R9 390 Crossfire

[latentjones]

The Beginning: How I found Crossfire 390's

 

I have just upgraded my rig to a Crossfire setup running two Sapphire R9 390's. I just wanted to start a thread that covered benchmarks and my first hand experience in games so that people who might be interested in this setup could find some detailed information. I tried to research benchmarks and gaming experiences before my upgrade but found that there really was very little. Hopefully this will prove beneficial to some and perhaps just curious and interesting to others.

 

So first things first, here is my rig:

 

System

• CPUIntel i5 4590 @ 3.3
• MotherboardMSI Sli Krait ed
• RAM2x 8gb Crucial CT102464BA160B DDR3
• GPUR9 390 (Sapphire) Crossfire
• CaseFractal Design Define S
• Storage1tb HDD 120gb SSD
• PSUEVGA Supernova GS 850w Gold
• Display(s)BenQ XL2730z 1440p 144hz Freesync
• Cooling2x 140mm Fractal Case fans 1x Non branded 120mm Case Fan
• KeyboardKane Pro
• MouseKane Pro
• SoundSades SA-902 Headphones
• Operating SystemWindows 10

 

N.B. I will update the specific model of Fractal fans I have but for now note that they are the ones that come with the Define S case as standard. I'm thinking of upgrading them to Fractal Venturis but I digress.

 

The first benchmark I ran was Unigine Valley. These were my settings:

 

Preset: Custom

API: DirectX 11

Quality: Ultra

Stereo 3D: Disabled

Monitors: Single

Anti-Aliasing: x2

Fullscreen: On

Resolution: 2560x1440

 

On the single Sapphire R9 390 at stock speeds the results were:

 

FPS: 51.0

Score: 2132

Min: 24.6

Max: 91.6

 

Reasonably respectable I thought.

 

FYI, Core was at 1040, Mem Clock was at 1500, Power Limit at 0 and Voltage at 0 (This is the card at stock/ Sapphire's factory overclock).

 

Temps on the single card never went above 76c and I ran the benchmark for at least an hour and a half. I also found this to be the case in while gaming. Credit to Sapphire's Nitro cooler.

 

The benchmarks for the Crossfire setup with both cards at stock and at the same settings as before:

 

FPS: 84.2

Score: 3525

Min: 25.8

Max: 143

 

Temps on the top card reached and stayed at around 86/87c in the three hours that I ran the benchmark, this is the card that the display port cable was connected; my primary card. The second card reached and stayed around 77/78c. Both cards had 100% utilisation.

 

Again, respectable and a clear improvement. I'm not sure how to work out if these gains are a reasonable return on my investment (RoI), I spent £268 on the second card. Regardless, I was going to do purchase the second card as I've never tried Crossfire before and I wanted to try and clear the mist surrounding the benefits and drawbacks. Some people rave about it and others say it is always more trouble than its worth. At this point I was more than happy I took the plunge. 

 

I should say that in most of the forums I inquired about crossfire the most common answer was 'Get a 980ti' or 'Wait for Pascal/Polaris'.

 

To respond to the first I want to make clear that I have  Freesync monitor, and feel I would be wasting a really enjoyable aspect of my rig by forfeiting Freesync and buying an Nvidia card.With that in mind I will want to stay with AMD until I replace my monitor which wont be soon as it is a recent purchase.

 

Hence my response to the second most frequent comment is that the first wave of Polaris GPU, the Polaris 10 and 11, will focus more on performance per watt and my research has led me to believe that the next enthusiast level cards that really show massive gains in performance wont be until Vega and that is due for release in 2017. This I suppose is my justification, if I even need one. 

 

To be continued.

 

 

In the next post I will be writing up real game benchmarks and experiences. 

 

Thanks for reading.

 

 

 

 

[latentjones]

Real Games One: The Witcher 3

 

One of the main reasons I wanted to get another R9 390 was to see if I could really get The Witcher 3 at ultra settings, at 1440p, and consistently above 60fps. 

 

I found that my single 390 really was performing well at 1440p with some settings turned down. For those that are running a single R9 390 at 1440p I found these settings to be the sweet spot for performance/quality:

 

Graphics Tab 

VSync: on

Max Frames Per Second: Unlimited

Resolution: 2560x1440

Display Mode: Fullscreen

Nvidia HairWorks: Off

Nvidia HairWorks AA: 0

Nvidia HairWorks Preset: Low

Shadow Quality; High

Terrain Quality: Ultra

Water Quality: High

Grass Density: High

Texture Quality: Ultra

Foliage Visibility Range: High

Detail Level: Ultra

Hardware Cursor: Off

 

Postprocessing Tab

Select 'High' and then adjust to taste. I find Motion Blur and DoF effects to be down to the player's palette. 

 

With these settings on a single 390 I was able to range between 50 - 80 fps depending on the frame but mainly stuck around 70fps. This was especially smooth and nice with Freesync when in the low 50's I barely noticed. 

 

But what about Crossfire?

 

So, as I said at the beginning of this post I was most excited to see what more I could squeeze out of the settings and frame rates when using two 390's. The first time I started the game I didn't check Radeon Crimson settings I just clicked play. 

 

After the title screen and the loading screen the first fame loaded up. I was pleasantly surprised. At the settings I stated above the fps counter was reading 115fps - Sweet. Then I started to move about. Oh not so sweet. I didnt experience any of the flickering i'd heard about before during research but the hitching was worse than playing at 30fps (personally speaking). 

 

i played for as long as I could stand it, maybe 25 minutes. Then it was back to desktop and I consulted the Radeon Crimson settings. I looked at the crossfire profile tab and found, to my luck and 'AMD pre-defined profile'. Bonus. 

 

I selected it and it changed all my Radeon settings. I booted the game up once more.

 

Yes the hitching had been reduced but it was certainly not gone entirely. It was more than enough to notice still. I Set everything but AA to full and managed 45-65fps. Pictures below.

 

Verdict

 

Even after playing around in Crimson settings and using various settings prescribed by people online I couldn't find any that offered smooth gameplay. I honestly felt I had a better experience running a my single 390 at the settings I mentioned at the start of this post with Freesync enabled. Sure I don't have all the eye candy turned up and my the gap between my minimum fps on a single card and my max in Crossfire could be as much as 50fps, the consistency and the playablity on a single card was miles away from the dual card setup.

 

I will continue to research Crossfire profiles for the Witcher 3 but in the meantime i will play using only a single card, Freesync and the aforementioned settings.

 

TL;DR

 

1-0 single card setup. Hitching in Crossfire was too unbearable. Pictures below for the word averse.

 

Thanks for reading.

 

Next I will be providing the same information for Crysis 3. 

Spoiler

It plays much better!

 

 

 

 

 

 

 

[latentjones]

The First Hurdle

 

My next post was going to cover Crysis 3 performance in Crossfire, but things don't always go to plan.

 

I was browsing the internet watching Youtube videos about watercooling and my PC shutdown. Weird I thought. I tried to turn it back on again but no joy. Having only just build the PC I wondered if I had missed something or the PSU was faulty. I opened up the rig and checked to see if the case power switch was properly attached - it was. I truned my attention to the power supply and tried to see if i could get a spare case fan to spin with a molex 3pin adapter hooked up straight into the PSU. The fan span so I took some solace in the fact that the PSU was at least working to some degree.

 

This is where some dread set in as I thought it could be an issue with my motherboard. I took both 390's out and - bingo! the power worked. CPU fan case fans etc. I hooked it up to my monitor and the computer was fine, displaying using the integrated graphics. So it was one of my 390's.

 

The Psu is 850w and I have no overclocks so I know I'm within limits and the setup had worked fine for a day. I suspected the newer card. But when I tried it worked. The second card, the oldest (but not by much maybe two weeks), wouldn't work in tandem or on its own in either PCIE lane. 

 

I rang Ebuyer and luckily they are going to pick it up and hopefully repair or replace it. 

 

Has anyone had any issues like this before? Where the card works fine for a day and then gives up the ghost?

 

Anyway, I can't run the benchmarks and playability tests I wanted to so I thought I'd give an update.

 

As soon as I get the new card I will continue to perform tests. I think if I don't get over sixty percent of success in the games i play I will sell one or both of the cards and buy a Fury X. Any thoughts?

 

To be continued

 

Thanks for reading.

 

TL;DR 

 

CF config is on hold as I had to RMA one of the cards. Will be continuing when replaced.

[ritchievillavicencio]

Hey, any update on this, do you get your card replaced? I´m trying to get my MSI R9 390 CF to work, but not sure if settings on Radeon settings are the correct ones, while CF can be enabled from Radeon settings most "guides" i found on the net are outdated and show it should be activated from old CCC panel.

 

Anyways, I will install a SSD for my OS today and I hope with a fresh Win10 install and a SSD performance improves.I have seen some improvements on some games but nothing WOW! so far, yet I need to do more tests. Every game I have runs smoothly with one 390, yet I wanted to squeeze some of the CF power so I can record or maybe stream and reduce the impact this can have on games, and as you just to realize if CF really worth it.

 

One question, I also need to buy a new screen, as my current one is vga (so i need an active convertor to use the 390) you thinks this can affect badly on the CF performance? I was thinking on trying my 1080 TV with hdmi and see if improves, but I'm not sure the vga is even a thing to consider on this equation.

[latentjones]

20 hours ago, ritchievillavicencio said:

Hey, any update on this, do you get your card replaced? I´m trying to get my MSI R9 390 CF to work, but not sure if settings on Radeon settings are the correct ones, while CF can be enabled from Radeon settings most "guides" i found on the net are outdated and show it should be activated from old CCC panel.

 

Anyways, I will install a SSD for my OS today and I hope with a fresh Win10 install and a SSD performance improves.I have seen some improvements on some games but nothing WOW! so far, yet I need to do more tests. Every game I have runs smoothly with one 390, yet I wanted to squeeze some of the CF power so I can record or maybe stream and reduce the impact this can have on games, and as you just to realize if CF really worth it.

 

One question, I also need to buy a new screen, as my current one is vga (so i need an active convertor to use the 390) you thinks this can affect badly on the CF performance? I was thinking on trying my 1080 TV with hdmi and see if improves, but I'm not sure the vga is even a thing to consider on this equation.

 

Hi there. My 390 replacement is being delivered tomorrow so hopefully I will be able to run some more tests straight away. 

 

Ok, so are you running the latest Crimson Drivers? I'm on the 16.5.2 but I was running the 16.5.1 when I had both cards. 

 

What games do you play and at what resolution? I would certainly try the 1080 tv over hdmi but if you can I would get a 1440 monitor or 4K one asap. It's here that you will see the benefits of the two cards and really the benefits of the 390 in general. In the meantime I would hook up your 1080 tv and I would enable DSR in the Crimson control panel/Radeon settings so that you can get a taste of 1440 on your 1080 tv/monitor.

 

I'm not sure if the vga directly affects performance but I would strongly suggest upgrading your monitor.

 

So to recap, let me know what games you play, if you have the latest Crimson drivers and what resolution your vga screen is. 

 

I will be straight back on the detailed gameplay and benchmark reviews as soon as I get my second card back tomorrow.

 

Good luck!

[ritchievillavicencio]

5 minutes ago, latentjones said:

 

Hi there. My 390 replacement is being delivered tomorrow so hopefully I will be able to run some more tests straight away. 

 

Ok, so are you running the latest Crimson Drivers? I'm on the 16.5.2 but I was running the 16.5.1 when I had both cards. 

 

What games do you play and at what resolution? I would certainly try the 1080 tv over hdmi but if you can I would get a 1440 monitor or 4K one asap. It's here that you will see the benefits of the two cards and really the benefits of the 390 in general. In the meantime I would hook up your 1080 tv and I would enable DSR in the Crimson control panel/Radeon settings so that you can get a taste of 1440 on your 1080 tv/monitor.

 

I'm not sure if the vga directly affects performance but I would strongly suggest upgrading your monitor.

 

So to recap, let me know what games you play, if you have the latest Crimson drivers and what resolution your vga screen is. 

 

I will be straight back on the detailed gameplay and benchmark reviews as soon as I get my second card back tomorrow.

 

Good luck!

Good to know your are getting your 2nd GPU back soon, I just re-installed Win 10 on the new SSD so I need to "re-install my games" aka make steam/origin/etc recognize the games back from the HDD. I have Witcher (all of them), Tomb Raider and Rise of the Tomb Raider, Crysis 3, Battlefield 3 and 4, GTA V, World of Warcraft, Diablo 3, Heroes of the Storm, Overwatch, Stracraft 2, Titalfall, Torchlight 2, and probably others I don't remember at the moment, plus a few others I have to install from my humblebundle library, like Bioshock infinite, and some others.

 

I play on 1080 on my VGA monitor, I do indeed change it for a better one, checking some options right now, especially since it is a vga and the GPUs have no vga port, so i'm using a active converter which it works but is not optimal no doubt about it.

 

I downloaded the latest AMD drivers last night (i'm not sure of the version and can't check right now) but I will confirm later today. yet I was wondering if I should try the beta ones or stick with the regular ones.

 

 

I did tried Witcher 3 last night after windows re-install and with hairworks off could hit around 60fps in 1080 on ultra settings/fullscreen, but there was some weird effect on the mountain rocks (like the ones near the balcony on the witchers castle aka first scene), and on the hair. I wanted to try your settings but it was late so I will try tonight.

 

I did notice a few days ago, some games behave better on virtual super resolution, yet I have to try some of those resolutions and see which one is better.

 

 

 

[ritchievillavicencio]

Radeon Software Version - 16.5.2.1
Radeon Software Edition - Crimson
Graphics Chipset - AMD Radeon (TM) R9 390 Series
Memory Size - 8192 MB
Memory Type - GDDR5
Core Clock - 1040 MHz
Windows Version - Windows 10 (64 bit)
System Memory - 16 GB
CPU Type - Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz
 

[ritchievillavicencio]

I tested with an hdmi Screen, there is an slight increase on FPS but witcher 3 still looks funny under crossfire, so I guess the profile for Witcher 3 still needs some work.

World of Warcraft runs fine, but there is not significant gain over crossfire, I will try The Division today and maybe some GTA V. is a shame but I think crossfire is not such a big gain, too bad because it looks awesome with the two MSI R9 390's inside my rig.

 

I wonder if there is any extra use for the crossfire besides bitcoin on which i´m not really interested. I was hoping I can use open broadcaster with it to reduce FPS drops while streaming but I don't see an increase on performance, maybe with another streaming app will work, also I want to try FRAPS or the MSI or AMD recording app.

[latentjones]

Sounds like you have been through it. I thought HDMI would be a better choice than VGA - impressed that that transpired as higher fps.

 

I got my second 390 back today. I put it in my rig and fired up Valley again but I found there was hitching now. The score and FPS benchamrked the same as before but now hitching... Not impressed. 

 

I ran GTA V and found that I gain very little FPS at all. In some places there were big boosts like standing in a corner near a building but running and gunning down a busy street was no increase and it was not smooth at all. There is no custom profile and there is little information online. i found a few fixes but that meant playing with Bois and registry files - so much hassle. 

 

I fired up The Forest (early access so I wasn't expecting much) and found the second card was at 0% usage the whole time.

 

Then I fired up Doom (new so I wasn't expecting much) and again no second gpu usage - 0%.

 

I fired up Skyrim without mods for fun and even then I was only dancing around the 60fps mark and below. Strange because when the the gpu I sent back was in the config I found that I was hitting the limit of my refresh at 144fps. When I disabled crossfire today and ran Skyrim again I was back at the 144fps but with only one gpu. Funny cos the carts in the into sequence flip around everywhere because of the engine's physics hating much above 100fps.

 

I must say, I'm really not impressed. Everything now plays worse that it did when I had the card I had to replace was in the config. All the settings are the same so are the drivers. I'm bemused. 

 

I'm also not a fan of crossfire. 

 

I'm selling both 390's and I'm going to get either a Fury x or a 1080 but here's my conundrum - I have a freesync monitor. 

 

Benchmarks online put the 1080 10-20fps ahead the of Fury x @1440 but does that matter if I'm going to just get screentearing or have to be forced into using vsync again after the creamy smooth of Freesync?

 

I could lock the 1080 at 60 or 120fps and forget about Freesync or buy a Fury X and not gain as many raw frames?

 

Dilemma.

[ritchievillavicencio]

I tried GTA V last night (1080p only) and it runned smooth, all settings to full except the ones in advance graphics sections (high def shadows, etc), I got around 80-90FPS average.

 

I Benchmark Rise of the Tomb Raider on both dx 11 and 12, and this are the results:

 

Settings      DX     FPS

High            11      83

Ultra            11     72

High            12     76

Ultra            12     59

 

It looks like DX12 is not well optimized just yet on this game, but it show less "fps drops" than DX11.

 

I hope to try some more games during this weekend. So far I believe crossfire and probably SLI is not as plug and play as many of us hope.