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

[Majestic]

If you take a moment to stop being a thick pr*ck I'd be arsed to have a proper conversation with you but alas that day is not yet here.

... congrats! You are disagreeing with a LOT of people! You win a prize! That prize? Being annoying and telling us to provide evidence when you dont provide any yourself! Good Job! 

 

and *criticize 

 

Oh i'm the prick for asking you people back up your claims that you've been making for months (and not the first time I might add)? This is some prime cognitive dissonance right here. I'm asking because you are making the claim, Mohenjo. It's called the burden of proof. 

 

Stop deflecting and answer the question i've asked three times now (in this post).

[don_svetlio]

Oh i'm the prick for asking you people back up your claims that you've been making for months (and not the first time I might add)? This is some prime cognitive dissonance right here. I'm asking because you are making the claim, Mohenjo. It's called the burden of proof. 

 

Stop deflecting and answer the question i've asked three times now.

I've been providing proof for a month now and every single time you dismiss it and bring up a random Polish site that nobody can read and say it's concrete proof. You are not worth anyone's time

[Majestic]

I've been providing proof for a month now and every single time you dismiss it and bring up a random Polish site that nobody can read and say it's concrete proof. You are not worth anyone's time

 

DigitalFoundry provides the same evidence. And none of that evidence shows 8GB being the main determining factor in the performance differences.

 

And again, if you can point me to a discrepancy in PCLAB.pl's results, please show me. Just saying "those are fake" isn't enough.

[Mohenjo]

Oh i'm the prick for asking you people back up your claims that you've been making for months (and not the first time I might add)? This is some prime cognitive dissonance right here. I'm asking because you are making the claim, Mohenjo. It's called the burden of proof. 

 

Stop deflecting and answer the question i've asked three times now (in this post).

Fine, Youtube is blocked at my school so ill just link you what i remember by name 

[don_svetlio]

DigitalFoundry provides the same evidence. And none of that evidence shows 8GB being the main determining factor in the performance differences.

Have I claimed the 8GB are going to give more performance? No

What I claim is that 8GB provides a better safety net for crappy ports and future games as well as offers better CF options and it's BETTER VALUE FOR THE MONEY

[Majestic]

Fine, Youtube is blocked at my school so ill just link you what i remember by name 

 

Fine, that's relative performance in a select number of games. Valid for choosing one or the other if one is your preferred game. It, however, does not provide any evidence that 8GB provides a tangible difference. Nor does it provide a counter argument to the things I've opted in my condensed summary (earlier). That looking at it from a product standpoint, what it offers to the consumer, the AMD cards offer very little else.

 

Might I add though, that those results are on a 1200mhz 390, which is not an average OC (mostly it's 1075-1150), and a severely crippled 970 (1442mhz, where 1520 is average). So it's golden sample vs. turd. And the differences are very small, which is what i've been saying all along. I know the cards stack up performance wise, but there is more to a card than that.

 

 

Have I claimed the 8GB are going to give more performance? No

 

The other guy did. And futureproofing isn't viable, look at the 8350 vs. 2500K. "Games will be more multithreaded in the future". Yeah, we've surely seen that happening. And as for those crappy ports. They use the memory you have and never scrub it or clean it up, if your card has 6GB it will fill up 6GB eventually (given the time). 

[Mohenjo]

Fine, that's relative performance in a select number of games. Valid for choosing one or the other if one is your preferred game. It, however, does not provide any evidence that 8GB provides a tangible difference. Nor does it provide a counter argument to the things I've opted in my condensed summary (earlier). That looking at it from a product standpoint, what it offers to the consumer, the AMD cards offer very little else.

 

Might I add though, that those results are on a 1200mhz 390, which is not an average OC (mostly it's 1075-1150), and a severely crippled 970 (1442mhz, where 1520 is average). So it's golden sample vs. turd. And the differences are very small, which is what i've been saying all along. I know the cards stack up performance wise, but there is more to a card than that.

jesus christ man, your clinging on nothing, what makes the god damn 970 better? Huh? what? is 4gb better than 8gb? F U C K N O. Plus when dx12 comes out its gonna kick the 970s ass. 

[Majestic]

jesus christ man, your clinging on nothing, what makes the god damn 970 better? Huh? what? is 4gb better than 8gb? F U C K N O. Plus when dx12 comes out its gonna kick the 970s ass. 

 

No you're projecting, you still haven't provided the evidence I asked for. 

 

EDIT; I've explained my standpoints here;

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

 

Looking at it from a consumer/product standpoint. Not mere performance, because that is (across the board) pretty much the same. Except for a few new games where AMD's drivers struggle. But i'm not even using those as a focal point.

[don_svetlio]

 

Fine, that's relative performance in a select number of games. Valid for choosing one or the other if one is your preferred game. It, however, does not provide any evidence that 8GB provides a tangible difference. Nor does it provide a counter argument to the things I've opted in my condensed summary (earlier). That looking at it from a product standpoint, what it offers to the consumer, the AMD cards offer very little else.

 

Might I add though, that those results are on a 1200mhz 390, which is not an average OC (mostly it's 1075-1150), and a severely crippled 970 (1442mhz, where 1520 is average). So it's golden sample vs. turd. And the differences are very small, which is what i've been saying all along. I know the cards stack up performance wise, but there is more to a card than that.

 

 

 

The other guy did. And futureproofing isn't viable, look at the 8350 vs. 2500K. "Games will be more multithreaded in the future". Yeah, we've surely seen that happening. And as for those crappy ports. They use the memory you have and never scrub it or clean it up, if your card has 6GB it will fill up 6GB eventually (given the time). 

 

Go look at recent games. 8350 outperforms the 2500K in every AAA title released in 2015

[Mohenjo]

No you're projecting, you still haven't provided the evidence I asked for. 

the 8gb helps the 390 in higher resolutions, Thats just common sense man, true in 1080p it wont effect much RIGHT NOW but still.

 

Here is evidence for that 

 

 

But its not the ONLY thing that helps it ok? jesus man you really like trying to be correct about shit. 

[don_svetlio]

Average OC as reported by HW Bot is 1187 for a 390 btw

[don_svetlio]

Most people get 1200 and some have managed 1270

[Majestic]

Go look at recent games. 8350 outperforms the 2500K in every AAA title released in 2015

 

 

Not really. I can provide more if promted. Buy I generally think DF does a good enough job.

[Majestic]

the 8gb helps the 390 in higher resolutions, Thats just common sense man, true in 1080p it wont effect much RIGHT NOW but still.

 

Here is evidence for that 

 

But its not the ONLY thing that helps it ok? jesus man you really like trying to be correct about shit. 

 

You are the person swearing, not sure why you think you can reach for the moral highground here. 

 

Well, those results corroberate your story and it's from DF so it's legit. However, you also see the 290 on the same graph. Which has about the same memory bandwidth as the 970 (when corrected for the delta coloring Nvidia uses), though it's a bit faster (which is why it still wins in some results). If you look at the 390 however, it's memory was overclocked from 5ghz to 6ghz. Same like what happened to the GTX770 when it was rebadged from the GTX680 and provided most of the performance gain (since kepler cards were crippled by memory bandwidth). Higher resolutions put more strain on the memory bandwidth, having upped it by 20% in terms of speed would certainly net some performance gains. 

 

Which is coincidentally just about the difference between the results. Also, those cards are pretty terrible at 4K. And will never in the future be viable for 4K. Even less so in the future, since games will only become heavier to run. I'll be willing to bet that overclocking the memory on the 970 from 7 to 8ghz would put them more closely together. I've seen as much from my own testing on 1440p on the 980 to say that it's memory bandwidth is a limiting factor to a certain degree (it has diminishing returns after 8ghz).

 

This is some evidence, but not conclusive.

[Mohenjo]

You are the person swearing, not sure why you think you can reach for the moral highground here. 

 

Well, those results corroberate your story and it's from DF so it's legit. However, you also see the 290 on the same graph. Which has about the same memory bandwidth as the 970 (when corrected for the delta coloring Nvidia uses), though it's a bit faster (which is why it still wins in some results). If you look at the 390 however, it's memory was overclocked from 5ghz to 6ghz. Same like what happened to the GTX770 when it was rebadged from the GTX680. Higher resolutions put more strain on the memory bandwidth, having upped it by 20% in terms of speed would certainly net some performance gains. 

 

Which is coincidentally just about the difference between the results. Also, those cards are pretty terrible at 4K. And will never in the future be viable for 4K. Even less so in the future, since games will only become heavier to run.

Im done here, you are obviously a -snip- who thinks theirs a "morel high ground" in an internet argument. I really dont care what you think and i dont care if you think me swearing means im somehow a lesser being than you, and using large words to insult someone all of sudden somehow makes you look like the better person, it really doesn't. You started this argument just to have it knowing full well this will happen. The only people i have seen do that in real life is my 6 year old cousin. Congrats, enjoy your life and i wish you the best. Also *corroborate If you're gonna use large words like that might as well spell them right, thats what spell check is for. Edited January 26, 2016 by alpenwasser
language

[TheElt]

If you want performance and don't care much for anything else, 390 definitely. If you care about stuff like Shadowplay, then go for the 970 (It's nice to be able to capture the past 5 minutes if something cool happens in a game), but the 970 will hit the 3.5GB issue in some games when you crank everything to max at even 1080p, and the 390 will scale better if you add another one in the future but draws a little bit more power than the 970 (But not that much more).

 

As said before, if you haven't bought the rig already I'd reccommend swapping that SSD out for something else, I have one and I'm disappointed with it to say the least, and the PSU is pretty overkill

[Majestic]

but the 970 will hit the 3.5GB issue in some games when you crank everything to max at even 1080p,

 

(But not that much more).

 

This was not really the conclusion i've heard every major outlet mention after ramgate passed. It was only in extreme scenario's at which the framerates were sub 30 anyway. If you have edivence proving otherwise, i'd love to see it. Lest I be speaking untruths..

 

(a good 80-100 watts more, differs by game). I'd say that is a tad bit more though.

[App4that]

You know what, I'm going to change this up.

Get the 390, have problems, end up with a Nvidia card in the end. Saves me energy.

[TheElt]

This was not really the conclusion i've heard every major outlet mention after ramgate passed. It was only in extreme scenario's at which the framerates were sub 30 anyway. If you have edivence proving otherwise, i'd love to see it. Lest I be speaking untruths..

 

(a good 80-100 watts more, differs by game). I'd say that is a tad bit more though.

GTA 5 cannot be put to full max at 1080p without running into the 3.5GB issue, at those settings the card is more than capable of 40+fps at stock but tanks to single digits. I would test on more games but I don't own any modern AAA titles that push VRAM that hard (Like Shadow Of Mordor). People can say it isn't an issue, but if you want to max some games you can't, and in SLI it's even more of an issue unless DX12 fixes that and lets VRAM stack

 

Stock speeds (1392MHz) and volts my card gets up to 63% TDP, or roughly 157.5W. Stock volts, 1580MHz core and 2050MHz memory gets to 72% TDP, or roughly 180W. And 1610MHz @1.265v with 2065MHz memory clocks in at 80% TDP, or roughly 200W.

 

Maxwell is efficient because voltages, touch the voltage and all that goes out the window. The power gap is at most 100w worst case, most of the time in the realm of 40-60W which is literally nothing, because there are no constraints on power draw or heat output that would make that difference an issue

[Majestic]

Stock speeds (1392MHz) and volts my card gets up to 63% TDP, or roughly 157.5W. Stock volts, 1580MHz core and 2050MHz memory gets to 72% TDP, or roughly 180W. And 1610MHz @1.265v with 2065MHz memory clocks in at 80% TDP, or roughly 200W.

 

 

What are you measuring this with? Because 63% TDP would mean 0,63*145=91W. 

[TheElt]

What are you measuring this with? Because 63% TDP would mean 0,63*145=91W. 

TDP values set in the card's BIOS. 100% TDP for my G1 Gaming is 250W, since board partners can set their own TDP values, if I set my card up to have the same TDP limits as a reference Nvidia card (145W being 100%, 156W being a very generous 108% max) it would throttle out the gate with no tweaks

[Majestic]

TDP values set in the card's BIOS. 100% TDP for my G1 Gaming is 250W, since board partners can set their own TDP values, if I set my card up to have the same TDP limits as a reference Nvidia card (145W being 100%, 156W being a very generous 108% max) it would throttle out the gate with no tweaks

 

Well, i'd still rather trust actual intercepted power measurements over software measurement..

[TheElt]

Well, i'd still rather trust actual intercepted power measurements over software measurement..

Nvidia's power target is hardware based, uses shunt resistors and some funky things to get a number out, which is why you can short them or half their resistance to increase the TDP limit if the BIOS doesn't change it

[Nena Trinity]

390 seems like the better GPU in most cases to me!

[ivan134]

 

390 seems like the better GPU in most cases to me! 

 

Except Linux

Seriously, do yourself a favor and dual boot