[VC] AMD Radeon HD 9000 Series Launching in October?

Quote:

---

June was surely a month of GeForce GTX 700 series. NVIDIA launched three cards in less than 30 days without any actual response from AMD.

If you thought that AMD engineers were on vacation then you are mistaken. Apparently they were working on new GPUs, especially a Hawaii silicon, which is believed to power the upcoming Radeon HD 9970 and HD 9950. Yes, that?s not a typo, AMD will supposedly drop the HD 8000 naming reserved for OEM-parts, and use HD 9000 instead. This was rumored for a long time, but people were not convinced enough to confirmed it. According to my sources this is true, there won?t be any Radeon HD 8970 non-OEM card, which if you think about actually makes sense. Imagine two cards having the same name, but different specs. Maybe now people will stop posting Radeon HD 8970 OEM datasheet as a news about a leaked document directly from AMD site (seriously don?t do that).

According to ChinaDIY, AMD could be preparing its Radeon HD 9000 series for October, which is the launch month for Battlefield 4, a Gaming Evolved title. My information is more optimistic. Unless the roadmap has changed, AMD could show off the new series in September. Some sources even claim it could be the first week of September. Regardless of what month is it going to be, we are two or three months until the new series arrive. Most likely AMD will start with HD 9970, which if the rumors are true will be faster than GeForce GTX 780, but slower than GTX TITAN. However the big question is not the performance, but the price. If the card would cost $550, then we have a clear winner.

The Chinese source also claims that AMD will resurrect the Radeon 9550, only now with the HD-part. Interesting fact is that it?s almost exactly 9 years since the launch of this card.

---

http://videocardz.com/43446/amd-rade...nching-october


Bring me the HD9990 goodness.

More Like Nvidia Story : Geforce 3xx and 4xx Series.

Edit :

Quote:

---

Most likely AMD will start with HD 9970, which if the rumors are true will be faster than GeForce GTX 780, but slower than GTX TITAN

---

What the hell? Slower than Titan After two generations ? Man , AMD HD9970 should be match At least Geforce 690 GTX just like HD7970 Is more powerfull than Geforce 490 GTX (?)

AMD needs to bring back the 9700Pro at $299, like the good old glory days :D

If the GPUs are still made on 28nm, I wouldn't be surprised if they could not quite beat the Titan.

Shouldn't these be on a smaller node? We are creeping up on dual 8pin at 28nm and that's just no fun.

Quote:

---

Originally Posted by behrouz

More Like Nvidia Story : Geforce 3xx and 4xx Series.

Edit :



What the hell? Slower than Titan After two generations ? Man , AMD HD9970 should be match At least Geforce 690 GTX just like HD7970 Is more powerfull than Geforce 490 GTX (?)

---

But titan is still $1000 while the AMD cards should be much cheaper than it, no real need to dump excessive R&D into such a project for a small market.
The 590 actually was a better card than the 7970 http://www.anandtech.com/bench/Product/508?vs=516

A $550 card that's ~5-10% slower than Titan depending on the game engine?...and that's not even being optimistic. This sounds like the AMD/ATI we've come to know, they always come through since AMD took over. These cards will be beast and probably get a bump in up in VRAM if the new console games are going to soak up that 8GB they have. Also, I'm pretty sure the 9000 series would have been on a smaller node if TSMC could of done it. I got high hopes for AMD's GPU division, I think over the next few years Nvida is going to become overwhelmed...they just don't price their cards right, they took the performance lead and instead of taking over the $550 price tier...they just start a new $650 and $1000 tiers??? WTF Nvidia, you have become to conceded or you simply can't produce affordable chips. I love how when Nvidia delivers the next gen first we get a huge price hike, but when AMD does it they do the right thing and take over the top tiers and lower the last gen's prices. Honestly, I think Nvidia did us consumers dirty this time because they didn't get the Xbox or Playstation deal...and they are worried. I'm not starting a flame war but, I really think Nvidia should have replaced their last gen. line-up with the new 700 series...not pull this OMG WE GOT THESE SUPER UBER SO FAST CARDS THAT AMD WILL NEVER CATCH UP TO SO WE GOT TO SELL THEM FOR $300 AND $500 MORE THAN LAST GENERATION OMG AMD DRIVERS WILL NEVER WORK AND WE CONTROL THE VIDEO MARKET SO GET DOWN ON YOUR KNEES AND PAY UP FOR ROYALTIES WE TOOK ALL YOUR CASTLES! Yeah, right Nvidia...2 months later AMD's drivers are fixed and 4 months later AMD's cards are just as fast but still $300 and $500 cheaper because your new price tiers only exist at Nvidia.

Yes, I realize Nvida has always price gouged early adopters and many people are aware of how Nvida plays. It's just so anoying that they do it though. They sit there and dangle the current fastest card in-front of your face and say,"oh but wait, we wan't $200 or $300 more than it's actually worth until we get some competition." Then a couple months later they drop the prices to $50 to $100 more then AMD's competing card and justify the extra price as being the premium that people who want physics and cuda have to pay. Which makes sense if you want that stuff I guess. Eh, I just don't care about proprietary gimmicks that really don't seem to deliver. Truly, Nvidia has in the past been the one with exclusive partnerships with game developers...Hence their "The Way it Was Meant to be Played" slogan. But, as of late, AMD seems to be the ones taking over that role with exclusive partnerships with developers. Nvidia, is just playing the big name reputation card and price gouging harder than they ever have before...We'll see how it plays out, but my $$$ is on AMD.

Quote:

---

Originally Posted by behrouz

What the hell? Slower than Titan After two generations ? Man , AMD HD9970 should be match At least Geforce 690 GTX just like HD7970 Is more powerfull than Geforce 490 GTX (?)

---

What two generations?
7970 was Jan 2012 and Titan was Feb 2013...

Quote:

---

Originally Posted by boxleitnerb

If the GPUs are still made on 28nm, I wouldn't be surprised if they could not quite beat the Titan.

---

This is the first source that has stated Hawaii is slower than Titan.
But I do agree with you and m_jones, it wouldn't be overly surprising and they don't need to overspend on R&D and manufacturing with 20nm ramping up in 1H 2014.

Quote:

---

Originally Posted by Chickenfeed

Shouldn't these be on a smaller node? We are creeping up on dual 8pin at 28nm and that's just no fun.

---

20nm won't be ready until the end of the year at the earliest, though it will likely be sometime 1H 2014 before we see any 20nm cards in retail.

Quote:

---

Originally Posted by LordEC911

This is the first source that has stated Hawaii is slower than Titan.

---

The most funny is the source of the original article ( Inpai ) absolutely dont write that. Look like they really want to write between the lines. In reality they even write the complete invert story .

Bringing about 30% more performance is going to be tough for AMD(to beat a gtx 780). Its going to take a bigger than 400mm2 die to do it if they stay on 28nm. Adding shaders doesn't do that much alone as similarly clocked 7970/ 7950 show. Add in AMD needs to add more compute stuff if they want to have people take it seriously as a professional card and it seems kind of improbable.

, even in more favorable conditions, AMD brought 15-20% more performance to the table with the 5870 to 6970 change. This was easier to bring about because the 5870 used a lot less power than a 7970 uses, was smaller and the 5870 had the tesselation weakness.

Scaling upward with die size, it doesn't seem GCN is particularly that good with power consumption and AMD needs high clocks for their 7xxx series to perform well.

Before jumping on the hawaii band wagon, people have to remember the jump between 5870 and 6970. People hyped it up like the second coming and look how that delivered.

Leaking this type of rumor seems more like a move from AMD to get people to wait for their cards as they have nothing to release for a while.

I think Nvidia should have a gtx 785 ready in october(2496 park with a ddr5 at 1750mhz) for whatever AMD releases if this rumors turns out true.

However, I think it would be best for AMD to focus their resources on getting their 20nm card off the ground first, then this 28nm project. Even Nvidia seem happy not making true revisions and they have a lot more money and alot more marketshare. With the little cash reserves AMD has, it doesn't make sense to make a big 450mm2vanity project as they don't have the professional market in place to support it yet. The apple deals isn't close enough. They need super computing projects in the 50 million plus region like the titan super computer.

Quote:

---

Originally Posted by Miwo

AMD needs to bring back the 9700Pro at $299, like the good old glory days :D

---

The $299 dollar high end is dead. R and D and silicon are too expensive. The closest thing we get to a glory card like that is something like the gtx 460 when it came out. A mid range performing closer to the high end and bringing prices down.

Big chips at 300 dollars are not profitable.

Wow its being optimistic but here's to AMD beating GTX 780 and being single card king !!

Quote:

---

Originally Posted by Zloyd

Wow its being optimistic but here's to AMD beating GTX 780 and being single card king !!

---

The difference is not so high, 22 % at 1440p, and surely a bit more at 1080p .. more or less 25% ... http://www.hardwarecanucks.com/forum...review-15.html


Maybe IF they release first a 28nm gpu around November, this could be the midrange, ( this can explain the " 9550 " gaming story ), followed in Q1-Q2 2014 by the 20nm parts, high end and computing...

Quote:

---

Originally Posted by Lanek

The difference is not so high, 22 % at 1440p, and surely a bit more at 1080p .. more or less 25% ... http://www.hardwarecanucks.com/forum...review-15.html


Maybe IF they release first a 28nm gpu around November, this could be the midrange, ( this can explain the " 9550 " gaming story ), followed in Q1-Q2 2014 by the 20nm parts, high end and computing...

---

That last bit was my first thought when the first rumors of VI came about, AMD making a "highend" gaming chip like GK104 that was compute-lite.
But with GCN development/release they have basically stated that they are sticking to a top-bottom compute architecture and won't be going that route.

I don't really see them having a problem beating GTX780 and competing with Titan.

whats ive heard, will be launched sooner then october and highest one-gpu model beats Titan

Looking at AMD so far, it makes sense that they do a refresh at 28nm now, and do a stable 20nm release in around Q3 2014.

No idea what the source is .. or from where he have take the info.

http://www.guru3d.com/news_story/amd...n_october.html

Quote:

---

More and rumors are circulating that AMD may introduce its Radeon HD 9000 as early as October. The company is anticipated to skip the 8000-series as it already sells Radeon HD 8000-branded parts to OEMs. The first Radeon HD 9000 family parts are anticipated to be Curacao and Hainan, two 28nm chips based on the GCN 2.0 architecture. The new architecture reportedly has an improved front-end with 4 asynchronous computing engines [ACEs] and 3 geometry engines, as well as a higher number of stream processors.

Since at present AMD sells re-badged Radeon HD 7000 products as Radeon HD 8000-series products to OEMs, it is logical for the company to utilize Radeon HD 9000 sequence for its new line of graphics cards. ATI Technologies’ original Radeon 9000-series products based on code-named R300-family graphics processing units revolutionized the market of graphics cards a decade ago and helped ATI (which now belongs to AMD) to become the producer of market leading graphics solutions.

The Curacao XT graphics processor is expected to feature 2304 stream processors (36 compute units), 144 texture units, 48 render back ends and 384-bit memory controller. The Hainan is projected to have 1792 stream processors (28 compute units), 112 texture units, 32 render back ends and 256-bit memory controller. Both chips will share the same front-end (just-like current-gen Radeon HD 7900 and 7800 do) with 4 asynchronous computing engines [ACEs], 3 geometry engines, command processor, global data share and so on.

---

How come Hainan (HD 9950?) sounds slower than a 7950?

I know this stuff is as yet unconfirmed, but does it look like there's going to be a Curacao Pro, as it were (2048 stream procs, 128 texture units, 48 ROPs, 384-bit mem bus)? That's what I want to buy: faster than a 7970 and around 400 dollars.

Quote:

---

Originally Posted by Miwo

AMD needs to bring back the 9700Pro at $299, like the good old glory days :D

---

I'd love for AMD to pay homage to the cards that really made them famous. HD 9700Pro - has a good ring to it!

Quote:

---

Originally Posted by [XC] Lead Head

I'd love for AMD to pay homage to the cards that really made them famous. HD 9700Pro - has a good ring to it!

---

I just bought one 2 weeks ago. Great card! :D It also OC well on my nostalgia rig (XP3000+, 1GB DDR400, nForce 2 MCP) and on AIR stock card 350MHz GPU is not a problem!

Back in old days I had 9500non-Pro modded to 9700 :) That was mega upgrade! :eek:

Quote:

---

Originally Posted by [XC] Lead Head

I'd love for AMD to pay homage to the cards that really made them famous. HD 9700Pro - has a good ring to it!

---

Well it was ATI at this time, but yes.. bring us back the 9700Pro ... I still have a 9700Pro Maya edition here ( + a 9800Pro )... ( One Maya was dead on overclocking but the second still working, just cant find an AGP board now )

What do you guys think...if AMD releases a 9970 that's as fast as Titan and a 9950 that's faster than the 780, will they replace the current higg-end price range with them? I think we will see a $100 bump so 9950 is $450 and the 9970 at $600 on release. That would be the sweet spot to undercut Nvidia because no way will Nvidia go that low and Nvidia fans will consider AMD's offerings at those prices if performance is comparable and the frame pacing driver fix works...AMD will turn some heads.

Lol heres comes the hype train. At 20nm I expect these expectations to be met. At 28nm its going to be derailed like every single other launch recently.

lol there went the de-hype-train

sigh..

Quote:

---

Originally Posted by STEvil

lol there went the de-hype-train

sigh..

---

When was the last time the hype train hasn't been derailed? Especially something like the revision like gcn1 --> gcn2. Only time it has been met was a new architecture or/and a new node. And even then the its been a crap shoot.

AMD has its work cut out when their 7970 ghz needs 30 percent(when u include a larger bench suite of games, as sleeping dogs, hitman, dirt showdown skew an average of 6 games) to beat a gtx 780.

7970 ghz already consumes 250 watts of power and scaling is already being limited as more components are being added as seen between the 7970 and the 7950.

If AMD tries to increase the efficiency of the GCN cores in a utilization sense, powers going to go up like it did with the 6970 and 5870 series. If they simply add more parts, they are going to run into a utilization of those cores issue(gk110 already runs into this problem and keplar is more efficient per core for gaming as seen with the competitiveness of gk104 vs tahiti). With the 5870, they were using so little power in the first place, that they could afford the first option. But with a 7970ghz using as much power as it does, the results are going to be more limited because they are stuck on 28nm.

A 2400 core gcn at 1000 mhz is not going to outperform a 2300 core keplar unless they are willing to break some power rules which increase utilization. One of the biggest coveats of gk104 performance is the 256bit bus for memory. With both hawaii and gcn both likely running on 384 bit buses, its going to come down to core efficiency and power consumption. GCN consumes too much power at higher clocks and the cores themselves are less efficient.

http://www.hardocp.com/article/2013/...4#.UdOaBW1u97Q

Look at this review for the gtx 760 vs an overclocked 7950. The overclocked 7950 uses 100 more watts, has a 384 bit memory bus, has 50 percent more cores, is running those cores about 10% slower than a overclocked gtx 760 but only out performs it by 5%.

For such a big monolith part, gk110 is really efficient. Efficient enough it kills the 7970 ghz for performance per watt and at the relevant revolution is on top of the performance per watt charts.

The 5870 had a massive performance per watt advantage over the gtx 480, yet its successor could not beat the fixed gtx 580. With tahiti running behind in performance(there a similar difference between a gtx 780 and 7970 ghz compared to the difference between a 5870 and gtx 580) and way behind in the performance per watt, its successor is going to have a hard time beating the gtx 780.

ROLLO go out of the body of Tajoh ... Watts / performance are extremely relative, specially if you compare the size of a 4Tflops 7970 vs a GK110 who are not even able to hit this ... on DP rate Nvidia got his lead.. but it is not really efficient.. they need 1/3 rate DP for just beat amd of about 10% of performance when AMD still use 1/4 rate .. ( you are speaking about efficiency in games.. not of the architecture ... games performance are extremely relative )...

The K20x need 520mm2 for just been able to match ( take the lead in DP ) of a 365mm2 chips ... ofc we can debate a long time who is better on games.. but if we speak about real computing thing ( graphics or not ) ... things change a lot.

If AMD can put the same efficiency on games have Kepler .. ( because the efficiency is here, the waste of power got GCN vs Kepler GK110 ( note for make work the GK110 correctly you need switch to K20 style)... here we are speaking about the efficiency of an architecture.. and this dont ask about more shaders ... Bigger L2 cache, more ROPs balance, bigger or faster front end, improve the branch units, a better use of scalar units ?

When you are speaking about efficiency, try to remember the 7970 or S9000 is justg 365mm2 vs a 520mm2 chip...

Architecture performance or efficiency of this architecture are not delivered by games average fps.. or start to write : " i dont think they can push more of a 20% more increase in Crysis 3 "

Quote:

---

Originally Posted by tajoh111

snip

---

You are skewing numbers and not taking into consideration many factors when comparing architectures/products.

What I'd expect to see, and being realistic about those expectations, are the 9950 beating / matching the 770 at $399 range and the 9970 in the 499-549 range and slightly behind the 780 (10% +/-). Pending game bundles and aftermarket options they could both end up being better value than Nvidia's offerings.

As you said tajoh, the head room on 28nm has nearly run its course so unless they've made some impressive gains efficiency wise, improvements will likely be akin to the 5870-6970.

Quote:

---

Originally Posted by LordEC911

You are skewing numbers and not taking into consideration many factors when comparing architectures/products.

---

I am being realistic here. The only time AMD has exceeded my expectations was the 4870(the 9700 pro days is too long ago) and that was a case of where AMD got to 2.5x everything about the 3870 because the 3870 was so small to begin with. The only time Nvidia has met my expectations were actually this generation and the gtx 8800. Both cases where we got a card faster than the dual chip predecessor and then some.

The gtx 760 and 7870 are very similar component for component. The 7870 has 12 percent more shaders, the gtx has 25 percent more bandwidth and they have the same clocks, but the gtx 760 is 20-25% faster. This is exactly why gk104 is able to catch up with tahiti which has way more shaders and computational power.

Component wise, a 2304 shader caracao xt(the 4024 shader rumor is ridiculous this generation) is going to have a hard time catching up to a 2304 shader gtx 780. GCN 2 better makes some pretty drastic improvements in efficiency all around, without cutting out the computing power.

AMD gave up something with the 7xxx series to make the chip more compute friendly. Unless they are going to cut some of that out, they are going to have a hard time beating gk110. gk110 552mm2 seems well utilized this generation and seems a bit more compute gaming orientated than the last generation. Combining both of these characteristics makes its so AMD has a hard time catching up to them with a 400nm or less chip, even 450nm might not do it. A fully spec'd gk110 has pretty scary performance potential.

If they were willing to push themselves to the point where some efficiency is lost like the 7970 ghz (which is partially made up with turbo) and used the same ddr5 in the gtx 770, they could probably make a card 50 percent faster than a ghz edition(15% faster than what they have now; a extra 100 mhz(some overclocked cards already have this like the evga acx), 14% more bandwidth and 192 more shaders). The gtx 580 or gtx 285 had no where near this potential.

Quote:

---

Originally Posted by tajoh111

I am being realistic here. The only time AMD has exceeded my expectations was the 4870(the 9700 pro days is too long ago) and that was a case of where AMD got to 2.5x everything about the 3870 because the 3870 was so small to begin with. The only time Nvidia has met my expectations were actually this generation and the gtx 8800. Both cases where we got a card faster than the dual chip predecessor and then some.

The gtx 760 and 7870 are very similar component for component. The 7870 has 12 percent more shaders, the gtx has 25 percent more bandwidth and they have the same clocks, but the gtx 760 is 20-25% faster. This is exactly why gk104 is able to catch up with tahiti which has way more shaders and computational power.

---

A HD 9870 (pitcairn update) with 1792 sp , 256 bit memory at 7 Ghz (224 Gb/s) , 32 rops and improved front end (4 ACE, 3 geometry engines) and clocked at 1.1 - 1.15 ghz will easily match or exceed a stock gtx 770. power consumption will be around GTX 680 (200w TDP). Pitcairn has already shown that GCN (stripped of DP fat) is very efficient. a Pitcairn scaled to compete with GK104 Kepler should be a very close fight in perf and perf/watt.

http://www.guru3d.com/news_story/amd...n_october.html

Quote:

---

Component wise, a 2304 shader caracao xt(the 4024 shader rumor is ridiculous this generation) is going to have a hard time catching up to a 2304 shader gtx 780. GCN 2 better makes some pretty drastic improvements in efficiency all around, without cutting out the computing power.

AMD gave up something with the 7xxx series to make the chip more compute friendly. Unless they are going to cut some of that out, they are going to have a hard time beating gk110. gk110 552mm2 seems well utilized this generation and seems a bit more compute gaming orientated than the last generation. Combining both of these characteristics makes its so AMD has a hard time catching up to them with a 400nm or less chip, even 450nm might not do it. A fully spec'd gk110 has pretty scary performance potential.

If they were willing to push themselves to the point where some efficiency is lost like the 7970 ghz (which is partially made up with turbo) and used the same ddr5 in the gtx 770, they could probably make a card 50 percent faster than a ghz edition(15% faster than what they have now; a extra 100 mhz, 14% more bandwidth and 192 more shaders). The gtx 580 or gtx 285 had no where near this potential.

---

firstly the current HD 7970 / HD 7970 Ghz chips use the same silicon except for higher voltage and BIOS changes. it was the first GPU chip on TSMC 28nm process and released on Jan 9, 2012. the Tahiti chip must have taped out in late Q1 2011. The design must have been extremely conservative as it was done on a immature 28nm TSMC process. the TSMC 28nm process has improved and so has AMD's learning of the 28nm process. AMD's HD 7790 Bonaire (160 sq mm) was able to fit in 30% more perf at the same 85W TDP of HD 7770 Cape Verde (123 sq mm). for a mere 37 sq mm the front end was doubled and the sp count went from 640 to 896. the power management was improved and the HD 7790 was a very power efficient chip. the HD 9970 / HD 9950 and HD 9870 / HD 9850 chips will incorporate the improved power magament of Bonaire (plus maybe even newer stuff) and have aggressive voltage binning (improved tsmc process) to keep TDP well under control to reach their respective perf and TDP targets.

the HD 9970 Hawaii XT is expected to a 2560 sp chip with the hawaii pro a 2304 sp chip. Tahiti was also a chip with bottlenecks. its front end was not able to keep the sp well fed. the ROPs too were bottlenecked. by improving the front end to 4 ACE , 3 geometry engines and increasing ROPs to 48 the per sp performance can easily increase by 10%. add to it 25% more sp and its not difficult to see a 30% perf increase at the same 250w tdp. remember the HD 7970 Ghz had a stock voltage of 1.25v as compared to HD 7970 which ran at 1.175v.

The Hawaii XT can easily run at 1 Ghz at 1.175v - 1.2v and provide 25 - 30% higher perf than HD 7970 Ghz and stay within 250W TDP with the improved power management and chip binning. if a 560 sq mm , 7 billion transistor, GTX Titan running at upto 1 ghz boost can stay within 250w tdp there is no reason why the hawaii chip which should be around 5.5 billion transistors and 440 sq mm cannot stay within 250w tdp while running at 1 ghz boost clocks.

http://www.anandtech.com/show/6774/n...nce-unveiled/2

"On our sample the top bin is 1.1625v, which corresponds to a 992MHz core clock. Overvolting Titan to 1.2 means unlocking two more bins: 1006MHz @ 1.175v, and 1019MHz @ 1.2v. Or put another way, overvolting on Titan involves unlocking only another 27MHz in performance."

both AMD / NVIDIA naming number make me sick

Quote:

---

Originally Posted by Chickenfeed

What I'd expect to see, and being realistic about those expectations, are the 9950 beating / matching the 770 at $399 range and the 9970 in the 499-549 range and slightly behind the 780 (10% +/-). Pending game bundles and aftermarket options they could both end up being better value than Nvidia's offerings.

As you said tajoh, the head room on 28nm has nearly run its course so unless they've made some impressive gains efficiency wise, improvements will likely be akin to the 5870-6970.

---

I think this is more realistic than some of what has been posted in here. But if they are able to compete (or come within 10%) of the Titan, then this will be fantastic news. I was just about to pull the trigger on a couple of 780's. But I think I will hold off to see what sort of numbers these cards can do.

"A few days ago, AMD released its latest Catalyst beta graphics driver ? and hidden within, AMD has helpfully provided a long list of new GPU and APU code names that the driver is compatible with, and provided a little more background on the highly anticipated Kaveri desktop APU. The successor to the Kabini and Temash mobile APUs will be called Mullins. On the desktop, discrete GPU front, we now know that there will be an 8000-series (Sea Islands) GPU called Tonga, and that the 9000-series (Volcanic Islands) will feature Hawaii and Vesuvius GPUs."

http://www.extremetech.com/tag/vesuvius
http://www.extremetech.com/computing...ealed-analyzed

I'm starting to get confused.

So this cards still aren't the real deal? They are the same as those new 8xxx cards, that apparently are only for OEM's, still on 28nm and still just a stopgap.


Are these cards even next-gen-crammed-on-28nm-while-waiting-for-20nm, or just tweaked-old-gen-with-bigger-die? I would say business wise, second option would make more sense with 28nm being more mature and all.


Oh Ok, just read the #16 post on this thread, it's old gen, makes sense.

Quote:

---

Originally Posted by raghu78

firstly the current HD 7970 / HD 7970 Ghz chips use the same silicon except for higher voltage and BIOS changes. it was the first GPU chip on TSMC 28nm process and released on Jan 9, 2012. the Tahiti chip must have taped out in late Q1 2011. The design must have been extremely conservative as it was done on a immature 28nm TSMC process. the TSMC 28nm process has improved and so has AMD's learning of the 28nm process. AMD's HD 7790 Bonaire (160 sq mm) was able to fit in 30% more perf at the same 85W TDP of HD 7770 Cape Verde (123 sq mm). for a mere 37 sq mm the front end was doubled and the sp count went from 640 to 896. the power management was improved and the HD 7790 was a very power efficient chip. the HD 9970 / HD 9950 and HD 9870 / HD 9850 chips will incorporate the improved power magament of Bonaire (plus maybe even newer stuff) and have aggressive voltage binning (improved tsmc process) to keep TDP well under control to reach their respective perf and TDP targets.

---

This is a really interesting approach to a new process that completely throws the "test low end cards first" out of the window.

You first deliver you high end arch, that despite being on a non mature process will still be a competitive decent card, you then start releasing the mid and low end cards, testing new features on them, and also and more important, measuring the process advances, being able that way to better calculate the performance\target TDP of a refresh high end card on a mature process.

Quote:

---

Originally Posted by Chickenfeed

What I'd expect to see, and being realistic about those expectations, are the 9950 beating / matching the 770 at $399 range and the 9970 in the 499-549 range and slightly behind the 780 (10% +/-). Pending game bundles and aftermarket options they could both end up being better value than Nvidia's offerings.

As you said tajoh, the head room on 28nm has nearly run its course so unless they've made some impressive gains efficiency wise, improvements will likely be akin to the 5870-6970.

---

Even then, it's still a hell of a bargain. I think 9970 will at least match the 780 though.

Wait what?
OK so I'm confused now. 9000 is coming in October? Did I blink and miss 8000?
I'm looking around for a new card ATM, and was considering 7970, but someone suggested waiting a few months for the 8000 cards. Now I'm reading 9000's are coming soon?

8000 series are rebadged 7000 series for OEMs. So they are skipping and going to 9000 to avoid "confusion"

At a recent sponsored LAN event I attended, power color representatives were asked about the new cards being 20nm, which they quickly confirmed. I cannot verify the accuracy of their answer so take it with a grain of salt.

Interesting. If this is true it's quite the feat.

Quote:

---

Originally Posted by Ace123

At a recent sponsored LAN event I attended, power color representatives were asked about the new cards being 20nm, which they quickly confirmed. I cannot verify the accuracy of their answer so take it with a grain of salt.

---

Based off of the specs/power consumption which I have witnessed, I can more or less confirm this...

If true, how did they manage this? Mass production hasn't started yet, so they must have opted for risk production which should be quite expensive.

Quote:

---

Originally Posted by boxleitnerb

If true, how did they manage this? Mass production hasn't started yet, so they must have opted for risk production which should be quite expensive.

---

Actually, thinking about it again I can see how 20nm kind of makes sense.

Originally the product line was supposed to be "stable through 2013" and somehow changed to a major refresh in the second half of the year. If VI is a 20nm product, then it's possible that the first test batch out of the oven came out far better than anyone at AMD expected and they pushed the launch up. It's still borderline unbelievable, but explains the massive shift in AMD's timetable better than a 28nm VI ever could.

If true, given the cost of the new node and the lack of capacity this early on I'd expect a 20nm VI to be very expensive, very supply constrained, or both. Could end up being the mother of all paper launches.

It's so confusing - some people say 20nm, others say 28nm. I hope concrete information is leaked, soon.

20 nano...cue the Jaws theme :cool:

Just want to point out that risk production doesn't necessarily mean that it is more expensive, they would be getting wafers for the same price as mass production, it would just mean that they are taking a slightly larger risk than usual that yields will be acceptable on the process with their design. This would mean either they have a near perfect design for the process or they have built in more redundancy to negate the effects of the new process.

Still finding it pretty hard to believe that the 20nm rumor is getting so much traction. I still find it highly unlikely but not impossible.

FYI- There was always a new product launch planned for 2H '13... Some people just didn't want to listen, looking at you Sky.

Quote:

---

Originally Posted by LordEC911

This would mean either they have a near perfect design for the process or they have built in more redundancy to negate the effects of the new process.

---

Or they could pull an Nvidia and have a few CUs disabled, enable them later on when yields improve and launch a refresh.

Quote:

---

Originally Posted by LordEC911

Still finding it pretty hard to believe that the 20nm rumor is getting so much traction. I still find it highly unlikely but not impossible.

---

Well, thanks to that Power Color rep there's actually some evidence for 20nm now besides just wishful thinking. Still seems incredible though.

Quote:

---

Originally Posted by LordEC911

FYI- There was always a new product launch planned for 2H '13... Some people just didn't want to listen, looking at you Sky.

---

Can you elaborate? The entire 'stable through 2013' thing seemed pretty cut and dry to me.

The "stable through 2013" slide only included Q3, not Q4.
As for the Power Color reps...who says they weren't lying or just misinformed? Seems unlikely to me that they would give away such an important piece of information just like that.

I thought the other rumor was 22nm. I thought 28nm was all but dead
Although a move like this could explain why kavari has also been delayed a bit.
The idea they are taking a move right out of nvidia's play book by disabling parts of the chip until yields are better makes perfect sense though. It would allow them to operate on a new fab, cores that don't make the cut for 9970 could be used on lower product lines based on how much of the core is good, and it leaves room for a product refresh on the same core once fab yields are better by enabling more of the core later. If you think about it in this sense, the risk is alot lower.

Kaveri is made at GF afaik, so this has nothing to do with TSMC's 20nm process.

Quote:

---

Originally Posted by hurleybird

Or they could pull an Nvidia and have a few CUs disabled, enable them later on when yields improve and launch a refresh.

Well, thanks to that Power Color rep there's actually some evidence for 20nm now besides just wishful thinking. Still seems incredible though.

Can you elaborate? The entire 'stable through 2013' thing seemed pretty cut and dry to me.

---

1. True, but that would add more power and potential leakage to their ASIC, not to mention the die size. IMO AMD can't really afford to do that.

2. Keep in mind that false/bad info is given out on purpose to diff members of the press and AIBs to try to find leaks.

3. The fact that they were in the process of taping out or had already taped out at least one, probably two, ASICs right around the time of the interview...

Quote:

---

Originally Posted by boxleitnerb

The "stable through 2013" slide only included Q3, not Q4.
As for the Power Color reps...who says they weren't lying or just misinformed? Seems unlikely to me that they would give away such an important piece of information just like that.

---

Most likely, they didn't know and just agreed. If they actually had heard something, its hard to dismiss the fact that it was probably misinformation.

Quote:

---

Originally Posted by Ace123

I thought the other rumor was 22nm. I thought 28nm was all but dead
Although a move like this could explain why kavari has also been delayed a bit.
The idea they are taking a move right out of nvidia's play book by disabling parts of the chip until yields are better makes perfect sense though. It would allow them to operate on a new fab, cores that don't make the cut for 9970 could be used on lower product lines based on how much of the core is good, and it leaves room for a product refresh on the same core once fab yields are better by enabling more of the core later. If you think about it in this sense, the risk is alot lower.

---

28nm or 20nm. 22nm was never even discussed/rumored since TSMC skipped 22nm.
Kaveri was rumored to have been pushed back, most likely it was a conservative move for "mass availability."

Huumm, dont forget AMD have offtly use risk production.... and they was really good at it ( mean the "so called" risk production" was a sucess ( on engeener side ) " ... they have offtly use it . it will not be the first time ..

Risk mean hazrdous in term of yields, this dont mean hazardous in term of cost.... The GPU team is really good to push this, and dont forget, if the 20nm is ready tomorrow, they work on it since 3 years allaready .. ( this dont mean i say i think they use 20nm )

Quote:

---

Originally Posted by boxleitnerb

The "stable through 2013" slide only included Q3, not Q4.
As for the Power Color reps...who says they weren't lying or just misinformed? Seems unlikely to me that they would give away such an important piece of information just like that.

---

Whistling past the graveyard?

Of course if it was a rumor that NV had pulled it off....:rolleyes:

Hmm im quite confident they will beat 780 ...... dont think anyone cares about SuckerTan. 7970 has 680 comfortably ATI engineers push and repeat the performance !!! (8950 insane bang for buck card incoming :D hope i can get 2))

For anyone in the know, when can we expect full cover water blocks to be available for the 9970 and 9950??? Thank you.

Lol ^

Quote:

---

Originally Posted by Final8ty

Lol ^

---

That is, when does E.K. or Swiftech usually release a block? And what are you laughing at...dude? There is no doubt in my mind that AMD will release a card faster then Nvidias and will still cost less even after buying the water block. Also, I just got my first Loop this weekend...so i'm a little restless.

I'm only laughing because the cards are not out yet and still away's off.

You've got the itch real bad :)

Quote:

---

Originally Posted by DefStar

That is, when does E.K. or Swiftech usually release a block? And what are you laughing at...dude? There is no doubt in my mind that AMD will release a card faster then Nvidias and will still cost less even after buying the water block. Also, I just got my first Loop this weekend...so i'm a little restless.

---

They usually get the card specs early enough that they can have blocks out right around launch, usually between a month or two for full coverage.

Ek should get a full waterblock when the card release or in the weeks who follow.. ( i dont remember have buy an AMD card at release without having my EK full cover block, the next week who have follow ).. they got the design schema really fast and they adapt their full cover block really quick with it...

I thought there was suppose to be a frame pacing driver released today or did I miss something?

It's just being tweaked a little. Have patience my young padawan

Whenever AMD launches a product, hardware or software doesn't matter, without any delays the last 2 years, a major earthquake happens somewhere...

Quote:

---

Originally Posted by PedantOne

whats ive heard, will be launched sooner then october and highest one-gpu model beats Titan

---

That's what is commonly referred to "wishful thinking". The only way they'd get close to Titan's performance is if the HD 9xxx series was manufactured using the 20nm process. And as we all know, TSMC is still 6 months away from even beginning "low-volume production". Which means high-volume production will be another couple months after that.

Transcript of the shareholder meeting where this info came from can be found here: http://www.semiwiki.com/forum/conten...4nm-track.html

no, Hawaii is still on 28nm, but more SPs and higher clocks will reach Titans performance level

Quote:

---

Originally Posted by Andrew LB

That's what is commonly referred to "wishful thinking". The only way they'd get close to Titan's performance is if the HD 9xxx series was manufactured using the 20nm process. And as we all know, TSMC is still 6 months away from even beginning "low-volume production". Which means high-volume production will be another couple months after that.

Transcript of the shareholder meeting where this info came from can be found here: http://www.semiwiki.com/forum/conten...4nm-track.html

---

As I've said before, 20nm is NOT the only way they can match/beat Titan.

As time passes 20nm seems more likely.

Quote:

---

Originally Posted by LordEC911

As I've said before, 20nm is NOT the only way they can match/beat Titan.

---

I find it unlikely they will go for a chip as big as GK110. So I agree with Andrew LB.

And I don't believe in magical huge increases on largely the same architecture...

Quote:

---

Originally Posted by zalbard

I find it unlikely they will go for a chip as big as GK110. So I agree with Andrew LB.

And I don't believe in magical huge increases on largely the same architecture...

---

It might not beat Titan, but it should be close. Titan is around 35% faster than 7970GE and higher clocks + cores + tweaks should be able to give it ~30% more performance. Should be similar to a GTX 780 while consuming similar if not less power.

Some say that the power consumption would be way too high in the above scenario but I dont think so. Aside from the core/process tweaks, they have a big room for improvement with regard to turbo power states/voltages and from what ive seen from Bonnaire, its pretty significant and can only get better.

Quote:

---

Originally Posted by Tao~

As time passes 20nm seems more likely.

---

Highly unlikely, unless they wait until 1st or 2nd quarter 2014.

Whatever it is I just hope its not $1000.

Quote:

---

Originally Posted by LightSpeed

It might not beat Titan, but it should be close. Titan is around 35% faster than 7970GE and higher clocks + cores + tweaks should be able to give it ~30% more performance. Should be similar to a GTX 780 while consuming similar if not less power.

Some say that the power consumption would be way too high in the above scenario but I dont think so. Aside from the core/process tweaks, they have a big room for improvement with regard to turbo power states/voltages and from what ive seen from Bonnaire, its pretty significant and can only get better.

---

Tweaks at this points are just that, tweaks. They hardly lead to big jumps because its not a new architecture. Adding more shaders is typically less efficient at this point because theirs a bottle neck in the architecture somewhere. The small difference between a 7970 and 7950 at equal clocks show this. Higher clocks and more cores = higher power consumption. Increasing utilization of existing shaders is going to increase power consumption further like what happened to the 6970 vs 5870.

30% is no small task at the same node, with no new architecture and GCN being kind of a energy pig as it is. Adding 25 percent more shaders an increasing clocks another 200mhz is going increase power significantly. I think adding any more shaders will require AMD to add more ROPS which will increase the size of the chip significantly.

Its no coincidence that Nvidia clocked their gk110 line lower. Making a bigger chip causes significantly more power consumption, more so than performance often because of leakage. It just happens that Nvidia's boost technology helps salvage a big more performance while it also seems Nvidia is less reliant on clocks this gen.

AMD brought 40 percent more performance initially over the 7970(not counting driver boosts). They had a new architecture and a new node. Although more came with drivers. GCN 2 is likely to be similar to gcn 1 and thus driver boosts won't be that much.

Getting the next 9970 to perform 30 percent faster while generally conditions are against it(gcn being big as it is and consuming as much as it is, no new node and no new architechture) make it something difficult to pull off, unless they want to make a 450mm2 plus chip.

If the 7970 was around 300mm and didn't consume so much power when overclocked, I could see AMD doing this. But the 7970 is 365mm and drinks power like a glutten at 1200mhz.

~440-480mm2 and roughly the same clocks, maybe a bit lower, should be enough.

Improved front end, more/stronger ROPs, 30-40% more shaders.

I'm not sure why you think ROPs take up a bunch of room, that was more true when ROPs were linked to the memory controller which isn't the case with GCN.

Also, with a mature process they can potentially increase transistor density, remove some redundancy that was needed for Tahiti and used the knowledge they got from the rest of 7x00 series and the 8x00.

Will it be easy? Nope.
Is it impossible? Nope.

Quote:

---

Originally Posted by tajoh111

Tweaks at this points are just that, tweaks. They hardly lead to big jumps because its not a new architecture. Adding more shaders is typically less efficient at this point because theirs a bottle neck in the architecture somewhere. The small difference between a 7970 and 7950 at equal clocks show this. Higher clocks and more cores = higher power consumption. Increasing utilization of existing shaders is going to increase power consumption further like what happened to the 6970 vs 5870.

30% is no small task at the same node, with no new architecture and GCN being kind of a energy pig as it is. Adding 25 percent more shaders an increasing clocks another 200mhz is going increase power significantly. I think adding any more shaders will require AMD to add more ROPS which will increase the size of the chip significantly.

Its no coincidence that Nvidia clocked their gk110 line lower. Making a bigger chip causes significantly more power consumption, more so than performance often because of leakage. It just happens that Nvidia's boost technology helps salvage a big more performance while it also seems Nvidia is less reliant on clocks this gen.

AMD brought 40 percent more performance initially over the 7970(not counting driver boosts). They had a new architecture and a new node. Although more came with drivers. GCN 2 is likely to be similar to gcn 1 and thus driver boosts won't be that much.

Getting the next 9970 to perform 30 percent faster while generally conditions are against it(gcn being big as it is and consuming as much as it is, no new node and no new architechture) make it something difficult to pull off, unless they want to make a 450mm2 plus chip.

If the 7970 was around 300mm and didn't consume so much power when overclocked, I could see AMD doing this. But the 7970 is 365mm and drinks power like a glutten at 1200mhz.

---

I mostly agree with what you said, but this was their first shot at GCN and a completely revamped architecture. There are bottlenecks that can be ironed out and as I said, Bonnaire showed what could be done simply with a better implementation of voltage/power states. Also, Tahiti consumes a lot of power while overclocked/volted but a new architecture coupled with process tweaks will surely enable it to operate at lower voltages.

I wonder if AMD are willing to make a ~450mm^2 architecture though.

ok, it can be maximaly 2560 SPs, 40 ROPs, same number of TEX units. Same memory capacity, bus, clocks maybe
with the same GPU clock as HD 7970 GHz, it can be +20-25 percents performance above Radeon HD 7970 GHz in best case

Quote:

---

Originally Posted by PedantOne

ok, it can be maximaly 2560 SPs, 40 ROPs, same number of TEX units. Same memory capacity, bus, clocks maybe
with the same GPU clock as HD 7970 GHz, it can be +20-25 percents performance above Radeon HD 7970 GHz in best case

---

I'm thinking (the biggest they would do):
More shaders- 44CUs or double the SIMDs and TFUs per CU and go with ~24CUs(3072ALUs and 192TFUs) [depends on how much a diff the stronger FE makes]
Better frontend- the 4ACEs and 3geometry rumor thats been floating around sounds like a decent increase.
Slightly more ROPs- I don't see them going with 48ROPs unless they move to 512bit w/ 6ghz GDDR5(unlikely) or starve/make the ROPs more bandwidth efficient.
If they use 7Gbps GDDR5, that would offer ~27% more bandwidth allowing them to do the 40ROPs, a 25% increase.

Core clocks would be under 1ghz, probably around 900-950mhz(one would hope).

That should fit in <500mm2(imo 440-480mm2) and I would assume stay around 250-265w TDP.

That should roughly match Titan.

Quote:

---

Originally Posted by LordEC911

I'm thinking (the biggest they would do):
More shaders- 44CUs or double the SIMDs and TFUs per CU and go with ~24CUs(3072ALUs and 192TFUs) [depends on how much a diff the stronger FE makes]
Better frontend- the 4ACEs and 3geometry rumor thats been floating around sounds like a decent increase.
Slightly more ROPs- I don't see them going with 48ROPs unless they move to 512bit w/ 6ghz GDDR5(unlikely) or starve/make the ROPs more bandwidth efficient.
If they use 7Gbps GDDR5, that would offer ~27% more bandwidth allowing them to do the 40ROPs, a 25% increase.

Core clocks would be under 1ghz, probably around 900-950mhz(one would hope).

That should fit in <500mm2(imo 440-480mm2) and I would assume stay around 250-265w TDP.

That should roughly match Titan.

---

as a cfx user with a 1600p screen if i wanted to match titan i would have bought a titan but titan will not do 1600p with a single card reliably

i want a card that is very close to a 7990 so i will wait for 20nm on both camps

cfx and sli was fine with cards that were sub 450 bucks but at 650 to 1000 no, i'll keep my 7970 cfx and 680 sli till they can do it with a single card

Quote:

---

Originally Posted by bill_d

as a cfx user with a 1600p screen if i wanted to match titan i would have bought a titan but titan will not do 1600p with a single card reliably

i want a card that is very close to a 7990 so i will wait for 20nm on both camps

cfx and sli was fine with cards that were sub 450 bucks but at 650 to 1000 no, i'll keep my 7970 cfx and 680 sli till they can do it with a single card

---

A 28nm maxwell at around gk110 size should do that.

gk110 with gk6000 specs(fully enabled), 100 more mhz and ddr5 at gtx 770 clocks should be within 10 percent of the gtx 690 or 7990. Heck a respin, with better memory on current gk110, should get this. Particularly if they are willing to raise the TDP and if the respin increases yields.

Adding the gains with maxwell should enable Nvidia to get even more than that.

Quote:

---

Originally Posted by tajoh111

A 28nm maxwell at around gk110 size should do that.

gk110 with gk6000 specs(fully enabled), 100 more mhz and ddr5 at gtx 770 clocks should be within 10 percent of the gtx 690 or 7990. Heck a respin, with better memory on current gk110, should get this. Particularly if they are willing to raise the TDP and if the respin increases yields.

Adding the gains with maxwell should enable Nvidia to get even more than that.

---

if maxwell is 28nm no thanks i like in phoenx if i whan that much heat i'll go outside :p:

and a "gk110 with gk6000 specs(fully enabled)" hey i guess the 1500 buck price point is open:down: pass

Quote:

---

Originally Posted by tajoh111

A 28nm maxwell at around gk110 size should do that

---

Maxwell won't be built on 28nm...

The initial Maxwell might. Depends on how ready 20nm will be in Q1 2014, how good yields are and how accurately Nvidia predicted this state at least 18 months when the decision what node to use was made.

Quote:

---

Originally Posted by boxleitnerb

The initial Maxwell might. Depends on how ready 20nm will be in Q1 2014, how good yields are and how accurately Nvidia predicted this state at least 18 months when the decision what node to use was made.

---

If AMD is willing to build a near 500mm2 chip, then 20nm is further off than we think and I could see some maxwell chips being 28nm.

It also might occur due to TSMC pricing on 20nm. 28nm was a huge jump in cost over 40nm. 20nm might be also a big jump which makes big chips too expensive to manufacture initially, similar to gk110.

Quote:

---

Originally Posted by boxleitnerb

The initial Maxwell might. Depends on how ready 20nm will be in Q1 2014, how good yields are and how accurately Nvidia predicted this state at least 18 months when the decision what node to use was made.

---

maxwell will be 28nm!!! Trust me :) First samples are done