The ACC GPU Benchmark Thread

Nice, this thread has some "gotta collect them all" appeal for me

It's getting there for sure.

 

New system:

2018-12-01 21:53:41 - AC2-Win64-Shipping
Frames: 6200 - Time: 60000ms - Avg: 103.333 - Min: 75 - Max: 144

 

2018-12-02 15:26:42 - AC2-Win64-Shipping
Frames: 4530 - Time: 60000ms - Avg: 75.500 - Min: 59 - Max: 106


I7 4790
16Gb Corsair DDR3
Asus GTX 1070
NV 4.17.01 drivers.

 

2018-12-02 10:31:18 - AC2-Win64-Shipping
Frames: 1506 - Time: 60000ms - Avg: 25.100 - Min: 17 - Max: 37

i7 4790k
16gb DDR3
Asus GTX 950

 

2018-12-02 10:44:59 - AC2-Win64-Shipping
Frames: 4790 - Time: 60000ms - Avg: 79.833 - Min: 58 - Max: 119

edit:nm

 

Looking at the results, how's possible that @D.Jankovic has AVG FPS below MIN FPS @LeDude83 ?
I'm considering to get a 290x so i was wondering.

EDIT :

Understood

 

I managed to squeeze a few more FPS from the Vega 64. I did it by "undervolting" it by 68 mV. This brought up the clock speed from ~1540 MHz to 1560 MHz and the Power consumption down from ~270 W to 250 W along with a temperature reduction of ~2 degrees. An efficiency tweak that improves performance. I'm Impressed.

2018-12-02 14:13:42 - AC2-Win64-Shipping
Frames: 4255 - Time: 60000ms - Avg: 70.917 - Min: 55 - Max: 98

 

post 1 updated.

@Jynnantonix I think you can achieve a better result. You should try to undervolt your GPU like I did and maybe adjust the fan curve to make sure the clock rate isn't reduced.

 

Now with the 290x (exactly the same config i7 4790k - 16gb DDR3) :

2018-12-02 16:33:40 - AC2-Win64-Shipping
Frames: 2018 - Time: 60000ms - Avg: 33.633 - Min: 24 - Max: 48

 

Lol at the results
nvidia 770=amd 470
amd 480 > 10fps less than 1060 3GB
amd 580 > 10 fps less than 1060 6GB
vega 64 > 10 fps less than 1070 and 20 less than 1080

AMD is definitely not the right choice for ACC (at this moment).

PS 2080ti is a beast!!!

 

Looking at the first system requirements on Steam, that seems to have been like...pick the 3rd strongest GPU from each manufacturer from the current series:



We know now: it has to be a Vega 64 and that one will even be slower than the 1070.

 

@LeDude83 What's your HBM clock set to, and whats your max temperature? Setting my 56 to 1100 mV core I'm getting 1573 mhz and around 195 W (In game, the menu somehow pulls 230W) and a peak temp of 70 C, and have 1000 mhz HBM clock. With these settings (plus closing other programs and actually running fullscreen instead of windowed) I'm getting these results now:

2018-12-02 14:55:24 - AC2-Win64-Shipping
Frames: 3933 - Time: 60000ms - Avg: 65.550 - Min: 49 - Max: 95

 

Wow, that's a really good result.

In Afterburner I set the core voltage to -100. Only by doing so I could achieve 1570 MHz, stable. The power usage has gone down from 260 W to ~230 W and the GPU temp is at 70. The temp is kind of worthless without the fan speed: 1900 RPM. That's fairly quiet on my GPU. At ~2100 RPM there's some sort of noise cliff after which it gets pretty annoying. The HBM clock is at the stock 945 MHz. I just increased it to 1000 MHz for a test run but the differences were within measurement uncertainty.

What model is your Vega?

And I have not much running other than ACC...the browser, FRAPS and Afterburner, Clamwin and RivaTuner (it always starts with AB).

The ACC menus currently take all the power your GPU can give. I had the power limit at maximum for a try and in the menus the GPU was at 300 W. Together with -100mV core this caused a system crash. I turned the power limit back down to +/- 0 - it makes no difference for FPS because I'm way below the standard 260 W.

 

Mine's a reference card from powercolor, but I flashed the 64 bios for the extra memory voltage and slapped a morpheus II onto it with a couple 120mm fans set to 2000 rpm, it's barely louder than my case fans under load now.

And yeah, i tried setting the memory to 900 mhz to check the scaling, we're entirely core limited. I had first tested with my browser, hwinfo64 and wattman open to keep an eye on things, but they do seem to sap performance more than i expected.

 

Ok - I'll give it another go just with Afterburner running in Background and FRAPS.

You surely have a well-running 56 there - 93% the AVG FPS of my 64.

MY GPU is also modded - with an extra heatsink +Fan on the backplate. Thermal gap filler was applied underneath the heatsinks between backplate and PCB. On the other side of the board, there's the core voltage regulators. They reach up to 120 degrees in stock condition and with this mod they peak at ~87 degrees. This also reduces the thermal energy from the regulators the main cooler needs to handle.

 

@LeDude83: in your excel list you added the Ryzen 5 1600, but i used the Intel Xeon E3-1231 v3 with the 2070 card. I know a little bit confusing due switching a lot between 2 PC's^^

Thanks

 

Very small but reproducable improvement by fiddling with the AMD driver settings. I think the main difference was to set texture filtering from "standard" to "performance". Setting it to "quality" decreases FPS further so this property (despite AA settings which seem to make nothing at all) appears to work and make a FPS difference.

I haven't noted a visual difference from changing the texture filtering property but I wouldn't ecxactly know where to look for it.

Anyways...a few FPS more is always welcome:

2018-12-07 10:12:30 - AC2-Win64-Shipping
Frames: 4349 - Time: 60000ms - Avg: 72.483 - Min: 55 - Max: 102

 

I didn't post first bench, but was same for me with Nvidia, changing in nvidia panel, acc profile to performanse mode made game more stable and avg fps was higher for me to. On stock setting in panel i had unusal frame drops, with 74fps cap, even alone my fps would drop on 70 for sec or two very often.

 

I actually already did that before running the bench, clocks are around 1530. Maybe I can try undervolt more and use a bit more tdp to get the clocks up. I'm at 1000mv. You?

 

Sorry guys...
But in the end, what's the conclusion?
Is better a new intel i5 (8th gen) or new a amd ryzen 5 (2nd gen)?

A new i7 or a new ryzen 7?