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

[RussianSensation]

dp

 

[RussianSensation]

So we are back to most bang per buck

I'll just remind you that people in the know aka sushiwarrior claims faster than Titan ULTRA!

You are talking about "breaking dreams" but when the competitor sells a card for $400 more, having a card 35% faster is not impressive, not even a little bit. Unless gaming is your primary hobby, 98% of people are not going to be spending $400 for only a 35% increase in performance. Also, 770 4GB selling for $440-450 is not impressing anyone but the most die hard NV fans/ignorant PC gamers. That's a huge waste of money for most of us. If say there is a card from AMD that delivers 90% of the performance for only $399, most of us could care less about an NV card for $650. If AMD releases a card for $599 and it's 90% of the performance of the after-market 780, well then most won't care about the AMD card.

When people choose to ignore price/performance, it really amazes me as if most of you have no other bills in life?

Everything goes back to hype vs. value vs. realistic exceptions. Price/performance, overclocking, stock performance and realistic specs are all valid points of discussion. Expecting only 1 scenario of AMD beating 780 with some hypothetical 500mm2 chip or else they fail is being close minded. Also, just because AT member ABCD states Card A > Card B, am I supposed to believe that? I heard that same for 6970 vs. 580 too.

 

[f1sherman]

Expecting only 1 scenario of AMD beating 780 or else they fail is being close minded.

I know. You have to have a plan B.

But now you are REALLY covering all your bases

 

[Zanovar]

So we are back to most bang per buck

I'll just remind you that people in the know aka sushiwarrior claims faster than Titan ULTRA!
^^ And that's EXACTLY how AMD brakes dreams (not promises)

IMHO

--

Im curious,did he actually say that?,i highly doubt it.not saying that he didnt,but the guy seems quite careful with what he says.

 

[RussianSensation]

I know. You have to have a plan B.

But now you are REALLY covering all your bases

In my mind I already said I don't expect R9 280X to beat 780 after-market versions because even if it matches them out of the box, NV's 780 cards overclock from 1.08-1.1Ghz to 1.3-1.4Ghz on air (HOF/EVGA Classy, etc.). I think NV got this round. :thumbsup:

On the point you made that sushiwarrior stated R9 280X is somehow faster than Titan Ultra, we don't even have the performance or specs of Titan Ultra/Titan II. Let's say Titan Ultra is 10-15% faster than Titan I, does that mean R9 280X will be 20% faster than the Titan then? I find that hard to believe.

 

[f1sherman]

Im curious,did he actually say that?,i highly doubt it.not saying that he didnt,but the guy seems quite careful with what he says.

Aims for Titan Ultra. NVIDIA won't be able to respond unless they clock GK110 to extremes. I believe those were the words.


You know... the usual fluff - like NV will remain noncompetitive through 2013 and most of 2014.
Quite lucid if only slightly controversial viewpoint that's often found on S|A

keywords: NVIDIA, unable to respond, uncompetitive, in deep crisis

 

[Jaydip]

You missed the best part fish, all of that in < $600 package.

 

[Black Octagon]

Wow, 23 September hasn't even arrived yet and already it's been proposed that AMD be judged by the 'dreams' of people on forums.

 

[seitur]

Never in the history of GPU making has AMD ever released any details about its future GPUs, not during 4870/5870/6970 or 7970 eras. You cannot break promises "almost consistently through the history" when the firm made no such promises in the first place. Let's not take random people posting specs online as some information AMD/NV promise to gamers.

I might as well make claims that Maxwell GTX880 will be 2x faster than Kepler GTX780 in Crysis 3 and when it fails, I scream that NV failed to meet their "promises"? AMD promised nothing at all for R9 280X not even a concrete release date.

Also, if to you getting 75% of the performance for $400 less is breaking dreams, I'll take that all day. When a 1.05ghz 7970 is going or $290 and GTX770 4GB sells for $440 while GTX780 sells for $650, NV can keep its overpriced lineup.

This basically.

Agreed with what you say.

-------------

Seriously folks. Stop with stupid unrealistic expectations and hype and for NV prices. Well if you're wanting to buy NV at premium prices and you ditch AMD GPUs even on timeframes when they sell for less. Well you have yourself to blame.

 

[Jacky60]

In the UK the 7970 has plummeted in price from a launch price of £550 to £240 today. I can't see AMD selling any of the new flagship at £400-£500 unless its twice as fast as the 7970. The 7990 is to be found under £500 so I will be buying two 7990's unless the 9970 blows the Titan far out of the water. I think the Titan is about to be ass raped.

 

[Hitman928]

In the UK the 7970 has plummeted in price from a launch price of £550 to £240 today. I can't see AMD selling any of the new flagship at £400-£500 unless its twice as fast as the 7970. The 7990 is to be found under £500 so I will be buying two 7990's unless the 9970 blows the Titan far out of the water. I think the Titan is about to be ass raped.

They could EOL the HD7XXX series and are clearing out inventory now, meaning there won't be stock of the HD7XXX series to buy (at least not used). It looks to me that the new series will launch right about the same price the HD7XXX series did, we'll find out soon enough.

 

[boxleitnerb]

&#8220;They&#8217;re coming in Q4. I can&#8217;t reveal a pricepoint but we&#8217;re looking at more traditional enthusiast GPU pricepoints. We&#8217;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.

&#8220;It&#8217;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.

&#8220;Another thing I can tell you is about the process node: this GPU is in 28nm. Some have speculated that it was 20nm and it&#8217;s not for a specific reason: At 28nm for an enthusiast GPU, we can achieve higher clock speeds and higher absolute performance.&#8221;

http://www.forbes.com/sites/jasonev...eon-cards-next-gen-consoles-7990-criticism/1/

 

[bystander36]

You are talking about "breaking dreams" but when the competitor sells a card for $400 more, having a card 35% faster is not impressive, not even a little bit. Unless gaming is your primary hobby, 98% of people are not going to be spending $400 for only a 35% increase in performance. Also, 770 4GB selling for $440-450 is not impressing anyone but the most die hard NV fans/ignorant PC gamers. That's a huge waste of money for most of us. If say there is a card from AMD that delivers 90% of the performance for only $399, most of us could care less about an NV card for $650. If AMD releases a card for $599 and it's 90% of the performance of the after-market 780, well then most won't care about the AMD card.

When people choose to ignore price/performance, it really amazes me as if most of you have no other bills in life?

Everything goes back to hype vs. value vs. realistic exceptions. Price/performance, overclocking, stock performance and realistic specs are all valid points of discussion. Expecting only 1 scenario of AMD beating 780 with some hypothetical 500mm2 chip or else they fail is being close minded. Also, just because AT member ABCD states Card A > Card B, am I supposed to believe that? I heard that same for 6970 vs. 580 too.

Keep in mind that performance is more than FPS numbers. Those who use multiple cards, 3D or feel a need for PhysX will come away with a different performance comparison than someone who only reads the FPS numbers on a FRAPS run.

Granted, Titan's and 780's are expensive, and may not offer a lot more than more value cards, but if that is the performance you want, you will not get it from using a cheaper card.

 

[iMacmatician]

http://www.forbes.com/sites/jasonev...eon-cards-next-gen-consoles-7990-criticism/1/

The die size range gives us reality to some supposed leaks / insider info claiming a &#8805; 470 mm^2 die (unless AMD has for some reason more than one upcoming &#8805; 400 mm^2 chip).

 

[ShintaiDK]

The talk about why 28nm instead of 20nm is just plain dumb. 20nm isnt ready and they couldnt make anything on 20nm.

The chip isnt much bigger than Tahiti GPU. I wonder if they axed the majority of the compute performance in favour of gaming performance. Perhaps also the chance of finally dividing the gaming and HPC segment for GPUs. The Titan for example is, while a great card, abit of a misfit with its heavy compute performance.

Its a shame node process for GPUs seems to be 3 years now.

 

[RussianSensation]

Keep in mind that performance is more than FPS numbers. Those who use multiple cards, 3D or feel a need for PhysX will come away with a different performance comparison than someone who only reads the FPS numbers on a FRAPS run.

Granted, Titan's and 780's are expensive, and may not offer a lot more than more value cards, but if that is the performance you want, you will not get it from using a cheaper card.

GTX760 SLI beats both the 780/Titan for less $. Also, does not explain how 770 commands such an unreasonable price premium over 7970 Ghz versions in the US.

&#8220;It&#8217;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.

561 / 1.30 ~ 432mm2 but he said nearly 30% bigger so let's say the next flagship from AMD is probably between 430-440mm2 max.

The chip isnt much bigger than Tahiti GPU.

It's going to be at least an 18% increase in die size. I think that's enough for this card to be 20-25% faster than 7970GE on avg. That's enough to bring it to parity with a reference 780. If AMD launches this for $550 and bundles BF4 with the card, that's like getting it for $500. It probably won't match the HOF or Classy 780 cards but with the game bundle and lower price, it'll have something to offer. AMD's biggest threat is Maxwell. It's supposed to be a far greater leap than Kepler was over Fermi based on all the tech in it.

With a wave of unoptimized games coming out (COH2, TWR2, ArmaIII), the entire 780 vs. R 280/290X is underwhelming. This gen is a pass for me until 20nm and Witcher 3.

 

[boxleitnerb]

SLI/CF cannot be compared to single-GPU. Profile dependency, microstuttering, input lag, incompatibility with some graphics options (in Crysis 3 for instance). Yet you ignore it again...and again...and again and make these silly comparisons. Just because Fraps says "40 fps" does certainly not mean it feels like 40 fps. You should leave these matters to people who actually have tried SLI/CF.

 

[Daedalus685]

http://www.forbes.com/sites/jasonev...eon-cards-next-gen-consoles-7990-criticism/1/

If GK110 is 30% larger (keeping in mind he likely means something like 28% since he is trying to pad a claim for better efficiency) the new AMD chip would be roughly 432mm^2. Which is, again roughly, 22% larger than Tahiti.

This of course tells us nothing about performance or anything. Kind of odd they mention area efficiency and not raw power... It would be a disaster of epic proportions if it is LESS efficient per mm than GK110 at gaming...

Edit: You know... if Matt Skynner knows anything about the next GPU and the best he can say is "it is the most performance for the die size" I'd have to guess that means not faster than some form of GK110. Which if it lowers prices makes me happy as it is upgrade time. But that won't be particularly exciting.

 

[RussianSensation]

SLI/CF cannot be compared to single-GPU. Profile dependency, microstuttering, input lag, incompatibility with some graphics options (in Crysis 3 for instance). Yet you ignore it again...and again...and again and make these silly comparisons. Just because Fraps says "40 fps" does certainly not mean it feels like 40 fps. You should leave these matters to people who actually have tried SLI/CF.

You forget I had multiple 7970s for more than 1.5 years now and Tri-SLI 470s before my 6950? Thanks for the reminder on SLI/CF. There are plenty of people with 670/680/780 SLI and 690 that do not complain about their performance. If the performance for SLI was so awful, why are so many gamers using SLI?

Here is a fun fact - in the most beautiful PC game now, 760 SLI beats a 780 without much effort. Overclock those 760s and a 780 OC would be left in the dust.

Again, you are not addressing what's being discussed. When cards like 7970 1Ghz/GE sell for $280-320 regularly in the US and the cheapest 780 is going for $650, its 35% advantage is not impressive in the slightest because 97% of the market will simply ignore its existence in the context of the performance increase it offers. Most people will just get a card with 75-80% of the performance and set aside $350-400 for a 20nm GPU upgrade that will beat their $300 card by 70-80%. You and a couple other guys on our forums dropping $1000 on Titans is not representative of the market. If NV released Titan II for $1,500 and it was 20% faster than Titan I, you probably would upgrade to that. For you price/performance is irrelevant but it isn't for 97% of GPU buyers who don't buy GPUs > $550.

Right now many newer PC games are poorly optimized which means they are not worth buying until future patches fix their performance. Most other titles if you turn down 2-3 settings, your FPS goes from 50%-200% higher. Most people understand that it's a poor strategy to buy a $700-1000 GPU and hold on to it for 5-6 years. As game developers start making newer games on newer game engines (Frostbite 3, latest CryEngine, UE4) and we see a wave of next gen PC games in the next 2-3 years, that's going to be the time to upgrade. Spending $700-1000 on GPUs now near the end of PS360 gen without a single next gen PC title, from a graphical perspective, doesn't make a lot of sense. Those who do are viewing PC gaming as their primary hobby or are completely price inelastic or use their GPU for work too/perhaps get their firm to finance their Titans for pro work. I would put this group at 3% or less of the entire desktop dGPU market. Let's not even forget that a lot of us got 2-3 7970s absolutely free so in that context a $1,000 GPU that's 35% faster is outrageous when the alternative was free.

It would be a disaster of epic proportions if it is LESS efficient per mm than GK110 at gaming...

I don't see how it can be worse. Right now GK110 is way worse than Tahiti XT if you start using the metric of performance/mm2. I think people care a lot more about performance/watt than performance/mm2 though. If someone gave you 2 GPUs and said 1 has a die size of 10mm2 and the other a die size of 900mm2, it really doesn't matter for the consumer unless the card uses 300W and needs water cooling from overheating. The performance/mm2 measurement is more tied to transistor density/engineering perspective.

 

[boxleitnerb]

Because people get used to it. Just like people play console games without AA and don't complain - yet if you show them proper image quality, it can open their eyes. AFR and non-AFR is simply not comparable, period. Add to that the fact (you yourself said that a couple of days ago) that there often aren't profiles for new games, thus you end up with only 50% of the maximum performance of your 2-way AFR-setup.

Stop making this about me (again) and just admit that your comparison with AFR is flawed. What cards I have and why I bought them is absolutely none of your business and it has nothing to do with my response to your post which specifically adressed your non-sensical comparison. Next time you drag me into this, I will report you, be warned.

 

[raghu78]

430 - 440 sq mm is a bit disappointing and now the chances of a Titan killer are zero. thats sad. even if AMD has made perf / sq mm improvements and has gone for a better process like 28nm HPM the best we can expect is parity with GTX 780 on a clock for clock basis. now its a matter of aggressive pricing and game bundles to add more value and differentiate from the competition. so now it looks like Hawaii XT for USD 550 with BF4 would be a good deal.

 

[Daedalus685]

I don't see how it can be worse. Right now GK110 is way worse than Tahiti XT if you start using the metric of performance/mm2. I think people care a lot more about performance/watt than performance/mm2 though. If someone gave you 2 GPUs and said 1 has a die size of 10mm2 and the other a die size of 900mm2, it really doesn't matter for the consumer unless the card uses 300W and needs water cooling from overheating. The performance/mm2 measurement is more tied to transistor density/engineering perspective.

Yeah, that's my point. It would be such a step backwards if it were not the best per/mm^2 in the enthusiast price range it is aimed at that I wonder why on earth he even mention it instead of saying "We have no comment about price or performance". Just struck me as very odd.

The cynic in me wants to attribute it to early damage control but it is likely just his way of saying nothing in as many words as possible.

 

[Crap Daddy]

So, judging by what Mr. Matt Skynner says " 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" and "I can&#8217;t reveal a pricepoint but we&#8217;re looking at more traditional enthusiast GPU pricepoints" it will be GTX780 level of performance while less expensive, available in a month. The title of this thread should be changed to ChipLoco humor.

 

[RussianSensation]

Stop making this about me (again) and just admit that your comparison with AFR is flawed. What cards I have and why I bought them is absolutely none of your business and it has nothing to do with my response to your post which specifically adressed your non-sensical comparison. Next time you drag me into this, I will report you, be warned.

I am not dragging you into anything. I am only using an example that 3% of the market that ignores price/performance wrt to $400-600 pricing gaps between lower level cards and ultra-enthusiast segment are not indicative of the general market. You seem to have trouble understanding this point time and time again. Based on this AMD does not need to beat 780/Titan or it somehow fails. Many posters in this thread continue to insinuate that unless AMD somehow beats 780/titan, then it has failed. There are plenty of other ways that AMD can deliver a competitive offering through game bundles and price/performance. And threatening to report me every time I post something that you disagree with makes you sound very immature. If you do not agree with something - i.e., SLI vs. CF vs. single GPU comparisons, please point it out. There is no need to make threats if you do not agree with an opinion of another poster. Furthermore, you often pick and choose what you want to address. You still didn't address the 770 vs. 7970GE pricing disparity in the US instead deflecting this point that in Germany this isn't the case. That doesn't explain why NV continues to charge such large premiums for its cards in the US, a market far larger than Germany for GPU purchases.

The point is AMD doesn't need to do anything with R9 single GPU against a $1,000 Titan. Their engineering and financial resources are better spend on the $550 and below line-up and saving whatever they have to compete with Maxwell.

430 - 440 sq mm is a bit disappointing and now the chances of a Titan killer are zero. thats sad.

If there is a card that delivers 90% of the gaming/DP performance of the Titan for $550 with BF4 bundle, it's not a Titan killer? That's an interesting observation considering so many posters on our forum claimed 7970 to be a failure when 680 undercut it by only $50 at launch and beat it by less than 10%.

 

[SiliconWars]

Very weak decision by AMD, 500mm2 would have beaten Titan but they didn't have the balls to go for it.