asus ROG 4080
VAM 1080 - 3-23-23 - Vanilla.jpg


-125 Undervolt
VAM - 125 UV.png
 
It could potentially be RAM, I assume you're running at the stock 4800 spd. I am facing a similar issue with my Asus G18 as I am trying to find a way to upgrade reliably to 5200 or 5600 ram spd. AMD's processors can see significant gains with faster Ram.

Have you considered an undervolt? I've had reasonable gains undervolting by -125mv on my system, and an improvement in the Baseline 3 Physics Time.


Vanilla Install on Asus ROG 4080
View attachment 245099

-125 Undervolt
View attachment 245100

Nice gains! What was your method for undervolting? I'm a novice at this kinda thing unfortunately.
 
Nice gains! What was your method for undervolting? I'm a novice at this kinda thing unfortunately.

Go to the timestamp of 22min and 22 seconds for a breakdown on AMD undervolting.

Download UXTU which is an AMD APU tuning utility:

Go to the curve optimizer setting and do a -10 undervolt and run benchmarks such as 3Dmark and Cinebench R23 to test for stability.
-30 is the highest you can go but stability at these settings will depend on silicon quality/lottery.

Revert you undervolt to factory settings whenever updating your bios, it's been known to cause issues for some.

Though operating outside of OEM parameters is a risk undervolting is "generally" considered safe as you are not running extra voltage.

Happy Hunting
 
Go to the timestamp of 22min and 22 seconds for a breakdown on AMD undervolting.

Download UXTU which is an AMD APU tuning utility:

Go to the curve optimizer setting and do a -10 undervolt and run benchmarks such as 3Dmark and Cinebench R23 to test for stability.
-30 is the highest you can go but stability at these settings will depend on silicon quality/lottery.

Revert you undervolt to factory settings whenever updating your bios, it's been known to cause issues for some.

Though operating outside of OEM parameters is a risk undervolting is "generally" considered safe as you are not running extra voltage.

Happy Hunting

Is there a particular directory it needs to be installed in? Trying to tune things but it's not having any effect.
 
Is there a particular directory it needs to be installed in? Trying to tune things but it's not having any effect.
I would install it on the C: drive but as long as it recognizes your CPU and allows you to place an undervolt the directory "shouldn't" be an issue. Download Cinebench R23 and go to advanced settings and turn off minimum test duration. Benchmark to get a default value, then set a -10 undervolt and apply it, run Cinebench again and your scores should improve. Then undervolt by -15 and see if your scores improve again. Gizmoslip tech managed 36k with his undervolt.
Try running the program with admin right (right click the executable) though this really shouldn't be the issue. Send me a private message and we'll work on it until we figure it out.
 
Guys, how much performance improvement would I be able to get if I overclock my DDR4 2400 to 2600? And a slight undervolt? I can't really OC the clocks because CPU temps are like 85ºC

Ryzen 5600 + RX6700XT
 
Guys, how much performance improvement would I be able to get if I overclock my DDR4 2400 to 2600? And a slight undervolt? I can't really OC the clocks because CPU temps are like 85ºC

Ryzen 5600 + RX6700XT
  • About 10°C lower temps from undervolting seem to be common for Ryzen 5th generation, that also means a few more % performance as you won't run int the temperature/power limit that quickly. However, keep in mind that, just like overclocking, undervolting voids your AMD warranty!
  • Make sure to test your system properly....and then test again. Not every calculation error of your CPU leads to an obvious crash, so it can be hard to notice with gaming benchmarks. You do NOT want to work with a CPU that thinks 1+1 is 3 in 1/10000 of cases. Some software like Prime95 that actually verifies the calculation results are actually correct may help.
  • I wouldn't recommend manual overclocking RAM. I would stick to modules with preset overclocking profiles. These are more expensive, but tested with the profile settings. All modules are tested in the factory. If some "slow" module would work overclocked, they would have sold it with overclock profile and made more money...they didn't, meaning the module proably won't work reliably.
  • DDR4-2400 is slow, though. I think that's the failsafe-default speed when you don't set the overclocking profile stored in the RAM module? 5th gen AMD officially supports up to DDR4-3200 RAM and is commonly know to support even more (although can't rely on it). Going from 2400 to 3200 I would expect some 5-10% speed boost in games with CPU bottleneck (like VaM).
  • Check the table here: https://en.m.wikipedia.org/wiki/CAS_latency#Memory_timing_examples
    • Find your RAM module, the first CL number corresponds to the CAS latency column. When you buy a DDR4-3200 module, its usually CL16, but you can spend extra bucks for example on CL14 modules. Time in the "First word" column should give you and idea of how much faster your memory will respond. Doesn't directly transform into more FPS, of course...but CPUs spend a LOT of time waiting on RAM, that's why the new 3D cache CPUs are so effective for gaming.
 
  • About 10°C lower temps from undervolting seem to be common for Ryzen 5th generation, that also means a few more % performance as you won't run int the temperature/power limit that quickly. However, keep in mind that, just like overclocking, undervolting voids your AMD warranty!
  • Make sure to test your system properly....and then test again. Not every calculation error of your CPU leads to an obvious crash, so it can be hard to notice with gaming benchmarks. You do NOT want to work with a CPU that thinks 1+1 is 3 in 1/10000 of cases. Some software like Prime95 that actually verifies the calculation results are actually correct may help.
  • I wouldn't recommend manual overclocking RAM. I would stick to modules with preset overclocking profiles. These are more expensive, but tested with the profile settings. All modules are tested in the factory. If some "slow" module would work overclocked, they would have sold it with overclock profile and made more money...they didn't, meaning the module proably won't work reliably.
  • DDR4-2400 is slow, though. I think that's the failsafe-default speed when you don't set the overclocking profile stored in the RAM module? 5th gen AMD officially supports up to DDR4-3200 RAM and is commonly know to support even more (although can't rely on it). Going from 2400 to 3200 I would expect some 5-10% speed boost in games with CPU bottleneck (like VaM).
  • Check the table here: https://en.m.wikipedia.org/wiki/CAS_latency#Memory_timing_examples
    • Find your RAM module, the first CL number corresponds to the CAS latency column. When you buy a DDR4-3200 module, its usually CL16, but you can spend extra bucks for example on CL14 modules. Time in the "First word" column should give you and idea of how much faster your memory will respond. Doesn't directly transform into more FPS, of course...but CPUs spend a LOT of time waiting on RAM, that's why the new 3D cache CPUs are so effective for gaming.
Thanks! I need to consider the price to upgrade to 3200Mhz, maybe it's better to save the money for a future upgrade to AM5 or whatever intel is at right now. My 2400Mhz modules are from 2019, that's their rated speed. (HyperX Fury, 8GB, 2400MHz, DDR4, CL15 - HX424C15FB3/8)
 
Thanks! I need to consider the price to upgrade to 3200Mhz, maybe it's better to save the money for a future upgrade to AM5 or whatever intel is at right now. My 2400Mhz modules are from 2019, that's their rated speed. (HyperX Fury, 8GB, 2400MHz, DDR4, CL15 - HX424C15FB3/8)
I just upgraded from 16GB of 2400 to 32 GB 3200. Frame rate did not change from my post #803 higher up. Physics times dropped slightly to 6.9 average over 3 runs.

Upgrading your GPU would be a better improvement over an AM5 , radeon cards don't seem to perform quite as well to RTX counterparts on VAM. Looking at game benchmarks the rx6700xt beats the 3060ti in many games.

My 5600 is auto oc to 4.6. 65ºC on just a wraith prism.
85ºC is nearing max temp, and probably is throttling down, any medium size cooler would help over the included stealth.
 
I just upgraded from 16GB of 2400 to 32 GB 3200. Frame rate did not change from my post #803 higher up. Physics times dropped slightly to 6.9 average over 3 runs.

Upgrading your GPU would be a better improvement over an AM5 , radeon cards don't seem to perform quite as well to RTX counterparts on VAM. Looking at game benchmarks the rx6700xt beats the 3060ti in many games.

My 5600 is auto oc to 4.6. 65ºC on just a wraith prism.
85ºC is nearing max temp, and probably is throttling down, any medium size cooler would help over the included stealth.
Good info! I was afraid the 3060Ti 8GB VRAM would be an issue. But looks like the AMD driver and encoder are a bigger issue than that. at least for VR (all games)

But I'm sure I'm CPU limited. Only physics heavy scenes are a problem. Sometimes I can even increase the resolution to 1.2 without any FPS drop.
 
Good info! I was afraid the 3060Ti 8GB VRAM would be an issue. But looks like the AMD driver and encoder are a bigger issue than that. at least for VR (all games)

But I'm sure I'm CPU limited. Only physics heavy scenes are a problem. Sometimes I can even increase the resolution to 1.2 without any FPS drop.

The 8 GB vram does fill up with 3 or more characters or from changing clothes . I tested a few scenes from the hub, physics had minimal impact compared to MSAA, pixel lights, and 3+ characters. CPU thermal throttling is holding performance back, but an X3D chip would be even better.
 
Moved to a 5600, a fair bit of improvement on baseline 3 for just a 100 upgrade, but frame averages slightly dropped everywhere else, tested before and after I made the swap back in february the results are still the same, I think this is its limit. Mem is only 2400, but I have not noticed this drop on more traditional benchmarks.

View attachment 243518

Both have 3060ti at mem 7700 core 1980
View attachment 243520


How does your game run in VR in a real scene?
 
How does your game run in VR in a real scene?
I don't have VR. Just tried 1440/2160, tried a couple real scenes results +/- 20frames to start the scene from these numbers a lot of variables to account for. I see used 3070ti are now less than this 3060ti I got used.
Benchmark-20230602-044550.png

Benchmark-20230602-045957.png
 
I don't have VR. Just tried 1440/2160, tried a couple real scenes results +/- 20frames to start the scene from these numbers a lot of variables to account for. I see used 3070ti are now less than this 3060ti I got used.
View attachment 253376
View attachment 253377

yeah I recently saw a new 6900XT go for the same price I got my 3060 ti last year. I'm not content with this card but I'm going for a cpu upgrade first. Even in VR i feel like it's the physics that kill my fps moreso than the res.

I did massively improve my fps by switching to the var browser bepinex plugin and disabling vam cache (it has its own cache) so now it's getting playable
 
Replaced my 5800X3D with a 7800X3D, here's a comparison of the benchmark results.

5800X3D, RTX4090, DDR4-3200 CL14
Benchmark-20230606-063142.png

7800X3D, RTX4090, DDR5-6000 CL30
Benchmark-20230606-071259.png
 
“cwayne1989, 帖子: 104644, 成员: 67527” said:
所以现在我目前的系统是
AMD R5-3600 @ 4.0GHz
32GB DDR4
AMD Radeon 5700 XT (蓝宝石硝基 +)

我想升级到 R7-5800X 或 R7-5800X3D

我已经查看了 X3D 的基准测试结果,它似乎是 VAM 的一个地狱处理器,我只是好奇是否真的值得额外支付 100 美元与 X3D 一起使用标准 5800。特别是如果我决定保留我目前拥有的同一显卡?


最后,我知道这可能是一个非常愚蠢的问题,但是从标准R5-3600到这些处理器中的任何一个的跳跃在VR中真的会那么明显吗?

我只是担心我最终会放弃这笔钱,而 VAM 仍然会......井。。像 VAM 一样运行,哈哈
[/引用]
 
Back
Top Bottom