AMD Ryzen 7 3800XT Put Through AotS Benchmark

[btarunr]

AMD's upcoming Ryzen 7 3800XT 8-core/16-thread processor was put through "Ashes of the Singularity" (AotS) benchmark, as uncovered by HardwareLeaks (_rogame). Paired with an NVIDIA GeForce RTX 2080 graphics card, the processor is able to put out CPU frame-rates of 113.2 FPS (averaging all batches); 135.9 FPS in the normal batch, 115.31 FPS in the medium batch, and 95.49 FPS in the heavy batch, with preset level set to "Crazy_1080p." An older article points to the 3800XT ticking at 4.20 GHz base with 4.70 GHz maximum boost (compared to 3.90 GHz base and 4.50 GHz boost of the 3800X), which means AMD aims to shore up gaming performance of its 3rd gen Ryzen processors with the XT series.



View at TechPowerUp Main Site

 

[cucker tarlson]

I hope it's both higher clocks and lower vcore

 

[Tomorrow]

These batch numbers are meaningless without comparison.

 

[Space Lynx]

These batch numbers are meaningless without comparison.

would be nice if they included what the regular 3800 gets in those same tests at the very least

 

[ratirt]

would be nice if they included what the regular 3800 gets in those same tests at the very least

I agree. you see numbers and it is ok to see them but it's just meaningless numbers if you have nothing you can compare these with.

 

[ShurikN]

I agree. you see numbers and it is ok to see them but it's just meaningless numbers if you have nothing you can compare these with.

Found on Komachi's twitter. Although this one is on High and not Crazy

113.2 AVG 92.9
135.9 normal 102.2
115.31 medium 96.2
95.49 heavy 82.5

 

[Space Lynx]

Thanks @ShurikN those are very decent gains really.

 

[ShurikN]

Thanks @ShurikN those are very decent gains really.

It's really hard to find comparable results. Different GPUs, different settings...
The CPU test should disregard the difference in GPUs but I've never run it, so I can't tell. Although 1080Ti = 2080 non super, more or less.

 

[Vayra86]

These batch numbers are meaningless without comparison.

Well... even with comparison... the AotS bench is pretty useless overall. What are you measuring? Some ingame situation no single other game has ever done? Great! You know your AotS performance...

 

[Tomorrow]

Ok so did some digging and here's the best result on 3800X+RTX 2080 @ 1080p Crazy using DX12:

And here's the the best 3800XT result using the same GPU and same settings:

Please note that i was unable to find up to date version runs of this combo using DX12. So the benchmark version is 2.80.62792.0 instead of 2.92.73675.0 used on the XT run.
There is an up to date version but it's a DX11 run on 3800X and the performance is much worse due to DX11 bottleneck.

Still comparing these two we can see some nice gains on the CPU side. Difference is between 6-20fps depending on the batch. Bigger gap in normal batch. Less so on Heavy batch.
Pure combined CPU+GPU framerate gains are less impressive. Between 0-8fps.

 

[Space Lynx]

ya why would you even run that bench for a leak, seems weird honestly. meh. we will find out in 8 days

 

[ShurikN]

Looking at the numbers it seems like a 10% increase.

 

[Assimilator]

AotS has been known to be trash as a benchmark from day 1... why are people still using it as such in 2020?

 

[droopyRO]

If it is a 10% increase for the same money and TDP, why not

 

[Vayra86]

AotS has been known to be trash as a benchmark from day 1... why are people still using it as such in 2020?

Same reason FFXV bench was used over and over. Its easy to manipulate with those numbers. In the case of AotS the main benefit is that people have no idea what a number means for any other game or application. It just applies to AotS and the game scales almost linearly with additional CPU grunt. 99% of other use cases especially in gaming do not

 

[Chrispy_]

ya why would you even run that bench for a leak, seems weird honestly. meh. we will find out in 8 days

I'm seeing a lot of stuff from _rogame at the moment. As a leaker he/she is new on the scene with no proven leak accuracy. The lack of any real history makes me extremely dubious, though I'd be happy to be proven wrong.

 

[Space Lynx]

If it is a 10% increase for the same money and TDP, why not

the problem is it won't be for the same money, as regular 3800x often dropped to $299 shipped or $305 shipped. They will make this at least 30-70 dollars more I am 100% certain, or back to the original price of $399, and at that point might as well wait for 5nm 4800x /shrug

 

[DemonicRyzen666]

Extrapolation of data

3800xt is from top to bottom in comparison

1.00 =%

1.0435
1.1065
1.0945
1.1726
1.0369
1.1012
1.0130
1.0657

Average in in this bench = 1.079 or 7.9% over all. Not quite 10%
The base is clock 7.6% higher the the single core is only 4.4% higher
I'll have to check my math. pretty sure this game of bench is showing scaling of almost 1.04 or 104% for the increase in clock speed. Unless infinity fabric is a bottleneck or current 3000 series or the l3 cache speed is another bottleneck. The last time scaling went like that was with phenom II with a NB overclock. That was rarely ever last past 1.02 scaling.