Anyone still choosing 970 over 390?

[KarateHottie93]

I get that the R9 390 may average out to be a little faster, but it's things like this that bother me:

That's a frame latency graph from a Tech Report benchmark. Those spikes from the R9 390 are sudden jumps in the time it's taking to render the next frame, sometimes spiking nearly to 50 ms. For reference, at 60 FPS it takes 16.6 milliseconds to deliver each frame. At 30 FPS it's 33.3 ms, and 20 FPS at 50 ms. While those spikes are very quick and wouldn't be noticeable if they only happened rarely, games feel terrible when it happens a lot.

AMD competes with Nvidia easily in performance per dollar if you just look at average performance, but we absolutely cannot ignore that this happens. It's not every game, but it's often enough that buying an AMD card feels like a risk to me right now.

At least it can be fixed with software unlike Ramgate.

[Morgan MLGman]

I'm very tempted to get a 970 just to see what the difference is. But it would mean not getting the 4790k, so probably won't.

Don't. Now it's my turn to give you some first-hand experience ^^ I've had one and I can tell you that the 290X is better. I used my 970 for 3-4 months and I've had one of the best 970's MSI Gaming 4G 100ME Limited Edition one

For example, Tomb Raider benchmark with the exact same PC with the exact same settings:

Average FPS on 290X is 8,5FPS higher than MAX FPS on a 970... Nothing besides the GPU and the power supply changed since then in my PC.

[Mew]

Don't. Now it's my turn to give you some first-hand experience ^^ I've had one and I can tell you that the 290X is better. I used my 970 for 3-4 months and I've had one of the best 970's MSI Gaming 4G 100ME Limited Edition one

For example, Tomb Raider benchmark with the exact same PC with the exact same settings:

Average FPS on 290X is 8,5FPS higher than MAX FPS on a 970... Nothing besides the GPU and the power supply changed since then in my PC.

 

One game isn't indicative of every game. 

 

Anyways, I went with the 970 because it was much cheaper than the 290x and 390 at the time (280 after shipping and tax vs 330ish before shipping and tax). Performance, at least according to Anandtech's very recent results show that the 970 performs slightly better than the 390 in games that I play.

 

So, with those results and reasons in mind, I don't regret getting the 970 at all. 

[Hemanse]

I traded in my 390 for a 970, got a very hot chip with bad coil whine, which i guess kinda made my first experience with AMD kinda meh. Very happy with my new MSI 970, runs cool and absolutely silent, even overclocked to 1520/2000.

[don_svetlio]

One game isn't indicative of every game. 

 

Anyways, I went with the 970 because it was much cheaper than the 290x and 390 at the time (280 after shipping and tax vs 330ish before shipping and tax). Performance, at least according to Anandtech's very recent results show that the 970 performs slightly better than the 390 in games that I play.

 

So, with those results and reasons in mind, I don't regret getting the 970 at all. 

I'm having a hard time believing those benches. 960 on par with a 380X? highly unlikely

[Morgan MLGman]

I'm having a hard time believing those benches. 960 on par with a 380X? highly unlikely

Yup, especially that a 960 is 10-15% slower than a 380, let alone the 380X...

[ThomasD]

Given that we are at the front of the holiday shopping season, yet 970 prices are not significantly lower than 390 prices it would appear that many people still are buying 970s.

[typographie]

At least it can be fixed with software unlike Ramgate.

 

Well, that particular game is nearly a year old. Maybe it can be fixed with software, but when do they plan to get around to it?

 

I'm having a hard time believing those benches. 960 on par with a 380X? highly unlikely

 

Arguments like this really start to go off the rails when we just arbitrarily stop believing benchmarks posted by legitimate, trusted websites. If there's something going on in the benchmarks beyond what's at face value, fine, go read the article and figure out what's up. But the other possibility is that maybe there's just less distance between the GTX 960-R9 380-R9 380X than the preconceptions this forum community tends to encourage. That's a really crowded market segment… how many frames difference were you really expecting for ~$30?

 

Yup, especially that a 960 is 10-15% slower than a 380, let alone the 380X...

 

Where did that 10–15% figure come from? What games?

 

http://techreport.com/review/29316/amd-radeon-r9-380x-graphics-card-reviewed

Just went through this and did the math. Focusing solely on average framerates in all games in this article, the 2 GB version of the GTX 960 averaged out to 90.35% of the performance of the R9 380X. I didn't do the same for the 380, but it's basically within a frame or two of the GTX 960 in the games tested.

[Morgan MLGman]

-snip-

You see, I found those results you posted unreasonable so I decided to research it myself, I found a 380X test on the best, independent Polish tech site from around a week ago (that only means that the drivers are up to date etc.) that involves a 970 and 390 scores in the same games you cherry-picked. @don_svetlio you might wanna look at the benchmarks too ^^

GTA V:

Battlefield 4:

Crysis 3:

As you can see, the 390 beats a 970 in all the games you selected, even in Nvidia-optimized ones. There also are games where the difference is bigger, sometimes like 10FPS in favor of the 390, Considering it's a card that costs the same, I see no reason to go for the 970, especially with a shaky DX12 future and VRAM limitations.

[Morgan MLGman]

-snip-

I didn't say you get 10-15% FPS more than with the 960, I said the card is 10-15% faster cause it is. You get more FPS in basically every game. In some, even 10FPS more with a card that costs exactly the same amount of money.

[typographie]

I didn't say you get 10-15% FPS more than with the 960, I said the card is 10-15% faster cause it is.

 

If 10–15% faster doesn't mean 10–15% better framerate, then what exactly does that figure measure?  :huh:

[Morgan MLGman]

If 10–15% faster doesn't mean 10–15% better framerate, then what exactly does that figure measure?  :huh:

 

Ugh, computing power for instance? It's a better card regardless, there is literally 0 logical reasons to pick the 960 right now. It's crap. Hoping for a 960Ti to be better.

[don_svetlio]

Well, that particular game is nearly a year old. Maybe it can be fixed with software, but when do they plan to get around to it?

 

 

Arguments like this really start to go off the rails when we just arbitrarily stop believing benchmarks posted by legitimate, trusted websites. If there's something going on in the benchmarks beyond what's at face value, fine, go read the article and figure out what's up. But the other possibility is that maybe there's just less distance between the GTX 960-R9 380-R9 380X than the preconceptions this forum community tends to encourage. That's a really crowded market segment… how many frames difference were you really expecting for ~$30?

 

 

Where did that 10–15% figure come from? What games?

 

http://techreport.com/review/29316/amd-radeon-r9-380x-graphics-card-reviewed

Just went through this and did the math. Focusing solely on average framerates in all games in this article, the 2 GB version of the GTX 960 averaged out to 90.35% of the performance of the R9 380X. I didn't do the same for the 380, but it's basically within a frame or two of the GTX 960 in the games tested.

I just trust DigitalFoundry a hell of a lot more since there I at least get raw footage and numbers. I also don't see how certain publications have a 380X on par with a 960 whereas others have it way ahead. This is the issue. We don't have consistent results

[typographie]

Ugh, computing power for instance?

 

Haha, "ugh?" If it's frustrating to have to explain what your claim actually means, what does that say about the claim? 

 

What is "10-15% better computing power" in this context? Do you mean actual GPU compute performance, as in OpenCL/CUDA, or do you mean an abstract, "one has more shader cores than the other and scores higher in Passmark" sense? In my opinion, actual framerates in specific games are all that really matters to the people buying a $200-250 video card. And maybe the R9 380 is still a better choice on those grounds, but 10-15% is overstated and oversimplified.

[App4that]

Don't. Now it's my turn to give you some first-hand experience ^^ I've had one and I can tell you that the 290X is better. I used my 970 for 3-4 months and I've had one of the best 970's MSI Gaming 4G 100ME Limited Edition one

For example, Tomb Raider benchmark with the exact same PC with the exact same settings:

Average FPS on 290X is 8,5FPS higher than MAX FPS on a 970... Nothing besides the GPU and the power supply changed since then in my PC.

I wouldn't be looking to replace my 390, I just want to be able to test the two in the same system. Anyone can cherry pick a graph off of google to make themselves feel good about their purchases. You and I want to know, and to know means putting your money where your mouth is and testing both. I mostly want to know because Nvidia fans are fucking rabbit. Makes me curious.

But, like I said. I'd have to resist the urge to get a i7 next month to do it. And i7s are cool.

[Morgan MLGman]

Haha, "ugh?" If it's frustrating to have to explain what your claim actually means, what does that say about the claim? 

 

What is "10-15% better computing power" in this context? Do you mean actual GPU compute performance, as in OpenCL/CUDA, or do you mean an abstract, "one has more shader cores than the other and scores higher in Passmark" sense? In my opinion, actual framerates in specific games are all that really matters to the people buying a $200-250 video card. And maybe the R9 380 is still a better choice on those grounds, but 10-15% is overstated and oversimplified.

My overall point still stands. 960 = rubbish (not actually rubbish, but considering its place on the market, it's rubbish)

[winningsince1337]

I would never choose AMD...................................................................................................................at least until a lot of software can use OpenCL.

I do a fair amount of rendering in Vegas 13 Pro, and using a 290 to render is a huge help when it comes to time compared to just a 4770k. Vegas 13 has some pretty good Open CL support.

[App4that]

I do a fair amount of rendering in Vegas 13 Pro, and using a 290 to render is a huge help when it comes to time compared to just a 4770k. Vegas 13 has some pretty good Open CL support.

My wife is a professional graphic designer and uses a 290. It's just fanboys being fanboys that don't even walk the walk. Pay no mind.

[Mew]

I just trust DigitalFoundry a hell of a lot more since there I at least get raw footage and numbers. I also don't see how certain publications have a 380X on par with a 960 whereas others have it way ahead. This is the issue. We don't have consistent results

The issue is gpu boost and some reviewers testing stock vs stock, factory oc vs factory oc, stock vs factory oc, and different reviewers using different parts of the game and settings.

[winningsince1337]

My wife is a professional graphic designer and uses a 290. It's just fanboys being fanboys that don't even walk the walk. Pay no mind.

Yeah, I'm interested to see how cards from AMD turn out now that AMD acquired CUDA liscencing