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

[blackened23]

Apparently, Maxwell is going to be 28nm yet again. Videocardz has an article about it - apparently it won't be a big jump in performance BUT it will introduce HUMA-like memory concepts and will have other optimizations.

So the performance of Maxwell at this point is questionable. Especially since it is going to be 28nm. Apparently, the cost of 20nm wafers is so ridiculously expensive that only a few players will be able to do it initially - and that volume production (eg parts that arrive in consumer hands) will be 2H 2014. I'd imagine that only large players such as Apple and Qualcomm will be able to do 20nm initially, apparently Maxwell won't be one of those players on 20nm.

If it is 20nm, DO NOT expect to see it in 3-4 months, that just will not happen. I wouldn't even expect 28nm Maxwell that soon - As mentioned, though, most sources are pointing to a 28nm Maxwell and that it won't be the same performance jump from GTX 500-600 or even 600-700 due to 28nm constraints. Most of the changes with Maxwell will be related to an entirely new memory structure and other optimizations / features. Not all-out performance. At least, that's the gist i'm getting from the article at videocardz.

 

[blackened23]

Not a good deal, except it uses less power and is about 23% smaller and faster (allegedly).

I don't care about die size, core count, all of that stuff is garbage for people to argue about on forums all day long. Do real consumers care? Not especially IMO.

The metrics that matter are performance, efficiency, price, and that's about it. As I said earlier I also have strong doubts about Maxwell since it is apparently going to be a 28nm part. The real changes with Maxwell aren't performance - but will be new features (new AA types?) and an entirely new memory structure. I also don't think it's going to be 3-4 months from now, but we shall see.

 

[BallaTheFeared]

R9 290X at 1020/1250 Vs. GTX Titan at 993/1500

http://udteam.tistory.com/539

If these are true it already lost to the aftermarket 780s.

 

[blackened23]

Yeah, i'm sure aftermarket R9X cards won't exist. Oh wait.

 

[Jaydip]

I won't be surprised if NV releases a bios update to spoil the party Though seriously I believe TU is incoming because this card is too close for NV's likings.Good time to purchase a gpu

 

[Tempered81]

A GPU can have one giant shader for all I care.

hah

 

[RussianSensation]

If these are true it already lost to the aftermarket 780s.

False. Only true if you are comparing 1.35-1.4ghz EVGA Classy/HOF cards. Let's wait to see how far this card overclocks. What if after-market 290X cards have 20-25% overclocking headroom on the GPU and 6.2-6.3Ghz on the memory? Many here expected this card to come in between 780 and Titan but it looks like it's beating the Titan.

You don't think there will be MSI Lightning 290X either?

 

[BallaTheFeared]

Yeah, i'm sure aftermarket R9X cards won't exist. Oh wait.

That's not the point, nobody cares about Titan or reference 780s anymore, just like nobody should care if AMDs reference design beats them they aren't the "benchmark" so to speak.

 

[ozzy702]

That's not the point, nobody cares about Titan or reference 780s anymore, just like nobody should care if AMDs reference design beats them they aren't the "benchmark" so to speak.

And if aftermarket R9X cards stomp the aftermarket 780's the goalpost has been moved once more. I seriously don't understand your thought process here. For all we know the aftermarket R9X cards will hit 1.3ghz on air, at which case they will be far ahead of the 780's.

Edit: You're saying that if the reference AMD card doesn't beat the very best aftermarket NVIDIA card, overclocked to it's maximum potential, that the AMD card is a let down, even though it will cost much less. The aftermarket AMD cards will cost less than a reference 780 from what it's looking like and those should be faster than cards costing 50% more.

 

[BallaTheFeared]

False. Only true if you are comparing 1.35-1.4ghz EVGA Classy/HOF cards. Let's wait to see how far this card overclocks. What if after-market 290X cards have 20-25% overclocking headroom on the GPU and 6.2-6.3Ghz on the memory? Many here expected this card to come in between 780 and Titan but it looks like it's beating the Titan.

Other websites paint a different picture, not sure if "false" is the right word here.


Wow, so from false to 1.4GHz 780s? As if you need 1.4GHz on a 780 to beat a 980MHz Titan now? Really?

You think a GPU that isn't so much faster is really going to benefit from that big of a memory bus?

 

[BallaTheFeared]

And if aftermarket R9X cards stomp the aftermarket 780's the goalpost has been moved once more. I seriously don't understand your thought process here. For all we know the aftermarket R9X cards will hit 1.3ghz on air, at which case they will be far ahead of the 780's.

Edit: You're saying that if the reference AMD card doesn't beat the very best aftermarket NVIDIA card, overclocked to it's maximum potential, that the AMD card is a let down, even though it will cost much less. The aftermarket AMD cards will cost less than a reference 780 from what it's looking like and those should be faster than cards costing 50% more.

No I'm specifically talking about stock aftermarket performance.

Nobody is buying Titans, why should anyone care if this is a "Titan Killer"? It needs to be an Aftermarket 780 killer, Titan was DOA.

 

[RussianSensation]

Wow, so from false to 1.4GHz 780s? As if you need 1.4GHz on a 780 to beat a 980MHz Titan now? Really?

You are saying after-market 780s will beat this card but the only valid comparison in this case is after-market 290X vs. after-market 780 unless you think a reference 290X will cost $680-750. You are also comparing 15-20% faster 780s out of the box to a reference card but fully ignore the possibility of after-market 290X cards. How does that even make sense? In that case, why not compare Asus DCU 290X vs. Asus DCU 780, etc.

If 290X cards are dog overclockers, then you may have a point. The game has changed entirely now if 290X reference is faster than the Titan out of the box. If this is indeed the case, then a 290X wouldn't need the same overclocking headroom as a the 780 since it would already be 7-12% faster than a stock 780.

 

[krumme]

Should be called uber mode not turbo mode. That was the intended name at least lol.

AMD will never send cherry picked or non-reference cards. Cards sent out for reviewing are only screened to make sure they aren't a lemon, they are never cherry picked. They just make sure no "bum" cards are sent out. No misleading tactics.

Sure its a nice attitude and i think it reflects amd engi culture. But its for other peoples money. Amd have wasted other people money pratically its entire lifetime. How ethical is that?
Uber oc evga 780 cards is hitting reviewers and amd personel probably thinks its "tactical". As if tactical is wrong in this game. They have the weak brand. And they behave like they are chosen by God and are the good guys. No wonder they cant make money and never will.

 

[raghu78]

If these are true it already lost to the aftermarket 780s.

complete rubbish. how did you draw that conclusion.

http://www.hardocp.com/article/2013...rectcu_ii_oc_video_card_review/9#.Uj8jb3_3x8E

firstly check the stock factory GTX 780 OC. its on par or slightly ahead of Titan. if the R290x is trading blows with Titan and doing consistently better at high res and high AA it means it does the same with GTX 780 OC. also Hawaii should have similar or better scaling wth OC given a powerful front end , lot of ROPs (44 or 48) and massive 512 bit memory bus.

i suggest you wait till official reviews are out. we will see how Hawaii XT scales with overclocking and the OC headroom.

 

[BallaTheFeared]

You are saying after-market 780s will beat this card but the only valid comparison in this case is after-market 290X vs. after-market 780 unless you think a reference 290X will cost $680-750. You are also comparing 15-20% faster 780s out of the box to a reference card but fully ignore the possibility of after-market 290X cards. How does that even make sense? In that case, why not compare Asus DCU 290X vs. Asus DCU 780, etc.

If 290X cards are dog overclockers, then you may have a point. The game has changed entirely now if 290X reference is faster than the Titan out of the box. If this is indeed the case, then a 290X wouldn't need the same overclocking headroom as a the 780 since it would already be 7-12% faster than a stock 780.

I'm saying compare them both for all I care, but leaving out the aftermarket cards of the 780 would be disingenuous to the current market conditions. Especially when you consider good aftermarket cards are the exact same price as reference.

I never saw in my handbook a clause that stated only reference could be compared to reference, nvidia has had this performance on market for almost a year so why not use the cards that have come out in that time to compare to this?

 

[blackened23]

Apparently you believe that comparing an overclocked card to a stock clock card is an "acceptable" review practice. That is just absurd.

The acceptable review practice is stock configuration vs stock configuration. If you want to analyze overclocking, then compare aftermarket to aftermarket which i'm sure will come in 1-2 months for the R9. Or, even better, you can just simply overclock the R9 card. I don't believe the R9X will have GPU Boost 2.0 type overclocking limitations, and it will have dual BIOS and 6 phase power. It should OC well, i'd imagine. But this isn't an acceptable way to review cards - reviewers make their conclusions on apples to apples comparisons which means STOCK CLOCKS VS STOCK CLOCKS.

Heck, why doesn't AMD just send reviewers golden samples that can overclock 300mhz above the stock with custom BIOS's? If you don't see the obvious problems and dishonesty associated with stock clocks vs overclocked comparisons, you're either arguing for arguing' sake or haven't thought it through. If you're going to move the goal posts, hell, why doesn't AMD send reviewers their highest clocked cards that go far above stock? In your book, it is fair game, apparently.

 

[RussianSensation]

I never saw in my handbook a clause that stated only reference could be compared to reference,

I never realized AMD doesn't make after-market cards either. I find it hilarious though since the same individuals now emphasizing after-market 780 vs. reference 290X are the same ones who ignored all after-market 7970GE cards for 18+ months & used reference 7970GE reviewer samples for power consumption, noise levels, temperature and overclocking assessment despite such version of 7970GE never being for sale in retails!

nvidia has had this performance on market for almost a year so why not use the cards that have come out in that time to compare to this?

Yup, at $1,000. No one says not to use after-market 780s when EVGA GTX780 is $10 more. However, the whole point is you keep ignoring the inclusion of after-market 290X cards too. Your comparison is completely illogical since it assumes only reference 290Xs will be for sale.

I don't recall you ever emphasizing that "fair" representation of 7970GE series was the Asus Matrix 7970 vs. reference 680s. Since Nov 2012 you ignored the existence of this card, and yet now you insist on comparing after-market 780s vs. reference 290X and imply that after-market 290X shouldn't even be considered? Why not focus on stock vs. stock and oc vs. oc? The latter covers what enthusiasts are looking for anyway.

 

[wand3r3r]

I'm saying compare them both for all I care, but leaving out the aftermarket cards of the 780 would be disingenuous to the current market conditions. Especially when you consider good aftermarket cards are the exact same price as reference.

I never saw in my handbook a clause that stated only reference could be compared to reference, nvidia has had this performance on market for almost a year so why not use the cards that have come out in that time to compare to this?

Why haven't you made any fuss that the 7970 matrix isn't in every comparison of the titan and then against the 780? That was the current "market situation" at that time. Oh wait, goalposts shifting detected.

Of course we want to see the whole situation, just reviews should be held to standards stock vs stock and oc vs oc, not in a biased manner.

 

[blackened23]

I don't recall you ever emphasizing that "fair" representation of 7970GE series was the Asus Matrix 7970 vs. all 680s. Since Nov 2012 you ignored the existence of this card, and yet now you insist on comparing after-market 780s vs. reference 290X and imply that after-market 290X shouldn't even be considered? Why not focus on stock vs. stock and oc vs. oc? The latter covers what enthusiasts are looking for anyway.

That's a good point. The 7970 IMO was a great overclocker despite it's launch issues, yet did anyone care about 1100mhz 7970s when the GTX 680 was launched? Absolutely not. Did anyone care that a 1100mhz 7970 could beat a 680? No. The comparison was stock to stock, and in that respect the 680 won. And that was the fair and proper comparison: stock clocks versus stock clocks, period. That is how the GTX 680 *should* have been and indeed was reviewed at the time - in a stock configuration vs other stock configuration cards.

This time around, the comparison will be stock to stock yet again. Anyone thinking that overclocked vs stock is a fair comparison is just in the wrong. That opens up doors to a lot of dishonesty in reviews - as I said earlier if that is fair game, then AMD could send their custom BIOS' golden samples to reviewers which OC far above stock.

 

[raghu78]

I'm saying compare them both for all I care, but leaving out the aftermarket cards of the 780 would be disingenuous to the current market conditions. Especially when you consider good aftermarket cards are the exact same price as reference.

I never saw in my handbook a clause that stated only reference could be compared to reference, nvidia has had this performance on market for almost a year so why not use the cards that have come out in that time to compare to this?

GTX Titan launched on Feb 21, 2013. GTX 780 launched on May 23rd. the really high end aftermarket GTX 780 versions started launching in August

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

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

http://www.techpowerup.com/reviews/MSI/GTX_780_Lightning/

so how the heck did 7 months from Titan launch become almost a year. btw GTX 780 has been in the market for 4 months with high end aftermarket versions for less than 2 months. there is a limit to crapping. you have an extreme bias towards Nvidia which reflects in every word of what you post.

 

[Keysplayr]

I don't care about die size, core count, all of that stuff is garbage for people to argue about on forums all day long. Do real consumers care? Not especially IMO.

The metrics that matter are performance, efficiency, price, and that's about it. As I said earlier I also have strong doubts about Maxwell since it is apparently going to be a 28nm part. The real changes with Maxwell aren't performance - but will be new features (new AA types?) and an entirely new memory structure. I also don't think it's going to be 3-4 months from now, but we shall see.

Where did you see any confirmation that Maxwell will be a 28nm part?

 

[blackened23]

There's no confirmation, but as I posted earlier there is an article on videocardz' front page about Maxwell.

http://videocardz.com/45417/why-maxwell-will-probably-launch-on-28nm-process

 

[Arzachel]

Where did you see any confirmation that Maxwell will be a 28nm part?

Rumours. I don't buy it either.

 

[BallaTheFeared]

I'd address each of your points in earnest, but since it's opinion based and I'm not changing mine and don't care what yours is on the topic I'll do us all a favor and move on.

 

[RussianSensation]

This time around, the comparison will be stock to stock yet again. Anyone thinking that overclocked vs stock is a fair comparison is just in the wrong. That opens up doors to a lot of dishonesty in reviews

It already happened with 650Ti and 660Ti where AT tested after-market NV cards against reference AMD ones.

I'd address each of your points in earnest, but since it's opinion based and I'm not changing mine and don't care what yours is on the topic I'll do us all a favor and move on.

That's nice but it still doesn't make your opinion logical. Why didn't you complain that an 1180mhz 7970 that uses a similar amount of power compared to a reference 7970GE was not tested at AT in any of 680/770 reviews? I have no qualms with including after-market 780s in a review but then I want to see after-market 290Xs too. It's only fair.