Wednesday, July 3rd 2013

Radeon HD 9000 Series Arrives This October: Report

When AMD re-branded most of its Radeon HD 7000 series SKUs to HD 8000 series, for OEMs, we saw this coming from a parsec away. AMD's next discrete GPU family for the retail channel will be placed along the Radeon HD 9000 series, and it debuts no later than this October, according to a Guru3D report. Interestingly, the report states that the first parts in the family will be based on existing 28 nanometer silicon fab processes, and will be codenamed "Curacao" and "Hainan."

We've had our run-ins with "Curacao," from time to time. It's been rumored to be an upgrade of existing "Tahiti" silicon, with 2,304 stream processors based on Graphics CoreNext 2.0 architecture, 144 TMUs, 48 ROPs, and a 384-bit wide GDDR5 memory interface. The Guru3D report adds to that with the mention of an improved front-end, which adds four asynchronous computing engines (ACEs), and three independent geometry engines.

"Curacao" will make up most of the top-tier, likely Radeon HD 9900 series; while "Hainan" succeeds "Pitcairn," filling up the performance-segment. "Hainan" reportedly features 1,792 GCN 2.0 stream processors, 112 TMUs, 32 ROPs, a 256-bit wide memory interface, and fewer ACEs and geometry engines than "Curacao."Source: Guru3D
Add your own comment

81 Comments on Radeon HD 9000 Series Arrives This October: Report

#1
jihadjoe
by: Super XP
I am not sure about this HD 9000 series? I mean, they are skipping the HD 8000 series for the performance desktop. I other words, this 9000 series better be at the very least a massive update/design to justify it IMO.
8000 series is exactly the same as the 7000 series. There aren't even any clock speed increases:

http://www.amd.com/us/Documents/AMD_Radeon_HD_8970_Feature_Summary.pdf
Posted on Reply
#2
alwayssts
by: HumanSmoke


The WCCF/Guru3D/Andwhoeverelseisreprintingthearticle seems to gloss over at least one fairly important fact - the reporting of the Hainan chip as being some performance/enthusiast part when AMD's Dave Baumann has already stated that Hainan is much lower down the product stack
This is quite true, although I reference it as Hainan as that is the accepted rumor-monger place-holder for the chip everyone expects to be 1792sp. In the end it's really just semantics until we know what it's actually called. It seems odd to reference it as simply the successor to Pitcairn.


As for 2304sp, I still do not understand how people believe this will be a native spec. It makes no sense. Not only from a product point of view (the first salvage sku is a 7970 with not needed rops?) but also efficiency.

With 2304 shaders, your mainline efficiency would be at 1111/6000 from a clock/unit pov. The same as 1536 or 768sp and respective smaller buses. Using faster ram doesn't make sense because a clock would not be reached to absorb the bw for efficiency...hence it almost certainly being a salvage part. That doesn't stop them from releasing it first and harolding the full part later as the value high-end second-coming to make as much money as possible off early adopters, but it's just not realistic (especially since 48 rops would allow for 40-44 units) for that to be 'it'. It would be a severe waste of resources.

As for everything else,

Bonaire at 1100/7000 would be essentially equal to 650ti Boost.
Not-Hainan Le/7870 will take on 660...the former probably with 1pci-e connector (150w)
1536 1xxx/6000 would be fairly similar to 760.
1792 11xx/7000 would be very similar to a 770, slightly slower but conceivably use less power (225w).
(Get why nvidia refreshed their lineup as such? AMD is predictable and they preempted each important product with similar-to-better than expected performance.)
780 is clocked expecting a 2560 part at 1000/6000. I think that AMD will take the over/under on that (2304 at 1xxx/6000 and 2560 at 1xxx/7000) hence creating a market between 770/780 and 780/Titan. Truth be told, 780 should be the most efficient with regards to units/rops/bw, but AMD's arch won't be far behind and will be good for clocks/units/bw...and in theory should be able to clock higher within the same tdp because of smaller die size granted somewhat off-set by need for greater external bw (higher mem clock).
Posted on Reply
#3
jigar2speed
by: crazyeyesreaper
which is fine and dandy but at the current power threshold they would need some huge clock speed gains along with the 300 or so shaders to make a huge improvement

its 15% shader increase with no known clock speeds but considering current designs most likely 1100 is the top clock speed to be seen on reference probably more likely to see 1000-1050

so from current looks about 15-20% if that, coupled with the fact there is still no Crossfire fix yet. Lets just say im not impressed, and I am willing to bet if the trail is followed long enough the source who gave this info is full of it.

This means AMD would release a card thats roughly on par with the GTX 780 but far more power hungry and hot, that said a game bundle like before would be nice,

Regardless this news seems a bit suspect. No disrespect you BTA but it just smells a bit to fishy but thats just my opinion.
I think i read somewhere that 7790 had GCN 1.1 due to which it had nice power saving - plus some of 7790 tweaking reached 7990 as well, due to which 7990 doesn't eat double power compared to 7970.
Posted on Reply
#4
tacosRcool
Maybe this series of cards may actually be new technology instead of Kelper 2.0 from Nvidia
Posted on Reply
#5
TheinsanegamerN
by: ViperXTR
Ahh the 9 series actually reminded me of the legendary 9700 Pro :D
oh man...the memories. that 9800 xt 256 mb ddr2 edition with copper heatsink. saved up for months to get that thing when it released. hands down, best gpu ever. stupidly overpowerful for when it came out though. (late 03?)
put that on a pair of oced 1.5 ghz pIII chips and 2 gb of sdram....man, that machine was awesome. if only the motherboard hadnt died....
Posted on Reply
#6
Ebo
The best videocard they have EVER made was a gigabyte Maya II 9700 PRO, a true piece of ingeneering. I still have mine laying behind me in the box with a Abit motherboard and a 462(AMD xp 2200, Oc'ed to 3.2Ghz) and it still rocks on win 98 with a good old game.

On to topic, i really hope the new engine CGN 2.0 till be better to take care of the microstuttering thats all that matters for me to get a upgrade.
I have 2x5850 in xfire and to my need in games thats just about enough playing in 1920x1200, but i want a new screen 2550x1440 or even the new kid on the block, the Asus 4K screen, im going to need GFX power in the end.
If 1 cant do it then i have to go xfire again.
Posted on Reply
#8
erocker
Those graphs speak of more speculation and rumor. Not to mention they don't seem very accurate as far as comparison.
Posted on Reply
#9
riffraffy
It was that 9700 Pro that converted me to ATI Radeon :D[/quote]I still have my 9800 pro on my bookshelf in fact I also have 2 radeon 4850s one 5870 next to it running out of storage room I'm still using 2 hd 6950s . AMD should send me a thank you note for putting all their kids thru college .
Posted on Reply
#10
TheHunter
by: erocker
Those graphs speak of more speculation and rumor. Not to mention they don't seem very accurate as far as comparison.
Maybe, but its aprox. timeline says that real VI 9000 (Q4 2014) is still a looong way to go..
Posted on Reply
#11
el etro
by: alwayssts
Bonaire at 1100/7000 would be essentially equal to 650ti Boost.
Not-Hainan Le/7870 will take on 660...the former probably with 1pci-e connector (150w)
1536 1xxx/6000 would be fairly similar to 760.
1792 11xx/7000 would be very similar to a 770, slightly slower but conceivably use less power (225w).
(Get why nvidia refreshed their lineup as such? AMD is predictable and they preempted each important product with similar-to-better than expected performance.)
780 is clocked expecting a 2560 part at 1000/6000. I think that AMD will take the over/under on that (2304 at 1xxx/6000 and 2560 at 1xxx/7000) hence creating a market between 770/780 and 780/Titan. Truth be told, 780 should be the most efficient with regards to units/rops/bw, but AMD's arch won't be far behind and will be good for clocks/units/bw...and in theory should be able to clock higher within the same tdp because of smaller die size granted somewhat off-set by need for greater external bw (higher mem clock).
This series sounds like a tunning on the HD 7000 series to get more perf/watt, by the way we saw the improvements on kabini igp and 7790.

I'm waiting a new pricewar... get a ~780 perf. card going to ~350$ would be nice.

And Amd can make an oficial overclocked 9970 edition to troll the Titan :laugh::laugh::laugh:
Posted on Reply
#13
Casecutter
Here's the one thing that makes no sense... What’s the wait? Obviously AMD has had such designs waiting in the wings.

Consider what they got from Malta; although we never knew if those aren't cherry picked Tahiti's, or had true improvements for power. It pretty much indicated once production could provide lower leakage chips... things have the ability to improve. While Bonaire, seemingly good return on performance and power usage, while only receiving GCN 1.1. Perhaps a more sophisticated boost algorithms, true GCN 2.0 enhancements, with higher quality production might be some if not most of this 9XXX series, but why the wait?

Figure even if AMD held off like mid-April; at that point it was "dried in concrete" to everyone what Nvidia had planned, a re-spin their GTX 6XX. Even if AMD held back to that point to finalize product specs and send to fab, wouldn't they receive volumes of die candidates, send to AIB’s and have SKU’s for sale, I though that was like 18-20 weeks meaning more August-Sept.

October’s pushing 6-7 months, what gives?
Posted on Reply
#14
erocker
by: Casecutter
Here's the one thing that makes no sense... What’s the wait? Obviously AMD has had such designs waiting in the wings.

Consider what they got from Malta; although we never knew if those aren't cherry picked Tahiti's, or had true improvements for power. It pretty much indicated once production could provide lower leakage chips... things have the ability to improve. While Bonaire, seemingly good return on performance and power usage, while only receiving GCN 1.1. Perhaps a more sophisticated boost algorithms, true GCN 2.0 enhancements, with higher quality production might be some if not most of this 9XXX series, but why the wait?

Figure even if AMD held off like mid-April; at that point it was "dried in concrete" to everyone what Nvidia had planned, a re-spin their GTX 6XX. Even if AMD held back to that point to finalize product specs and send to fab, wouldn't they receive volumes of die candidates, send to AIB’s and have SKU’s for sale, I though that was like 18-20 weeks meaning more August-Sept.

October’s pushing 6-7 months, what gives?
Sales of current series is still good. It's about making money afterall.
Posted on Reply
#15
ThE_MaD_ShOt
So basically what I am getting from this is that I am good with my 7850's for awhile.

The Ati Rage Fury pro and K6-2 is what pulled me over to the Ati/Amd kingdom and been with them since.
Posted on Reply
#16
theoneandonlymrk
by: erocker
Sales of current series is still good. It's about making money afterall.
Im personally ! Dreaming! of a 20nm miracle and it would be a miracle.:)
Posted on Reply
#17
Prima.Vera
The top cards will have 2 or 3 GB of VRAM I wander...
Posted on Reply
#18
NeoXF
by: crazyeyesreaper
which is fine and dandy but at the current power threshold they would need some huge clock speed gains along with the 300 or so shaders to make a huge improvement

its 15% shader increase with no known clock speeds but considering current designs most likely 1100 is the top clock speed to be seen on reference probably more likely to see 1000-1050

so from current looks about 15-20% if that, coupled with the fact there is still no Crossfire fix yet. Lets just say im not impressed, and I am willing to bet if the trail is followed long enough the source who gave this info is full of it.

This means AMD would release a card thats roughly on par with the GTX 780 but far more power hungry and hot, that said a game bundle like before would be nice,

Regardless this news seems a bit suspect. No disrespect you BTA but it just smells a bit to fishy but thats just my opinion.
1 GCN 1.0 core < 1 GCN 2.0 core

Also, GTX 780, and even GTX 770 for that matter, use more power than a R7970... so I dont know what you mean...
Posted on Reply
#19
crazyeyesreaper
Chief Broken Rig


much faster same power draw at 28nm theres no way AMD can increase performance to match while dropping power consumption also know one knows how good GCN 2.0 will it seems to be more of a VLIW 5 to VLIW 4 change to me which resulted in little performance difference

Going from previous AMD gpu changes

at the same process node

they will gain 10-15% max and increase power draw at peak by about 40 watts.

THe GTX 780 is roughly 18-22% faster than the 7970 GHz edition, this would mean going by previous AMD design changes and performance increases that if GCN 2.0 is much like the VLIW 5 vs VLIW 4 change then the new HD9000 series would be slightly slower than the GTX 780 unless AMD is willing to push the power draw much higher.

THis means AMDs gpu will probably peak at 260-270w in a game and hit a maximum of 300-315w in likes of Furmark while being slower. Only way to counter act that would be higher clocks, which requires better binning = higher prices or even higher power consumption.

Having owned every high end AMD gpu since the HD 4000 series, my past experiences tell me AMD will equal the GTX 780 while being slightly cheaper but will run hotter and have a much higher power draw to achieve it.

4870 1gb was 13% slower than the 4890
5870 was 12 % slower than the 6970
7970GHz edition will most likely be roughly the same difference.

4890 was 27% slower than the 5870
6970 was 23% slower than the 7970 GHz edition added 9% on to that for a 31%

Thus looking at past changes at the same manufacturering process it can be expected that the 7970 GHz edition will be around 12-15% slower than the HD 9970 or w.e AMD calls it which as it stands is not enough to make the leap to beating the GTX 780, AMD wont risk pushing out the GPUs untill they can beat it soundly by a few % points in order to prove they are "better"

Regardless it wont matter much if AMD can rival the 780 while being cheaper COnsumers win. I just do not expect the gains everyone else does.
Posted on Reply
#20
theoneandonlymrk
by: crazyeyesreaper
http://tpucdn.com/reviews/MSI/GTX_780_TF_Gaming/images/power_peak.gif http://tpucdn.com/reviews/MSI/GTX_780_TF_Gaming/images/power_maximum.gif

much faster same power draw at 28nm theres no way AMD can increase performance to match while dropping power consumption also know one knows how good GCN 2.0 will it seems to be more of a VLIW 5 to VLIW 4 change to me which resulted in little performance difference

Going from previous AMD gpu changes

at the same process node

they will gain 10-15% max and increase power draw at peak by about 40 watts.

THe GTX 780 is roughly 18-22% faster than the 7970 GHz edition, this would mean going by previous AMD design changes and performance increases that if GCN 2.0 is much like the VLIW 5 vs VLIW 4 change then the new HD9000 series would be slightly slower than the GTX 780 unless AMD is willing to push the power draw much higher.

THis means AMDs gpu will probably peak at 260-270w in a game and hit a maximum of 300-315w in likes of Furmark while being slower. Only way to counter act that would be higher clocks, which requires better binning = higher prices or even higher power consumption.

Having owned every high end AMD gpu since the HD 4000 series, my past experiences tell me AMD will equal the GTX 780 while being slightly cheaper but will run hotter and have a much higher power draw to achieve it.

4870 1gb was 13% slower than the 4890
5870 was 12 % slower than the 6970
7970GHz edition will most likely be roughly the same difference.

4890 was 27% slower than the 5870
6970 was 23% slower than the 7970 GHz edition added 9% on to that for a 31%
Vliw 4-5 ,??, I don't see it, and I think all the focus being placed on shaders might mean naught.
To me the details aren't there on gcn2 but I'm expecting a design aimed to differentiate via Hsa tweeks , after all ioummu or whatever it is has arrived and at rev 2 + ie established, and they are pushing hsa hard now , there Will imho be full dx11.2 support in architecture and therein another possible advantage.
Posted on Reply
#21
Am*
Sounds like a pretty mediocre card -- I sure hope it is more powerful than it sounds (they need it to be at least 80% faster than 7970). 4K resolution monitors/TVs should hit mainstream markets next year at latest and tests done by various websites proved that even a single Titan cannot pull off any decent framerates at that resolution. In fact, even BF4 alpha testing shows that current gen cards fail miserably at getting anywhere close to high settings in next gen games, even at 1080p. Current GPUs have all had great value thus far against current 7+ year old consoles but when next gen consoles hit, Nvidia and AMD need to bring their A-game and at least double the performance of every tier of cards sold for the same price as current gen cards, otherwise nobody will buy them (who wants another repeat of ATI's 2900XT? Nobody, but we all want another Nvidia 8000 series to follow up to next gen consoles).
Posted on Reply
#22
Casecutter
by: erocker
Sales of current series is still good. It's about making money afterall.
Well things somewhat went in "free-fall" with the 760 release and subsequent stock and price that GTX660Ti had been holding. As soon as folks got a look at the 760, price on GTX660Ti tumbled and pressed the last of the GTX660 to fall on the GTX650Ti Boost.

On the 78XX, now even 7850 2Gb are like $165 -AR and 7870 have been at $180 –AR having to slip with the GTX660 GTX650Ti Boost sell-off. Now yes AMD probably still has substantial chips for both in the channel, you figure all such wafer capacity shouldv'e moved off the Pitcairn and onto Hainan. I think they'll slowly bring them to a trickle to try to re-attain a pricing of $160/$210. As much as the 7790 was to back-fill the 7850 1Gb's, I think they're here to stay, vying what is the GTX650 Boost 1Gb.

I think a bulk of 79XX Sku's started slowly drying up after the 780, with most AIB's holding with just a few nicer Boost/Ghz offerings while maintain acceptable stock levels. Most all the more generic stuff (not Boost/Ghz) have already been waning. I think AMD and there AIB will finish out the summer aiming to recapture most of the what's the respective $270/410 price points.

I don't think AMD can see enjoying making less money on Tahiti/Pitcairn parts, and possibly even some sales being taken by "suppose" new models from Nvidia in the sweet spot of the enthusiast market. Sure if they've huge volume in the channel, but why are they in that position? There's been an abundance of time to work out such a transition. It's as if someone dragged their heels on switching scheduled 28Nm wafer starts from Tahiti to Curacao / Pitcairn to Hainan. The process is the same, they certainly held a schedule wafer starts, the design/engineering has been in place by all accounts, but somehow here they are 5-6 months in a holding pattern. Is it only me that it’s not making sense?

The only thing is if they had made it like August it could be 6 months till a Volcanic Islands, if they stay with October they’ve just a 3 month lull.
Posted on Reply
#23
Prima.Vera
by: Am*
Sounds like a pretty mediocre card -- I sure hope it is more powerful than it sounds (they need it to be at least 80% faster than 7970).
Hard to believe they will keep the same performance increase like they did back in the day.

3870->4870->5870 Ah good days.
Posted on Reply
#24
Frick
Fishfaced Nincompoop
by: Prima.Vera
Hard to believe they will keep the same performance increase like they did back in the day.

3870->4870->5870 Ah good days.
Historically it has been up and down. 80% would be fantastic for sure, but it won't happen and it can't happen.
Posted on Reply
#25
Casecutter
by: Prima.Vera
3870->4870->5870 Ah good days.
3870->4870 had even held to the same 55nm process but it really was like 36% increase. But that I believe was more what GDDR5 brought, AMD even surprised themselves... The good old days when those who first jumped on the GTX260 found their AIB’s feeling obligated to print reimbursement checks! ;)

by: Frick
Historically it has been up and down. 80% would be fantastic for sure, but it won't happen and it can't happen.
True, one might figure CuracaoXL offering about 25% above a Tahiti Ghz, while holding to a similar if better efficiency. Figure 10% from increase Sp and TMU’s, while there's conceivably something from alleged 50% increase in ROP’s. Weigh in low leakage production (Malta) parts curtail power usage, possibly even while presenting 10% increase in clocks. All that while on the same or close to the same 28nm die size, so conceivably for a $500 MSRP. That places it about par with a 780 working from a chip that is 30-35% smaller than the GK110, for like 28% less money.

But seems it just won't come soon enough... Had AMD had them last May such printers might have ran agian. :ohwell:
Posted on Reply
Add your own comment