Benchmark Result Discussion

[geo_gan]

I wonder what kind of boost we're gonna get from a 5090. I bet 25-35% or so compared to 4090.

Probably at most... since a lot of the new AI based upscaling/multiplying features are unusable in VAM and VR unfortunately. Only speed ups would come from an increase in shader cores and the increase in power to run them faster.

 

[GenericUsername1]

Interesting. Judging from Seraphim's 7800x3d results (back on page 30) it seems like you're gpu bottlenecked with the 3080 Ti?.
I have an RTX 4070 and an old intel 9700k, and from these results it seems like there's no point in my getting a CPU above a ryzen 9700X or so with my current gpu? (as far as vam 1 is concerned). Also first post hello everyone!

For me with VaM, yes, I feel the 3080ti is the bottleneck with this CPU. VaM profits a lot from X3D cpus as its FPS is highly dependant on single core speed, and the huge CPU cache helps keeping the core busy by reducing the need to reach out to relatively slow RAM. But please also note: a few pages back 40 Seraphim upgraded to an 9800X3D and got again an impressive FPS increase (thank you Seraphim btw for providing these consistent benchmark results ). So the 7800X3D appears to bottleneck the 4090. The 4090 is 89% faster than my 3080ti according to userbenchmark.com, so that doesn't suprise me.

With regards to the 9700X, I'm very sure it'll give you quite an increase in FPS. I've went with my 3080ti from a 9700K -> 7700X -> 7800X3D and got a nice Baseline 3 (the important one) FPS boost with every upgrade: the Min 1% FPS went from 58.49 -> 78.44 -> 97.45. But when comparing the 9700X vs. the 7800X3D in game benchmarks, the latter seems to perform better. Personally I'd pick the 7800X3D over the 9700X based on game benchmarks I find online; the newer architecture and higher clock speed doesn't compensate for the lack of X3D cache. Of course in general purpose workloads instead of games, the 9700X performs better.

 

[mostvanvege]

Probably at most... since a lot of the new AI based upscaling/multiplying features are unusable in VAM and VR unfortunately. Only speed ups would come from an increase in shader cores and the increase in power to run them faster.

I think so too. I have just collected a 4090 and I would say if anyone wanted but delayed an upgrade now is the time to look for bargains. New 50xx gen wont give significantly performance upgrade (aside the 5090 obviously).
I think like
5070 -> 4070 Ti (nonS)
5070 Ti < 4080 or even
5080 < 4090 close, but I do not expect it can match in vam

However individual reviews are due after jan 21, after that and when street prices are revealed I expect to rise the ada 2nd hand market again.

 

[lolmao500]

Probably at most... since a lot of the new AI based upscaling/multiplying features are unusable in VAM and VR unfortunately. Only speed ups would come from an increase in shader cores and the increase in power to run them faster.

Yeah I was talking about raster with zero AI based stuff. GDDR7 will really help compared to the 4090. The 4090 was memory starved... even with gddr7 and a 512 bits bus it might not be enough... maybe nvidia will release a 5090ti or something when gddr7 matures and goes up by 50% in speed. We'll see.

But yeah I dont think fake frames will work on vam 2 but upscaling probably. Anyway im not buying a 5090 so it wont concern me haha

 

[Substrata_]

@1080p 60hz:

Now eyeing the 9800x3d as an upgrade...

 

[perkel]

how to get older version of VAM ?
CPU patch says it doesn't work with latest and i have 1.20.22, updating it would go to latest instead of specific one.

Also how does that benchmark works. I load scene and it is just a painting and nothing else no stats buttons etc.

edit: ok found how to get older version.
edit: figured out benchmark too.

 

[johndoe1221]

How am i stacking up here? fresh install, EXPO on, I was wondering why my performance was terrible before, then i realized steam was pushing my resolution to some ridiculous levels.....

 

[Dominanto]

I have strange issue when I select any of benchmark from Benchmark scene (Official / custom run) there is nothing happened. I downloaded all dependences tried PC/VR game launch with no luck. Maybe anyone can suggest what it can be? Will try clean client.

 

[Emosion]

My Testing with CPU patch shows very small gain going from 7800X3D to 9800X3D for 1080p and slower Baseline 3 on 2k

7800X3D RTX4090 32G DDR5-6000 w CPU patch

9800X3D RTX 4090 32G DDR5-6000 w CPU patch

 

[Emosion]

For 2K results
7800X3D RTX4090 32G DDR5-6000 w CPU patch

9800X3D RTX 4090 32G DDR5-6000 w CPU patch

 

[Emosion]

For VR Test using Virtual Desktop Ultra Mode, 9800X3D has very small gain over 7900X3D

7800X3D RTX4090 32G DDR5-6000 w CPU patch

9800X3D RTX 4090 32G DDR5-6000 w CPU patch

 

[Emosion]

Upgrading from DDR5-6000 to 6400 made slight gain for both 1080p and 2k

9800X3D DDR5-6400 1080p w CPU patch

9800X3D DDR5-6400 2K w CPU patch

 

[johndoe1221]

Doubled my Ram, and got considerably faster stuff too... and I lost frames?

 

[Seraphim]

Doubled my Ram, and got considerably faster stuff too... and I lost frames?
View attachment 451852
View attachment 451853

You got more ram, but its configured wrong if its faster ram. your cpu is now slower. Plus you are mostly gpu bottlenecked, so cpu and ram will make little difference in vr.

 

[trety]

Doubled my Ram, and got considerably faster stuff too... and I lost frames?
View attachment 451852
View attachment 451853

If by 'doubling' you meant adding another 2 sticks, so running 4 sticks now, then yeah, yours RAM is slower after that.

 

[Emosion]

Doubled my Ram, and got considerably faster stuff too... and I lost frames?
View attachment 451852

@trety is correct. For the AM4 motherboard, it increases the performance going 4 x RAM stick. But it decreases the performance using 4 x RAM on AM5 Motherboard comparing to 2 x RAM stick.

 

[soverm]

Test platform
CPU: 13900KS
GPU: MSI RTX 4090 SUPRIM LIQUID X
RAM: KINGSTON FURY DDR5 8000 C38 16GB*2
MOTHERBOARD: ROG MAXIMUS Z790 APEX
SSD: Micron Crucial T500 2TB

1. Compare the performance differences of DDR5 8200, 7600, 6800, and 6000 at 1080P, 2K, and 4K:

DDR5 timings are all fixed at C38-48-48-128, making it easy to compare the impact of simple frequency differences.

Use CPU performance patch, turn off E core and HT, all 8 P cores are fixed at 5.9G to reduce CPU bottleneck

Because there is a lot of data, in order to facilitate you to compare the differences more intuitively, I tested the AVG and MIN1% of baseline3 twice and took the average, using 6000 frequency as 100% as the benchmark, and the percentage of each frequency increase in baseline3 ( The scenario pressure of baseline3 is relatively high and closer to my daily usage, so I use the data of baseline3 for conversion)

1080P:
AVG / MIN1%
6000 100% / 100%
6800 101% / 104%
7600 102% / 106%
8200 102% / 108%

2K:
AVG / MIN1%
6000 100% / 100%
6800 100% / 103%
7600 98% / 106%
8200 98% / 106%

4K:
AVG / MIN1%
6000 100% / 100%
6800 100% / 98%
7600 100% / 99%
8200 99% / 99%

Conclusion: Increasing the DDR5 frequency at 1080P and 2K resolution can significantly increase MIN1%, but the AVG increase is not so big that it can even be considered as an error margin. At 4K, neither MIN1% nor AVG increases


Because English is not my native language, I use Google Translate to translate the content. I hope you can understand what I mean.

 

[soverm]

Test platform
CPU: 13900KS
GPU: MSI RTX 4090 SUPRIM LIQUID X
RAM: KINGSTON FURY DDR5 8000 C38 16GB*2
MOTHERBOARD: ROG MAXIMUS Z790 APEX
SSD: Micron Crucial T500 2TB


2. Compare the differences caused by different single-core performance at three CPU frequencies: 5.0G, 5.5G, and 5.9G:

RAM uses overclocked DDR5 8200 C38 to reduce RAM bottleneck

CPU uses performance patches, turns off E cores and HT, and fixes the 8 P cores to the following three frequencies: 5.0G, 5.5G, and 5.9G

5.0G: PerformanceTest single-core score: 4120, simulating the single-core performance of i5 13600K

5.5G: PerformanceTest single-core score: 4503, simulating the single-core performance of i7 14700K

5.9G: PerformanceTest single-core score: 4900, representing extreme single-core performance

Because there is a lot of data, in order to facilitate you to compare the differences more intuitively, I tested the AVG and MIN1% of baseline3 twice and took the average, using 5.0G as 100% as the benchmark, and the percentage of each frequency increase in baseline3 ( The scenario pressure of baseline3 is relatively high and closer to my daily usage, so I use the data of baseline3 for conversion)

1080P:
AVG / MIN1%
5.0G 100% / 100%
5.5G 111% / 106%
5.9G 114% / 112%

2K:
AVG / MIN1%
5.0G 100% / 100%
5.5G 101% / 107%
5.9G 101% / 110%

4K:
AVG / MIN1%
5.0G 100% / 100%
5.5G 99% / 99%
5.9G 98% / 96%

Conclusion: Improving single-core performance at 1080P can significantly improve AVG and MIN1%.
The improvement of AVG under 2K is not big and can even be regarded as the margin of error, but the MIN1% value has a significant improvement.
Neither MIN1% nor AVG improved at 4K

 

[soverm]

Test platform
CPU: 13900KS
GPU: MSI RTX 4090 SUPRIM LIQUID X
RAM: KINGSTON FURY DDR5 8000 C38 16GB*2
MOTHERBOARD: ROG MAXIMUS Z790 APEX
SSD: Micron Crucial T500 2TB


3. By limiting the power consumption of RTX4090, the performance of four graphics cards, RTX4060, 4070, 4080, and 4090, is simulated to compare the differences between different graphics cards:
Limited to 150W, time spy score 11604 emulating RTX4060
Limited to 185W, time spy score 17799 simulated RTX4070
Limited to 260W, time spy score 28113 emulating RTX4080
Original 475W, time spy score 38000 for RTX4090

CPU performance patch, turn off E core and HT, all 8 P cores are fixed at 5.9G to reduce CPU bottleneck

RAM uses overclocked DDR5 8200 C38 to reduce RAM bottleneck

Because there is a lot of data, in order to facilitate you to compare the differences more intuitively, I tested the AVG and MIN1% of baseline3 twice and took the average. I used RTX4060 as 100% as the benchmark, and the percentage of each frequency increase in baseline3 (baseline3 The scenario pressure is relatively high and closer to my daily use, so I use the data of baseline3 for conversion)




1080P:
AVG / MIN1%
RTX4060 100% / 100%
RTX4070 151% / 144%
RTX4080 251% / 209%
RTX4090 284% / 213%

2K:
AVG / MIN1%
RTX4060 100% / 100%
RTX4070 153% / 140%
RTX4080 244% / 214%
RTX4090 344% / 267%

4K:
AVG / MIN1%
RTX4060 100% / 100%
RTX4070 149% / 155%
RTX4080 247% / 247%
RTX4090 342% / 329%


Conclusion: Except for RTX4090, which cannot fully release its performance at 1080P, the performance improvements brought by other graphics cards are similar to their theoretical performance.

 

[HolySchmidt]

Dude this is TOO much man!
Who's gonna read all this?