AMD Radeon HD 9970 Specifications Leaked – Twice as fast as GTX 780 (ChipLoco rumor)

[BoFox]

GK110 being "nearly" 30% bigger (actually less) means that Hawaii will be over 423.8 mm^2, no less.

This DOES make it the biggest GPU silicon AMD (or ATI) has ever made, beating HD 2900XT at 420mm^2.

This also means that it will be at least 20% bigger than Tahiti's 352mm^2.

It seems that:

15-25% more shaders/TMUs
and
48 ROPs
(or even more shaders/TMUs while still having only 32 ROPs which might hurt @ higher resolutions with AA)

is just about what AMD can do with this increase in silicon die size, without some other magical optimizations or expansions (remember, HD 4890 was actually LARGER than HD 4870 despite having practically identical GPU layout).

It certainly does seem like 30% increase in performance is not out of reach, as long as AMD is willing to allow a 300W TDP headroom on this single-GPU card.

However, I'd expect a modest TDP ceiling, and a 20% increase out of the box rather than 30%... because going anywhere close to 300W is just pushing it too hard. There are too many risks (longevity, warranty, noise, notoriety of heat output, etc..) for this to become a practically viable option.

25% increase at 265W might, just might be possible. AMD did it really nicely with Bonaire, after all, so I think they can do it (hint, that is, beat GTX 780).

.. I'm just hoping that AMD will no longer stubbornly stick to 32 ROPs, and just move onto a symmetrical ratio of 48 ROPs to 384-bit bus (so that there's full access to the RBEs rather than limited crossbar access). That's why Tahiti's 384-bit bus seems rather lackluster, like as if the GPU had less than 320-bit bus access in the real world. Just guessing - there seems to be about 5% overall performance overhead that could be regained due to having a straight 48 ROP design alone. Yet it would probably also translate to 5% increase in silicon die size as ROPs are notoriously difficult to shrink with each new fab process nodes, from what I have been told. After all, it was a logical move by AMD to do that with Tahiti early on 28nm process in order to get it out ASAP while keeping the size and power consumption down. I have no idea on exactly how much the ROPs affect power consumption though (up to 5 percent for this 50% increase in ROPs??).. but if AMD manages all of this well (48 ROPs, +25% performance, say 15+% reduction in power consumption-to-performance ratio, etc..) then I'd be amazed! Or more!!!

 

[3DVagabond]

You are getting pretty hung up on the SLI comments. I said that SLI isn't necessarily better. I did not say it is never better, or usually not better, just that it isn't always better.

Read the whole post, and consider it all as a whole. The point is, SLI is not perfect, even if it is generally pretty good. If you jump on board with SLI, you have to make concessions on some inconsistencies. I gave an example, one which I cannot find the source of any longer, but it was one of the video reviews during the early stages of FCAT testing.

Tell me, so we can end this. Do you want me to say SLI is perfect, or horrible and not worth ever using? I think it is in a gray area, personally.

EDIT: Here is an example of where it isn't perfect, in particular the surround example: http://www.pcper.com/reviews/Graphi...0-6GB-Review-Malta-Gets-Frame-Rated/Far-Cry-3

You may also find that games that aren't AAA titles, tend to have more issues, but are not tested, due to them not being so popular.

Nothing's perfect. So, no, I'm not saying that or that it's horrible. I just think that 760SLI overall will give better performance than a single 780. Probably somewhere approaching ~90% of the time, and it's cheaper? The 760 is one of nVidia's better perf/$ gpu's, better than the 780. I think 2 of them are still better value than a single 780. I would recommend them over a single 780 for most users. That's all. The way you posted that SLI was too problematic to consider it with 760's but fine with 780's just didn't make sense to me and I thought you could clarify.

Far Cry 3 is just a buggy game in general.

 

[bystander36]

Nothing's perfect. So, no, I'm not saying that or that it's horrible. I just think that 760SLI overall will give better performance than a single 780. Probably somewhere approaching ~90% of the time, and it's cheaper? The 760 is one of nVidia's better perf/$ gpu's, better than the 780. I think 2 of them are still better value than a single 780. I would recommend them over a single 780 for most users. That's all. The way you posted that SLI was too problematic to consider it with 760's but fine with 780's just didn't make sense to me and I thought you could clarify.

Far Cry 3 is just a buggy game in general.

The 780 is a pretty expensive choice, so ya, it is not a bad way to go with 760 SLI, but that is not to say that many people wouldn't still prefer the 780.

 

[3DVagabond]

The 780 is a pretty expensive choice, so ya, it is not a bad way to go with 760 SLI, but that is not to say that many people wouldn't still prefer the 780.

True. Some people won't even consider multi-GPU.

 

[KevinH]

I didn't read the full thread but are #'s out? Is that why people feel let down?

 

[hyrule4927]

All they need to do is put a gazillion gigs of vram on this thing to make the morons on the forums happy. Doesn't seem to matter to a lot of people on forums how fast the gpu is anymore, but how much vram it has. Looks like I'll be getting some sli GTX 760's now while I wait out the true next gen. :whiste:

If you are talking about gaming forums, then yes, most people there have no idea what they are talking about and think VRAM is the sole determiner of performance.

Don't think your statement applies to most tech forums though . . .

Also, did anyone else see this: http://www.forbes.com/sites/jasonev...adeon-cards-next-gen-consoles-7990-criticism/

 

[3DVagabond]

I didn't read the full thread but are #'s out? Is that why people feel let down?

If you read the article linked below, it might help explain.

“They’re coming in Q4. I can’t reveal a pricepoint but we’re looking at more traditional enthusiast GPU pricepoints. We’re not targeting a $999 single GPU solution like our competition because we believe not a lot of people have that $999. We normally address what we call the ultra-enthusiast segment with a dual-GPU offering like the 7990. So this next-generation line is targeting more of the enthusiast market versus the ultra-enthusiast one.

“It’s also extremely efficient. [Nvidia's Kepler] GK110 is nearly 30% bigger from a die size point of view. We believe we have the best performance for the die size for the enthusiast GPU.

Here's a snip from it. A few things to notice here.
1, It's not targeting the $999.99 ultra enthusiast market. Read between the lines and it's likely saying it's not as fast as GK110. It could just be a backhanded way to jab at nVidia's pricing, but lately AMD has been a bit more in your face with these types of comments. If Hawaii was going to score a clean kill over the GK110 I think AMD would just say so and then brag on the pricing.
2, They say it's ~30% smaller. This also could be taken to say it can't compete with GK110. 30% die size advantage is hard to make up. Especially considering that Kepler and the GK110 aren't dogs in the perf/mm² department.
3,They go on and say they have the best perf/mm² for the enthusiast GPU. This gives a bit of hope for high performance.
4, The reference to a dual GPU card is good. It means they might have actually lowered the TDP compared to Tahiti. Mainly I say that because the 7990 is stretched a bit too much already when you consider that when installed in a typical case it throttles from heat. Even with that giant Steam Punk cooler.

If you are talking about gaming forums, then yes, most people there have no idea what they are talking about and think VRAM is the sole determiner of performance.

Don't think your statement applies to most tech forums though . . .

Also, did anyone else see this: http://www.forbes.com/sites/jasonev...adeon-cards-next-gen-consoles-7990-criticism/

I'd seen the part I quoted above, but not the whole article. Thanks for posting this. :thumbsup:

 

[sontin]

I thought gk110 actually was somewhat bad when it came to gaming perf/mm2 which kind of left the door open for AMD

Perf/mm^2 is irrelevant. Power consumption is the limitation. To stay under 250W you need a bigger die and lower clocks.

Perf/mm^2 is great if you really have a low power and high performance chip.

 

[raghu78]

1, It's not targeting the $999.99 ultra enthusiast market. Read between the lines and it's likely saying it's not as fast as GK110. It could just be a backhanded way to jab at nVidia's pricing, but lately AMD has been a bit more in your face with these types of comments. If Hawaii was going to score a clean kill over the GK110 I think AMD would just say so and then brag on the pricing.
2, They say it's ~30% smaller. This also could be taken to say it can't compete with GK110. 30% die size advantage is hard to make up. Especially considering that Kepler and the GK110 aren't dogs in the perf/mm² department.
3,They go on and say they have the best perf/mm² for the enthusiast GPU. This gives a bit of hope for high performance.
4, The reference to a dual GPU card is good. It means they might have actually lowered the TDP compared to Tahiti. Mainly I say that because the 7990 is stretched a bit too much already when you consider that when installed in a typical case it throttles from heat. Even with that giant Steam Punk cooler.

I'd seen the part I quoted above, but not the whole article. Thanks for posting this. :thumbsup:

GK104 is the stand out in terms of perf /sq mm but GK110 is actually poor in terms of perf / sq mm. Its roughly 90% larger than GK104 and only 30 - 35% faster on a clock for clock basis. Tahiti also is poor in terms of perf /sq mm.

Hawaii looks to achieve two goals - better perf / sq mm than GK110 / Tahiti and competitive perf/watt with GTX 780/Titan. let see what AMD has achieved when actual reviews are out.

 

[Erenhardt]

GK104 is the stand out in terms of perf /sq mm but GK110 is actually poor in terms of perf / sq mm. Its roughly 90% larger than GK104 and only 30 - 35% faster on a clock for clock basis. Tahiti also is poor in terms of perf /sq mm.

Hawaii looks to achieve two goals - better perf / sq mm than GK110 / Tahiti and competitive perf/watt with GTX 780/Titan. let see what AMD has achieved when actual reviews are out.

Yes. I don't know you people extrapolate hawaii performance based on thaiti's performance/mm2 ratio. These are two different architectures, you can't directly compare those.

Die increase alone will be resposible for a nice performance boost. Can't wait to see how will that scale with arch improvements!

 

[3DVagabond]

Yes. I don't know you people extrapolate hawaii performance based on thaiti's performance/mm2 ratio. These are two different architectures, you can't directly compare those.

Die increase alone will be resposible for a nice performance boost. Can't wait to see how will that scale with arch improvements!

They aren't all that different. Still GCN.

 

[insertcarehere]

GK104 is the stand out in terms of perf /sq mm but GK110 is actually poor in terms of perf / sq mm. Its roughly 90% larger than GK104 and only 30 - 35% faster on a clock for clock basis. Tahiti also is poor in terms of perf /sq mm.

Considering that GK110 has significant chunks of the die sealed off in both titan and gtx780, using these as basis points isn't exactly that relevant.

 

[DownTheSky]

You're making a highly optimistic performance estimate and call this very estimate of yours disappointing? What's wrong with your logic unit?

It is disappointing because I hoped for AMD to aim for something bigger and better. To actually go for n1 spot and all the perks that come with it, not just be on par with the competition (like they traditionally like to aim for).

 

[raghu78]

Considering that GK110 has significant chunks of the die sealed off in both titan and gtx780, using these as basis points isn't exactly that relevant.

Titan has only 1 SMX disabled. the rest of the chip is fully working. still Titan to GTX 770 is atmost 35 - 40% perf diff at same clocks. you can see how GTX 780 performs only around 25 - 30% faster clock for clock. add another 10% max for Titan.

http://www.hardocp.com/article/2013/08/14/galaxy_geforce_gtx_780_hof_edition_review/3

http://www.hardocp.com/article/2013..._directcu_ii_video_card_review/4#.Ujgr13_3x8E

Metro 2560 x 1600 Very High settings AAA 16x AF

GTX 780(1.3 Ghz) - 65
GTX 770(1.3 ghz) - 50.1

http://www.hardocp.com/article/2013/08/14/galaxy_geforce_gtx_780_hof_edition_review/6

http://www.hardocp.com/article/2013..._directcu_ii_video_card_review/7#.UjgsV3_3x8E

Farcry 2560 x 1600 Ultra 2x MSAA

GTX 780(1.3 Ghz) - 50.2
GTX 770(1.3 ghz) - 39.4

http://www.hardwarecanucks.com/foru...s/63051-evga-gtx-780-classified-review-8.html

http://www.hardwarecanucks.com/foru...ews/62594-galaxy-gtx-770-gc-4gb-review-8.html

Crysis 3 2560 x 1600 Very high FXAA

GTX 780(1.32 Ghz) - 47.64
GTX 770(1.31 Ghz) - 37.99

the GTX 780 is 25 - 30% faster than GTX 770 at same clocks. even with Titan that gap is 35 - 40%. the die size difference is 90%. the Tesla / Quadro centric design takes a lot of die space for features which benefit HPC , but not gaming.

 

[boxleitnerb]

Are you still touting bandwidth-limited benchmarks under extreme OC? You are funny.

 

[Keysplayr]

Are you still touting bandwidth-limited benchmarks under extreme OC? You are funny.

Agreed. Raghu, come back with some stock numbers. Even if the performance difference is exactly the same, it will be more conclusive rather than overclocking with bandwidth limits.

 

[raghu78]

Are you still touting bandwidth-limited benchmarks under extreme OC? You are funny.

read the hardwarecanucks review

http://www.hardwarecanucks.com/foru...s/63051-evga-gtx-780-classified-review-8.html

By moving to a core speed of 1.32GHz, we were able to all but eliminate the ultra high temperatures and ensure a continual clock speed, regardless of the situation. Temperatures normalized as well. After 20 minutes of constant load (see the image above), the core frequency didn’t budge one iota, though we were still hitting the Classified’s Voltage Limit.

"Did we forget to mention memory? Well, that topped out at 7426MHz before the GDDR5’s error correction stepped in, which is nothing short of incredible."

http://www.hardwarecanucks.com/foru...ews/62594-galaxy-gtx-770-gc-4gb-review-8.html

"Unfortunately, Galaxy doesn’t seem to offer their own home-grown overclocking software anymore so MSI’s AfterBurner was used for these tests. With it, we were able to hit a constant Boost clock of 1313MHz which is within spitting distance of the results achieved with samples from MSI and Gigabyte. Memory speeds weren’t quite as flexible with frequencies topping out at 7404MHz before error correction kicked in. "

so you see GTX 780 was running 1.32 Ghz core and 7.4 Ghz memory on a 384 bit memory bus with 48 ROPS while GTX 770 was running 1.31 Ghz core and 7.4 Ghz on a 256 bit memory with 32 ROPs.

same core clocks and memory clocks on both chips. but 50% more cores, 50% more bandwidth, 50% more ROPs with GTX 780. but only 25 - 26% more performance in Crysis 3. so stop giving excuses.

the reason is GTX 770 has 4 GPC, 4 raster engines for 1536 cc while GTX 780 has the same 4 GPC, 4 raster engines for 2304 cc. the perf/cc on GTX 780 is lower than GTX 770 due to the same amount of front end resources feeding 50% more cores. The GK104 is a extremely well balanced chip. :biggrin:

 

[Keysplayr]

read the hardwarecanucks review

http://www.hardwarecanucks.com/foru...s/63051-evga-gtx-780-classified-review-8.html

By moving to a core speed of 1.32GHz, we were able to all but eliminate the ultra high temperatures and ensure a continual clock speed, regardless of the situation. Temperatures normalized as well. After 20 minutes of constant load (see the image above), the core frequency didn’t budge one iota, though we were still hitting the Classified’s Voltage Limit.

"Did we forget to mention memory? Well, that topped out at 7426MHz before the GDDR5’s error correction stepped in, which is nothing short of incredible."

http://www.hardwarecanucks.com/foru...ews/62594-galaxy-gtx-770-gc-4gb-review-8.html

"Unfortunately, Galaxy doesn’t seem to offer their own home-grown overclocking software anymore so MSI’s AfterBurner was used for these tests. With it, we were able to hit a constant Boost clock of 1313MHz which is within spitting distance of the results achieved with samples from MSI and Gigabyte. Memory speeds weren’t quite as flexible with frequencies topping out at 7404MHz before error correction kicked in. "

so you see GTX 780 was running 1.32 Ghz core and 7.4 Ghz memory on a 384 bit memory bus with 48 ROPS while GTX 770 was running 1.31 Ghz core and 7.4 Ghz on a 256 bit memory with 32 ROPs.

same core clocks and memory clocks on both chips. but 50% more cores, 50% more bandwidth, 50% more ROPs with GTX 780. but only 25 - 26% more performance in Crysis 3. so stop giving excuses.

the reason is GTX 770 has 4 GPC, 4 raster engines for 1536 cc while GTX 780 has the same 4 GPC, 4 raster engines for 2304 cc. the perf/cc on GTX 780 is lower than GTX 770 due to the same amount of front end resources feeding 50% more cores. The GK104 is a extremely well balanced chip. :biggrin:

Stock clocks please.

 

[raghu78]

Stock clocks please.

what do you mean. GTX 780 is 25 - 30% faster than GTX 770 at the same clocks. at stock clocks the gap is around 17 - 20%.

http://www.hardwarecanucks.com/foru...369-nvidia-geforce-gtx-770-2gb-review-13.html

http://www.computerbase.de/artikel/grafikkarten/2013/nvidia-geforce-gtx-770-im-test/4/

http://www.hardware.fr/articles/896-22/recapitulatif-performances.html

 

[tweakboy]

So if a 780 gives you 40fps in a certain situation,, your saying this new ATI would be twice as fast,, so 80fps ??? hmmmm something smells fishy, thx gl

 

[RussianSensation]

What I hope you can understand is that performance has many layers and people are going to have a different opinion on what is best for them. Just because they choose a 780 doesn't make them a moron, it just means they don't want the hassle of multiple GPU and have top end performance. If they want more performance, they may opt for 2 or more.

I am not saying people who buy 780s are morons. My point is people on our forum have this idea that unless AMD delivers a card for X and it beats 780/Titan, it's automatically underwhelming. dGPUs in the $650-1000 bracket comprise a very niche market segment, even more so than $499-549 cards. AMD doesn't necessarily need to beat 780/Titan to offer a viable choice. As you yourself stated, and with which I agree, there will be people who just want the best performance. Those gamers will spend $700-1000 on EVGA Classy 780s/HOFs and Titans of this world. There will be another group of people whose budget might be $400 but if they see a card that falls within 90% of the performance of the after-market 780s and costs $550+BF4, they might just stretch their budget a little bit but no way to $700.

What amazes me is people on our forums first complained that AMD couldn't compete on performance during 4870/5870/6970 eras and yet AMD offered unbeatable price/performance. Yet now that AMD is releasing cards at $550 starting with 7970, they are not happy with that either. And that's exactly my point about loyal NV buyers - they are 90% committed to NV anyway. Price/performance, beating NV to market, game bundles, none of these factors sway them. They don't care about price/performance (proof is 4850/4870 vs. 9800GTX/260, 5870 vs. 480, 6950 unlocked vs. 570/580, etc.). They also would still buy a more expensive slower card (proof is 680 vs. 7970GE, 1Ghz 7970 vs. 760s).

Between all the NV vs. AMD bickering, what really happened is a mid-range GPU like 560Ti went up from $249 to $499 in the form of 680/7970 and now everyone is pretending as if $650-1000 are the new normal for single GPU flagship prices. NV & AMD are the winners, not us. Unfortunately next gen there is no more bitcoin mining which means we'll probably be stuck paying $650-1000 for flagship Maxwell. I guess I could swallow that if there were mind blowing next gen PC games, but there are none.

The other thing for someone who got free 7970s from mining, a single 780/R9 280 is not an upgrade. Now you are looking at spending $1100-$1,400 for 2 of those 28nm 780/R9 280 cards for a truly tangible performance increase but there are a total of 0 next generation PC games out now. I just can't get excited for 780 or R9 series right now. They feel more like a move from iPhone 5 to 5S but what I really want is an iPhone 6 if you get the analogy. There needs to be a game like Far Cry 1, Crysis 1 to make me want to spend $1-1.4K on GPUs from now on. There is just nothing like that right now and in the foreseeable future. Metro LL, Tomb Raider, while pretty games aren't revolutionary enough. BF4 won't be it either. 780 and R9 280 feel like stop gaps between next gen PC games and next gen PC GPUs. Yes they are faster by 30-35% vs. 7970GE but when next gen PC games launch, that performance increase is a drop in the bucket. For people who can throw $1-1.4K and want every setting @ max and not care that in 12 months there will be 20nm GPUs I guess it doesn't matter

 

[imaheadcase]

“It’s also extremely efficient. [Nvidia's Kepler] GK110 is nearly 30% bigger from a die size point of view. We believe we have the best performance for the die size for the enthusiast GPU." from Forbes article.

Translation: We believe with what we got its a good performer, just not as good as Nvidia because they got a bigger die.

 

[ams23]

@ raghu78, what you are saying is highly misleading because a 1300MHz core clock frequency for GTX 780 is not even close to the sweetspot for perf. per MHz for that card. In fact, in the same Hardwarecanucks review you linked to, you can see that a >30% overclock for GTX 780 Classified gives <10% improvement in Crysis 3 framerate! If you compared perf. per MHz at 900MHz core clock frequencies, the result would be much more favorable to GTX 780 (relatively speaking).

 

[Jaydip]

I am not saying people who buy 780s are morons. My point is people on our forum have this idea that unless AMD delivers a card for X and it beats 780/Titan, it's automatically underwhelming. My main point is cards in the $650-1000 bracket comprise a very niche market segment, even more so than $499-549 cards. AMD doesn't necessarily need to beat 780/Titan to offer a viable choice. As you yourself stated, and with which I agree, there will be people who just want the best performance. Those gamers will spend $700-1000 on EVGA Classy 780s/HOFs and Titans of this world. There will be another group of people whose budget might be $400 but if they see a card that falls within 90% of the performance of the after-market 780s and costs $550, they might just stretch their budget a little bit but no way to $700.

What amazes me is people on our forums first complained that AMD couldn't compete on performance during 4870/5870/6970 eras but now that AMD is releasing cards at $550 starting with 7970, they are not happy with that either. And that's exactly my point about loyal NV buyers - they are 90% committed to NV anyway. Price/performance, beating NV to market, game bundles, none of these factors sway them. They don't care about price/performance (proof is 4850/4870 vs. 9800GTX/260, 5870 vs. 480, 6950 unlocked vs. 570/580, etc.). They also would still buy a more expensive slower card (proof is 680 vs. 7970GE, 1Ghz 7970 vs. 760s).

Between all the NV vs. AMD bickering, what really happened is a mid-range GPU like 560Ti went up from $249 to $499 in the form of 680/7970 and now everyone is pretending as if $650-1000 are the new normal for single GPU flagship prices. NV & AMD are the winners not us.

1. Couple of points RS, though 480 was ~15% faster than 5870 at launch it aged much better compared to 5870.Look at any recent review and you will see most of the time 480 is faring much better than 5870.It also had 512MB of extra vram lets not forget that.

2.Same way 580 aged much better than 6970 though this time the vram advantage went to AMD.Another important thing is overclocking, Fermi scaled way better than 69xx series.

 

[KingFatty]

1. Couple of points RS, though 480 was ~15% faster than 5870 at launch it aged much better compared to 5870.Look at any recent review and you will see most of the time 480 is faring much better than 5870.It also had 512MB of extra vram lets not forget that.

2.Same way 580 aged much better than 6970 though this time the vram advantage went to AMD.Another important thing is overclocking, Fermi scaled way better than 69xx series.

When you say it aged much better, what do you mean? Is it like how the silicon resists flexing or something due to heat cycling over time?