R260/270/280/290/290x Review thread

[Borealis7]

TPU:

http://www.techpowerup.com/reviews/AMD/R7_260X/
http://www.techpowerup.com/reviews/MSI/R9_280X_Gaming/
http://www.techpowerup.com/reviews/AMD/R9_270X/

the 270X @ 200$ looks good in my opinion. 260X is weak-sauce, and the 280X is sure to make waves in the pricing of other high end cards.

 

[Silverforce11]

The 7950 @ $200 is a good buy, not the 270X, when OC vs OC, Tahiti pwns Pitcarn.

Can't wait for the 290/X review, AMD's website listed the 290 spec as having identical shader counts as the 290X.. really if that is true, it would be one heck of a card.

 

[3DVagabond]

The 7950 @ $200 is a good buy, not the 270X, when OC vs OC, Tahiti pwns Pitcarn.

Can't wait for the 290/X review, AMD's website listed the 290 spec as having identical shader counts as the 290X.. really if that is true, it would be one heck of a card.

Link? I haven't seen the listing for the 290.

 

[Gikaseixas]

The 7950 @ $200 is a good buy, not the 270X

Nvidia just lowered GTX660 and GTX660ti because of the 270X, not a bad card at all. The 7950 is just an amazing one...

 

[Silverforce11]

Link? I haven't seen the listing for the 290.

http://www.amd.com/US/PRODUCTS/DESKTOP/GRAPHICS/R9/Pages/amd-radeon-hd-r9-series.aspx#4

Looks unfinished since it doesn't have the non X version listed, but the columns next to the 290X has what looks to be a cut down variant, ie. the 290.

Its listed as the 280X but we know those specs aren't for Tahiti v2.

 

[hyrule4927]

http://www.amd.com/US/PRODUCTS/DESKTOP/GRAPHICS/R9/Pages/amd-radeon-hd-r9-series.aspx#4

Looks unfinished since it doesn't have the non X version listed, but the columns next to the 290X has what looks to be a cut down variant, ie. the 290.

Its listed as the 280X but we know those specs aren't for Tahiti v2.

Well, looking at the clockspeed, VRAM, CF bridge requirement, and lack of TrueAudio, seems like that column is the 280X and SPU count was just a typo.

 

[raghu78]

Unacceptable, they will use the same chips(Tahiti and Pitcairn) for almost 4 years (early 2011 until the end of 2014).

sorry you got it wrong. HD 7970 launched on Jan 9, 2012 and HD 7870 on Mar 5, 2012. so even if these chips are going to be here till Sep 2014 that would mean around 33 months or below 3 years. the problem here is TSMC 20nm node is not ready for volume production and should only be ready in H2 2014.

After two years i would expect 290X to replace the 7970GE at $299 and a new even bigger die to take the high end place at $549+.

NVIDIA can release a new GTX785 with the same specs as current TITAN at $699 and then even release a Full die specs TITAN 2 at the same $999.

are you serious ? the R9 280X at USD 299 is a commendable product and a good value. You realize Nvidia is selling GTX 770 at USD 400 for the same perf. also without waiting for R9 290X reviews you are saying Nvidia can release a GTX 785 with same specs as Titan at USD 699 and a Titan 2 for USD 999.

 

[ShintaiDK]

I was thinking that from the outset as well, but the 280X is actually clocked higher. So it is "very very" slightly better than the 7970 GE.

That said, the performance jump over the 7970GE is minuscule to non-existent since it was only a 20mhz core and 400mhz memory jump. Not very impressive. Not the same jump as the 680 > 770 - the performance jump there (and the addition of GPU Boost 2.0) were much more impressive. Then again, the 280X is 300$, not a bad deal.

HD7970GE is faster than regular R9 280X as you can see in Anandtech. So lets not sugarcoat it anymore. Less for more is what AMD wanted.

 

[ShintaiDK]

Unacceptable, they will use the same chips(Tahiti and Pitcairn) for almost 4 years (early 2011 until the end of 2014).
They already made two new chips(260 and 290), they had to make another two and have the same features for the entire family.

You mean early 2012 (Paper launch on the 22nd december 2011 doesnt count.) to end of 2014, if not early 2015. Welcome to 3 (or even 4) year cycles and it will only get worse. Specially when we get 20nm non finfet before 20^H^H16nm with finfet 1½-2 years later.

dGPUs are really struggling now.

 

[Imouto]

Boring.

Now I want to know how are they going to fit the 290X in that price scheme. If we take the 270X as the best bang for the buck they're asking for a 50% price increase for ~35% perf increase with the 280X.

Are they going to ask for a 100% price increase for the 290X over the 280X with barely the same perf improvement?

Are they going to release a 290 that would make seem the 290X even more overpriced at a $600 price tag? I know they're aiming for the GTX 780 and Titan but they have to look at their lineup too and it would make zero sense with such pricing.

Good times for 1080p gaming. Not so much for those looking for top tier GPUs.

 

[SPBHM]

one question, will the 7790 get the audio thing enabled by driver updates?


comparing all the cards one thing is clear, 7950 is a better option than all the 270 and 280 cards.

and the 7850 is a better choice compared to the 260x, so yes... I guess AMD will be able to sell the pile of cards with the old name.

 

[pandemonium]

I still think the R280X will be a valuable purchase, given it's comparable performance to the 7970GHz edition, at a slightly cheaper price, with additional features enabled. Sure, it's a rebadge, but still a win for the consumer.

At this point, I really fail to see how anyone can justify spending the extra money on a GTX770 or 760 for gaming purposes. (You can leave the typical CUDA and PhysX arguments out of this.)

Edit: Also, TR's shortbread has reviews in no small amount.

Edit #2: The ASUS CoolTech fan really has me impressed. [To me] This helps explain the tremendously quiet operation and low load temperatures along with the larger heatpipe diameters.

 

[ShintaiDK]

Hardwarecanucks on overclocking:

So, with AMD’s Catalyst Control center in tow, we eventually achieved a core speed of 1108MHz and a respectable GDDR5 frequency of 6.454GHz. Whether or not these results were voltage limited is anyone’s guess but percentage-wise they’re among the lowest we’ve ever achieved with a Tahiti-based product.

I wonder if the regular 280X was simply all the speedbin failed Tahiti chips...

 

[AtenRa]

sorry you got it wrong. HD 7970 launched on Jan 9, 2012 and HD 7870 on Mar 5, 2012. so even if these chips are going to be here till Sep 2014 that would mean around 33 months or below 3 years. the problem here is TSMC 20nm node is not ready for volume production and should only be ready in H2 2014.

Yea my mistake, it will be a 3 year cycle for the same chip.

are you serious ? the R9 280X at USD 299 is a commendable product and a good value. You realize Nvidia is selling GTX 770 at USD 400 for the same perf. also without waiting for R9 290X reviews you are saying Nvidia can release a GTX 785 with same specs as Titan at USD 699 and a Titan 2 for USD 999.

280X at $299 is a good value because GTX770 is at $399. The problem is that GPU prices went up the last two years by a high margin. Middle sized chips like GTX680/770 and HD7950/70 were used to be at $299-350 in 2011.
I understood the need for higher price at the original launch of 28nm GPUs from both AMD and NV in early 2012. But today after two years, 28nm production cost should have been at the same level 40nm was at the same period.
So, it is only logical to expect the same price we had in 2011 for the same size of silicon in 2013. But my biggest problem is that 280/X and 270/X bring nothing new to the table except a price reduction to the price level it had to be from the start. It would be better if 270/280 had the same features of 260/290 and bigger performance up to 20% than Pitcairn/Tahiti for the same price of $299 or 349. And make 290/X an even bigger die to bring even higher performance of Last gen (HD7970).
At 40nm I upgraded from HD5850 to HD6950 because it brought a new architecture, better Tessellation and higher performance. Now I don’t have any reason to upgrade from HD7950 to R9 280. But if I would get the same features as 290 and another 10-20% performance I would make the move.
Well, things change and GPU cycles will be a 3 year from now on. It will be the first time in years I will not do an intermediate upgrade between nodes.
As for Titan, I just gave you NV’s options at the matter and what a bigger die can do if your opponent only has a smaller die.

 

[GaiaHunter]

Hardwarecanucks on overclocking:



I wonder if the regular 280X was simply all the speedbin failed Tahiti chips...

Considering that the hardwareheaven R9 280X hit 1250/1650, I doubt it.

We found our 280X DirectCU 2 Top hit a maximum overclock of 1250/1650...

 

[ShintaiDK]

Considering that the hardwareheaven R9 280X hit 1250/1650, I doubt it.

Thats already a premium factory OCed card that cost extra. It better be better.

 

[3DVagabond]

Considering that the hardwareheaven R9 280X hit 1250/1650, I doubt it.

I can't believe you bothered to indulge that post with an answer. Good on you for taking the time, though. :thumbsup:

 

[AtenRa]

1100MHz is only with default vcore, ASUS CUII is Vcore unlocked and brings higher voltage at default settings.

280/X will OC at the same levels as HD7950/70

 

[Haserath]

Yea my mistake, it will be a 3 year cycle for the same chip.



280X at $299 is a good value because GTX770 is at $399. The problem is that GPU prices went up the last two years by a high margin. Middle sized chips like GTX680/770 and HD7950/70 were used to be at $299-350 in 2011.
I understood the need for higher price at the original launch of 28nm GPUs from both AMD and NV in early 2012. But today after two years, 28nm production cost should have been at the same level 40nm was at the same period.
So, it is only logical to expect the same price we had in 2011 for the same size of silicon in 2013. But my biggest problem is that 280/X and 270/X bring nothing new to the table except a price reduction to the price level it had to be from the start. It would be better if 270/280 had the same features of 260/290 and bigger performance up to 20% than Pitcairn/Tahiti for the same price or $299 or 349. And make 290/X an even bigger die to bring even higher performance of Last gen (HD7970).
At 40nm I upgraded from HD5750 to HD6950 because it brought a new architecture, better Tessellation and higher performance. Now I don’t have any reason to upgrade from HD7950 to R9 280. But if I would get the same features as 290 and another 10-20% performance I would make the move.
Well, things change and GPU cycles will be a 3 year from now on. It will be the first time in years I will not do an intermediate upgrade between nodes.

If we are to believe Nvidia's price per transistor:

28nm GPUs are still more expensive to make than the 40nm ones. The transistor increase is more than price per transistor decrease.

Then they have to account for longer product cycles as node lengths increase.

 

[Imouto]

Thats already a premium factory OCed card that cost extra. It better be better.

It's $10 more and it's already a way bigger cooler. You serious?

 

[Dribble]

Was quite positive at first then basically worked out it's the same cards at the same prices you can currently buy them for only someone removed the 7950 - so it's a smaller selection of the same cards at the same prices minus the previous high end price/performance king.

 

[tg2708]

Even with all this talk about value and the like I am still considering an nvidia card (I do have a 7970). For some reason I think nvidia just offers a better overall experience and this is where the "what if" moment kicks. But the only way I'd know is to compare both cards myself as reviews and the like are not indicative of the actual performance increase I might gain without doing my own testing. Also nvidia has so much more proprietary enhancements that adds value to their products. If I see a price drop for the 770 I'm going to pick one up.

 

[AtenRa]

If we are to believe Nvidia's price per transistor:

28nm GPUs are still more expensive to make than the 40nm ones. The transistor increase is more than price per transistor decrease.

Then they have to account for longer product cycles as node lengths increase.

The graph shows 28nm to have lower price than 40nm. It simple saying that 20 and 14nm will not have lower prices than 28nm.
Also, longer cycles help lower the prices. Just take Tahiti (HD7970) die for example, it will be in production for 3 years instead of 12-14 months. Same for NVs GK104 and 110. After two years in production, you expect to have made all the money you spend on RD and more. Now you pocket the extra you made and you can lower the prices because Manufacturing cost is also lower.
What used to be a 12-14 months return of investment now becomes more than 24 months periods. That also brings RD cost lower or you can spend more for a longer time. If i remember correctly, NV spend 2Bil for the Kepler family. They clearly knew they are going to need more than 1-2 years period to have a RoI.

 

[GaiaHunter]

Thats already a premium factory OCed card that cost extra. It better be better.

And is the XFX DD a reference card?

We all know that AIBs generally save some of the best bins for their premium cards.

 

[Abwx]

The graph shows 28nm to have lower price than 40nm. It simple saying that 20 and 14nm will not have lower prices than 28nm.
Also, longer cycles help lower the prices. Just take Tahiti (HD7970) die for example, it will be in production for 3 years instead of 12-14 months.

The same graph says that Tahiti was launched when cost/die
was 50% higher than it is currently yet we see here people
complaining about a high launching price...