I am about to buy a graphics card. 970 vs R9 390.

[TheElt]

Those are 3Dmark scores, meaning the entire test. Including the combined test. It's a 4.7 vs. a 4.8ghz 5960X, that accounts for the difference. 3Dmark is also prone to variance. No two tests are the same.

Still nothing on 4GB vs 8GB. Also, those people also overclock memory heavily, something you can't do on air. Something a little less anecdotical and prone to variance would be better.

You see that Fire strike result? You see the little subheadings under the score? You see that little one marked "GPU Score"? That's the score from the GPU benchmark, not the combined that is the end reported number.

Both those fire strike scores are on air, on stock cooling. My 970 does 1630/2060 on air, memory does not benefit from water cooling, only the core does. Show me a point where 8GB is worse than 3.5GB. Fire strike is as consistent as you can get comparing GPUs, the only variance with come from minor driver tweaks (Anything major and the result will be invalid) and Windows optimisations (Which with GPUs it's just disabling Aero and moving to classic)

[App4that]

Has nothing to do with backdoor politics or developer relationships. They just have higher drawcall throughput. It shows in the API overhead tests of Futuremark.

http://images.anandtech.com/graphs/graph9112/73050.png

http://images.anandtech.com/graphs/graph9112/73019.png

Look at DX11ST and MT with the nvidia cards and amd cards. Nvidia pulls 2.2million on DX11MT, whilst amd only gets 1.1million. Half the drawcalls.

Yes, but the day one performance is directly tied to the relationship with the developer.

And my 980ti isn't overkill at 1080p. I actually have to sacrifice settings. More than I think I should have to honestly but that's because of my CPU not the 980ti. You can mod most games to bring any graphics card to its knees.

[Majestic]

You see that Fire strike result? You see the little subheadings under the score? You see that little one marked "GPU Score"? That's the score from the GPU benchmark, not the combined that is the end reported number.

Show me a point where 8GB is worse than 3.5GB. 

 

So its 15628 vs 15108. 3.4% higher. Is this supposed to be some type of smoking gun?

 

And no no no, the 8GB is used as an argument why the 390 is better. As a counter against the lower power consumption, less heat and noise and shadowplay. So it must be atleast of some relevance beyond one game on one specific setting.

 

 

Yes, but the day one performance is directly tied to the relationship with the developer.

And my 980ti isn't overkill at 1080p. I actually have to sacrifice settings. More than I think I should have to honestly but that's because of my CPU not the 980ti. You can mod most games to bring any graphics card to its knees.

 

True, but you have more headroom with that drawcall throughput. Especially when games on consoles are designed to already work with low-overhead API's.

 

And true to some extent. It's designed for 1440p though. 16xAA is something that will always bring cards down

[App4that]

So its 15628 vs 15108. 3.4% higher. Is this supposed to be some type of smoking gun?

And no no no, the 8GB is used as an argument why the 390 is better. As a counter against the lower power consumption, less heat and noise and shadowplay. So it must be atleast of some relevance beyond one game on one specific setting.

True, but you have more headroom with that drawcall throughput. Especially when games on consoles are designed to already work with low-overhead API's.

And true to some extent. It's designed for 1440p though. 16xAA is something that will always bring cards down

LOL, I hack the ini files very aggressively. To the point of losing track of how many I write. Really need to start keeping track of shit like that.

I'd run DSR but I don't like it as much as VSR. Though if I have time this weekend I'm going to give it another shot.

[TheElt]

So its 15628 vs 15108. 3.4% higher. Is this supposed to be some type of smoking gun?

And no no no, the 8GB is used as an argument why the 390 is better. As a counter against the lower power consumption, less heat and noise and shadowplay. So it must be atleast of some relevance beyond one game on one specific setting

3.4% faster, for cheaper, with more VRam, better multi GPU scaling. That's two good clocking chips, the gap stays roughly the same at equivalent clocks for each

A non-reference 970 consumes only 20-60W less than a 390, compare the numbers I gave you for my card to numbers people have put up for the 390 and you'll see. Even 100W is a tiny bit extra on your electricity bill, and will not be an issue unless you're going extreme SFF where you're power and space limited to a big degree. Hotter and louder point is moot, board partner coolers vary, my G1 970 is louder than say an XFX 390 with the stock fan profile because shitty fans, and with my custom profile it runs hotter than an XFX 390 but is quieter. Heat and noise are a reference cooler issue, which is why they're avoided like the plague.

Shadow play is the one benefit that keeps me from swapping over to AMD, it's convenient and records in good quality though it would be nice to support 120fps@1080p at the least. Raptr is cancer and needs to be fixed or die

[Nena Trinity]

Seeing as you two are so convinced you're giving the correct advice you don't even feel the need to explain yourselves, you'll have no trouble making counter arguments for this.

http://linustechtips.com/main/topic/533899-r9-390-vs-gtx-970/page-3#entry7084736

Or rebuttals to what transpired in this topic.

Only the 3 first is right, also unless he play Just Cause or any other tesselation heavy game I see no reason to not grab 390 over 970!

[ivan134]

Only the 3 first is right, also unless he play Just Cause or any other tesselation heavy game I see no reason to not grab 390 over 970!

You've fallen into the trap. Good luck.

[Majestic]

3.4% faster, for cheaper, with more VRam, better multi GPU scaling. That's two good clocking chips, the gap stays roughly the same at equivalent clocks for each

A non-reference 970 consumes only 20-60W less than a 390, compare the numbers I gave you for my card to numbers people have put up for the 390 and you'll see. Even 100W is a tiny bit extra on your electricity bill, and will not be an issue unless you're going extreme SFF where you're power and space limited to a big degree. Hotter and louder point is moot, board partner coolers vary, my G1 970 is louder than say an XFX 390 with the stock fan profile because shitty fans, and with my custom profile it runs hotter than an XFX 390 but is quieter. Heat and noise are a reference cooler issue, which is why they're avoided like the plague.

Shadow play is the one benefit that keeps me from swapping over to AMD, it's convenient and records in good quality though it would be nice to support 120fps@1080p at the least. Raptr is cancer and needs to be fixed or die

 

Better crossfire scaling, i refuse to believe that. Especially since CPU's already have a hard time saturating one 390 in games like Witcher 3, Fallout 4, etc. Maybe in synthetic benchmarks. Yes it has more vram, but show me something (please) that shows this to be more than just a marketing tactic. 100W for 4 hours a day over a year is 145kw/h, for Dutch energy pricing that's 40 euro a year. Assuming you keep it for 2-2.5 years, that's 80-100 euro. It is a big deal if you don't live in America..

 

But the card still puts out 100W more energy. Meaning your case has to dissipate this heat. It means that you need more airflow. Also, you can't make coolers that efficient that extra heat won't add to any noise.

 

http://nl.hardware.info/reviews/6176/11/amd-radeon-r9-390x-review-asus-vs-club-3d-vs-msi-vs-sapphire-koeling-geluidsproductie-en-stroomverbruik

 

Quietest 390 is 48dB, quietest 970 is 38dB. Yes, there is also a loud 970 in the mix, but no 38dB 390 in the mix.

 

 

Only the 3 first is right, also unless he play Just Cause or any other tesselation heavy game I see no reason to not grab 390 over 970! 

 

Because? If performance is moot and pretty much the same across the board (with some wins to the 970 and some to the 390), and you agree on the first 3 points. Why again is it the better choice? Especially given AMD's lacklustre driver support on new games.

 

I thought we were advising products for average gamers, not benchmarkers.

[Kinda Bottlenecked]

As the title says, I am just about to buy a new graphics card for my system... I am between the GTX 970 or the R9 390 but I can't decide which one and why should I get. I have around 350 U$D and my computer has this specs.

 

http://pcpartpicker.com/p/24RBwP

 

Thanks in advance. 

 

Depends on the features you like and what games you play. You don't have to worry about driver overhead since your CPU would handle it just fine. 

 

http://linustechtips.com/main/topic/532947-970-or-390-overclocking/#entry7073233 

You can reference my thread to help you decide.  

[Majestic]

You don't have to worry about driver overhead since your CPU would handle it just fine. 

 

http://pclab.pl/zdjecia/artykuly/chaostheory/2015/12/gfx_sapphire_fury/charts/oc/gta5w_1920vh.png

 

4.7 GHZ 6700K.

 

And yes I know the Fury-X isn't slower than the 970.

[ivan134]

LOL, I hack the ini files very aggressively. To the point of losing track of how many I write. Really need to start keeping track of shit like that.

I'd run DSR but I don't like it as much as VSR. Though if I have time this weekend I'm going to give it another shot.

Whats wrong with DSR? There are also other ways to downsample that people have been using for ages before AMD and Nvidia supported it officially. Nvidia actually supported it their control panel for a long time before they gave it a name.

[Kinda Bottlenecked]

http://pclab.pl/zdjecia/artykuly/chaostheory/2015/12/gfx_sapphire_fury/charts/oc/gta5w_1920vh.png

 

4.7 GHZ 6700K.

 

And yes I know the Fury-X isn't slower than the 970.

 

4.5Ghz i7 6700k

 

 

 

 

 We tested the game with everything set to very high and MSAA off

[App4that]

Whats wrong with DSR? There are also other ways to downsample that people have been using for ages before AMD and Nvidia supported it officially. Nvidia actually supported it their control panel for a long time before they gave it a name.

Honestly, I'm lazy. With VSR you enable it and select the resolution in the game's launcher. With DSR you have to pick the percentage you want to go over the native, and play with the sharpness. I got bored with having to leave the game to adjust the sharpness so turned it off.

 

1 point for team red.

[Majestic]

4.5Ghz i7 6700k

 

 

 

They get scaling to 110 fps in GTA 5. They were sure as hell not in a city, rather in some desert facing the floor. Even DigitalFoundry doesn't get those results in their CPU tests. (80ish for a 6700K 4.4ghz)

[App4that]

4.5Ghz i7 6700k

 

980 right there with a Fury X? Damn, go 980!

[Kinda Bottlenecked]

They get scaling to 110 fps in GTA 5. They were sure as hell not in a city, rather in some desert facing the floor. Even DigitalFoundry doesn't get those results in their CPU tests. (80ish for a 6700K 4.4ghz)

 

They probably tested in a mix of areas. Point is, The driver overhead ain't as bad as what you showed in the chart. How long ago was that benchmark anyway? 

[Majestic]

They probably tested in a mix of areas. Point is, The driver overhead ain't as bad as what you showed in the chart. How long ago was that benchmark anyway? 

 

Probably? 

 

Point is, in a game that is so reknown for being CPU heavy, PCLAB.pl's results make more sense. 980 almost rivalling the 980TI because both are CPU bound and reach diminishing returns. Whereas the 970 SLI scales 80% in your test. Sorry, PCLAB's results are much more realistic. Specially given this video, showing the exact same numbers;

 

 

Results have to make sense, not only corroborate your story. 

 

Recent test. 23dec 2015

[Nena Trinity]

Better crossfire scaling, i refuse to believe that. Especially since CPU's already have a hard time saturating one 390 in games like Witcher 3, Fallout 4, etc. Maybe in synthetic benchmarks. Yes it has more vram, but show me something (please) that shows this to be more than just a marketing tactic. 100W for 4 hours a day over a year is 145kw/h, for Dutch energy pricing that's 40 euro a year. Assuming you keep it for 2-2.5 years, that's 80-100 euro. It is a big deal if you don't live in America..

But the card still puts out 100W more energy. Meaning your case has to dissipate this heat. It means that you need more airflow. Also, you can't make coolers that efficient that extra heat won't add to any noise.

http://nl.hardware.info/reviews/6176/11/amd-radeon-r9-390x-review-asus-vs-club-3d-vs-msi-vs-sapphire-koeling-geluidsproductie-en-stroomverbruik

Quietest 390 is 48dB, quietest 970 is 38dB. Yes, there is also a loud 970 in the mix, but no 38dB 390 in the mix.

Because? If performance is moot and pretty much the same across the board (with some wins to the 970 and some to the 390), and you agree on the first 3 points. Why again is it the better choice? Especially given AMD's lacklustre driver support on new games.

I thought we were advising products for average gamers, not benchmarkers.

GTX 970 pros:

*Better day-one driver support.

*Less heat

*less power

•Does not suck on Linux!

•Tessellation!

R9 390 pros:

•8GB VRAM!

•Async Compute!

•Better at higher resolutions! (1440p & 4K)

•Better performance mostly... (1080p)

•Crossfire > SLI...

•FreeSync! (Free unlike Gsync...)

•Often cheaper! (Not always...)

[ivan134]

Honestly, I'm lazy. With VSR you enable it and select the resolution in the game's launcher. With DSR you have to pick the percentage you want to go over the native, and play with the sharpness. I got bored with having to leave the game to adjust the sharpness so turned it off.

1 point for team red.

Go to the nvidia control panel and look for the "change resolution" menu. Enable custom resolutions and create the resolutions you want. The new resolution options wil appear automatically in your games.

[Majestic]

GTX 970 pros:

*Better day-one driver support.

*Less heat

*less power

•Tessellation!

R9 390 pros:

•8GB VRAM!

•Async Compute!

•Better at higher resolutions! (1440p & 4K)

•Better performance mostly... (1080p)

•Crossfire > SLI...

•FreeSync! (Free unlike Gsync...)

•Often cheaper! (Not always...)

 

8GB vram! yes, more #numbers, more #cores. AMD's only selling point. How does it matter though beyond the marketing value, for the 10th time i've asked you people.

Async compute, unproven tech for games that haven't released yet. Strong point.

both 390 and 970 suck at 1440p/4K high settings, even my 980 @ OC struggles to maintain 60 in even 2013 titles (tomb raider)

1080p it's pretty much dead even on factory clocked cards at best, 970 winning by landslide at worst (GTA 5, Fallout 4, "gimpworks" titles that were nominated games of the year but i'm sure noone plays witcher 3)

crossfire scaling is lacklustre given the driver overhead in some games. Not that SLI is much better, single GPU ftw.

Freesync, AH ys that technology that was only available thanks to G-sync, and is still requiring R&D from scaler manuf. to approach G-sync. The panels that do 1-144hz on freesync aren't fucking cheap, only the 75-90hz panels are.

 

And how can you say better day-one driver support, but at the same time say the card performs better. Obviously not with bad day-one driver support. Or are AMD buyers always months late with buying games?

 

Ah what the fuck, who the hell is even listening on this forum. This forum is a giant AMD echo-chamber anyway.

[Kinda Bottlenecked]

Probably?

Point is, in a game that is so reknown for being CPU heavy, PCLAB.pl's results make more sense. 980 almost rivalling the 980TI because both are CPU bound and reach diminishing returns. Whereas the 970 SLI scales 80% in your test. Sorry, PCLAB's results are much more realistic. Specially given this video, showing the exact same numbers;

Results have to make sense, not only corroborate your story.

Recent test. 23dec 2015

Alright, I looked through some gta 5 benchmarks from other sites and the pclab one is really fishy. Even on 1080p a 980 ti should be getting more than 68fps.

Even df's videos contradict pclabs testing. The fps was very well above 70 all the time. Even at a point it went to 140fps. But yet all the cards from pclab is underperforming by alot. I get an average of 60fps on my 390x with msaa turned on. The 980 ti should be considerably faster than that... Why is it only getting an average of 68fps on 1080p?

[ivan134]

Alright, I looked through some gta 5 benchmarks from other sites and the pclab one is really fishy. Even on 1080p a 980 ti should be getting more than 68fps.

Even df's videos contradict pclabs testing. The fps was very well above 70 all the time. Even at a point it went to 140fps. But yet all the cards from pclab is underperforming by alot. I get an average of 60fps on my 390x with msaa turned on. The 980 ti should be considerably faster than that... Why is it only getting an average of 68fps on 1080p?

You fell in into the trap too.

[Kinda Bottlenecked]

You fell in into the trap too.

Trap?

[App4that]

Trap?

shhhh! Ivan's got their hat on, the shiny one made of tin.

[ivan134]

Honestly, I'm lazy. With VSR you enable it and select the resolution in the game's launcher. With DSR you have to pick the percentage you want to go over the native, and play with the sharpness. I got bored with having to leave the game to adjust the sharpness so turned it off.

 

1 point for team red.

Also, you need to make sure you create different refresh rates for each of the resolutions if you have a monitor that's more than 60 Hz.