videocardzAMD Radeon R9 290X confirmed to feature 64 ROPs

[ams23]

That is incorrect. There was never GTX 680 4GB editions at newegg for $550 when it first came out, or even at the time Titan came out. Do a search history at this forum where we compared the cheapest GTX 680 4GB version at $629-649 to Titan.

 

[Stuka87]

Deleted

I misread the quote.

 

[ozzy702]

Actually that is not necessarily correct. Due to having 4GB RAM in R9 290X instead of 3GB RAM as in GTX 780 SC, NVIDIA's RAM production cost will be lower. GTX 780 SC should be able to easily compete with R9 290X when it comes to overall perf. and perf per dollar.

AMD will be running slower (less expensive) memory modules thanks to the 512bit bus. I'd say the memory expense is probably a wash.

 

[Vesku]

I agree with this but i'd also say for the more extreme guys (like us) overclocking becomes very much a value added proposition. I'm hoping the 290X does overclock as well on air as some have hinted at, if that is the case then a 600$ 290X would definitely be a hell of a card. I've mentioned it before but the only real pricing target for the 290X are the pre-overclocked GTX 780 cards. I'd think differently if NV weren't better than AMD in terms of software, but it is what it is in that respect. AMD can't charge what NV can for the same performance because of software (IMHO).

I know I sound like a broken record at this point with regard to software, but AMD must get their software on par with nvidia's level if they want the same type of brand strength/recognition that they get. Right now AMD's software isn't quite there.

Have yet to have any issues with my 7950 except for one driver update causing a snowy pixel effect in Skyrim with AA (solved by rolling back and waiting for the next release. My GTS 450 was fine too except for annoyingly getting stuck in 2D clocks after a while without a reboot, every other driver update.

Would be nice if AMD and Nvidia could settle on a cross-platform Physics engine, though. Or have AMD push their own physics enhancements alongside Mantle.

 

[Grooveriding]

Have yet to have any issues with my 7950 except for one driver update causing a snowy pixel effect in Skyrim with AA (solved by rolling back and waiting for the next release. My GTS 450 was fine too except for annoyingly getting stuck in 2D clocks after a while without a reboot, every other driver update.

Would be nice if AMD and Nvidia could settle on a cross-platform Physics engine, though. Or have AMD push their own physics enhancements alongside Mantle.

I've been wondering the same, if AMD will pull an nvidia with Mantle and start making available exclusive visual features via Mantle using the additional performance they gain with the API.

 

[ams23]

AMD will be running slower (less expensive) memory modules thanks to the 512bit bus. I'd say the memory expense is probably a wash.

The production cost of 4GB GDDR5 RAM @ 5Gbps should be more than 3GB GDDR5 RAM @ 6Gbps in this day and age.

 

[Silverforce11]

I've been wondering the same, if AMD will pull an nvidia with Mantle and start making available exclusive visual features via Mantle using the additional performance they gain with the API.

Hope not, I really hate dev wasting time on a vendor specific fluff graphics when they could spend that time on overall gameplay improvement or even game performance optimizations as a whole.

Mantle should only be used to specifically speed up rendering. *In theory* Its available in the game engine and its all on the engine to compile so the extra work required is not major.

 

[AtenRa]

Yeah cause AT is bias now right? Geeze...

Your words not mine, I simple made an observation that AT changed the Tessellation benchmark when TITAN was released.

 

[Skurge]

Hope not, I really hate dev wasting time on a vendor specific fluff graphics when they could spend that time on overall gameplay improvement or even game performance optimizations as a whole.

Mantle should only be used to specifically speed up rendering. *In theory* Its available in the game engine and its all on the engine to compile so the extra work required is not major.

The people who make the engine are not the same people who make the gameplay. Besides, Since FB3 will be powering all their games going forward so any extra work or money spent on FB3 will not effect the other games, so I say they should add as much stuff as they want/can.

 

[yacoub]

Looks like the 290 will be a decent 7970 upgrade in a couple years for around $250. Perfect!

 

[toyota]

Yes it was. When the GTX 680 4GB was originally listed at newegg, it was somewhere around $629-649 while GTX 680 2GB was around $449-469. Anyway, this is a semantical argument because the price difference was relatively large and much more than $50 as asserted earlier.

I never asserted $50 dollar difference. and the 680 was $500 and when 4gb models came out most were only $100 difference or less. and actually I linked you to a review of a 680 4gb that was $560 while the 680 2gb reference was still $500 so stop with your silly $200 exaggerations.

 

[Vesku]

Looks like the 290 will be a decent 7970 upgrade in a couple years for around $250. Perfect!

The 20nm $200-250 cards should give you close performance but not necessarily the 512 bit memory, amount of RAM, and GPU compute of the 290. That is if Nvidia and AMD don't reduce the die size on us for that price range of card.

 

[Nintendesert]

I wouldn't know what to do with that many ROPs.

 

[ruhtraeel]

Looks like the 290 will be a decent 7970 upgrade in a couple years for around $250. Perfect!

At that point I would probably just get the 20nm card. The difference between 7xxx and R9 will be smaller than R9 and next gen. Added to the fact that it will be cheaper since they don't need to do the whole big die strategy to get performance increases

 

[SolMiester]

Metro: LL with tessellation:

Tomb Raider with tessellation:

Metro: LL is an nvidia TWIMTBP game and Tomb Raider is a Squeenix game. Both use heavy amounts of tessellation. Obviously you're not very objective in your opinions on this.

That said, this is in contrast to the situation with Cayman XT where tessellation heavy games, without exception, ran like a dog. The GTX 580 ran circles around Cayman XT in tessellation heavy games. This is not the case with Tahiti - there are some synthetic benchmarks where it pulls ahead in tess performance, and others where it does not. Where it matters, GAMES, it performs just fine and on par with the NV competing product. Your characterization of Tahiti not performing well in tessellated games has no basis in reality. Unless you cherry pick a NV skewed title - if you average all titles out, the Tahiti is on par with the 770 in terms of GAMING TESSELLATION performance. Cayman XT was not - Cayman was ridiculed for it's poor tess performance and that was justified. Tahiti basically corrected tessellation issues in hardware , and it performs just fine where tessellation matters. GAMES.

TBH, I thought Ams23 was discussing Titan DP and tess, not GK104 770.

 

[NostaSeronx]

I wouldn't know what to do with that many ROPs.

More antialiasing/downsampling?

 

[Rvenger]

TBH, I thought Ams23 was discussing Titan DP and tess, not GK104 770.

Well then you are comparing an apple to an orange because 7970 is the competitor to the 680.

 

[Bubbleawsome]

Someone needs to get a farm built on these for F@H! They look amazing. I wish I had money.

 

[Mastercommander]

Oh, you mean the one where AMD always uses "application preference" and obeys the application mediated tessellation settings by default. Is that what you're referring to?

You can change it in the control panel if you want, but by default application settings are always used. And i'm sure HardOCP did not manipulate tessellation settings whatsoever.

Actually, it defaults to AMD optimized in 13.10 and 13.11

If HardOCP "manipulated" tessellation settings, it would be to turn AMD cheating on or off

 

[blackened23]

Okay, I remember the babble about this ages ago from what I owned my 7970s. To be clear: that setting does not manipulate tessellation to lower the effect or remove it. That is what Ibra implied, he implied the default setting is a "cheat", which is most certainly not the case. This is what a lot of benchmark freaks do to get world records, they go in CCC and turn the slider all the way down to remove tessellation vertices - this requires *manual intervention*. The default setting in CCC is not a cheat. It does not remove tessellation vertices at all, it obeys the application setting in 3d apps. I think you're slightly confused as to what it does.

The only way to remove tessellation vertices is manually turning the "slider" down. By default, AMD does not manipulate tessellation settings. If they were, don't you think review websites would call them out on it? Yes, yes they would - the fact is, tessellation is not manipulated or removed with AMD drivers. (unless you intentionally go in and tell it to). It without exception, by default, obeys the application setting and is not a cheat.

What many have complained about in the past is that AMD users can basically remove tessellation completely from 3d benchmarks by manually turning the slider down, and indeed this is what world record benchmark nuts do and have done for years. But that is only if you turn the slider down. By default, as mentioned, the 3d application setting is ALWAYS obeyed and tessellation is not manipulated. If you want to "cheat" and remove tessellation you must turn the slider down. No reputable website would do this. HardOCP aren't idiots. Brent @ HardOCP knows his stuff and he knows how to benchmark properly, he would not manually turn the slider down to enable a cheat (again, that is what Ibra implied). Neither are the numerous other sites that do real world benchmarking on AMD / NV hardware - the tessellation settings are always apples to apples between NV and AMD.

 

[SolMiester]

Well then you are comparing an apple to an orange because 7970 is the competitor to the 680.

The discussion was the price of Titan and that not all owners used it just for gaming, as it had DP and better tess etc.....then we got a graph of GK104 as if to prove the statement wrong!?...IDK, maybe read it wrong?!

 

[blackened23]

The discussion was the price of Titan and that not all owners used it just for gaming, as it had DP and better tess etc.....then we got a graph of GK104 as if to prove the statement wrong!?...IDK, maybe read it wrong?!

Maybe I read it wrong as well, but there was an implication along the way that AMD's architecture is "weak" at tessellation. Aside from one OpenGL synthetic, that is not the case - real world usage shows otherwise. I think the 280X vs 770 is a good comparison here as to show real world usage of tessellation in games - it is true that the Cayman uarch was very weak in tessellation, and the GTX 580 ran circles around it in every game that used tessellation. Tahiti corrected those issues and performs just fine in tessellation heavy games or other benchmarks such as Unigine, Firestrike , among many others. Generally speaking, you may find a cherry picked benchmark here or there which may favor one architecture (OpenGL favors NV heavily, or rather - AMD's driver is focused on DX11 since there has been one OpenGL game of note in the past 3 years), but averaged out among all tessellation heavy games, Tahiti 280X and GK104 GTX 770 are on par with each other in terms of tess performance. Therefore there is no tessellation performance deficit with AMD's GCN as there was with the Cayman XT.

I guess the appropriate comparison for the GK110 would be the 290X which obviously isn't out yet. Based on Tahiti, though, i'd expect the 290X to do just fine in tessellation heavy games. I certainly don't expect a "Cayman XT" situation where any and all games using tessellation perform horribly - that wasn't the case with Tahiti and i'm about 99% sure it won't be the case with the 290X.

 

[Elfear]

That is incorrect. There was never GTX 680 4GB editions at newegg for $550 when it first came out, or even at the time Titan came out. Do a search history at this forum where we compared the cheapest GTX 680 4GB version at $629-649 to Titan.

"The Gigabyte GeForce GTX 680 4GB costs 17% more than the 2GB version, which equates to an $80 price premium." Source


Back OT - Did not expect Hawaii to have 64 ROPs after all the rumors of 44. As someone else stated I am shocked AMD was able to fit so much onto a 438mm2 die.

 

[Cookie Monster]

Im still somewhat puzzled by the low 320GB/s bandwidth on the R290.. especially since they are pushing for 4K gaming.

 

[tolis626]

Im still somewhat puzzled by the low 320GB/s bandwidth on the R290.. especially since they are pushing for 4K gaming.

Define low.All high end cards until now (7970,GTX780,Titan) have below 300Gb/s of bandwidth,if I'm not mistaken.And I don't believe that they were bandwidth starved when running at very high resolutions in the first place.Plus,if we're lucky and the 290x's memory overclocks well,that 512bit bus is gonna yield closer to 400GB/s of bandwidth at the same clock speeds as the 384bit GPUs.