-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Chiphell - AMD HD8000 series GCN2.0 in Q3
- Thread starter Boomstick777
- Start date
- Joined
- 2 Jan 2012
- Posts
- 12,453
- Location
- UK.
It's not really that amazing, they just nerfed the compute side of the card so overall it uses less power. The 7970 is a more complete type of monster, hence uses a lil more power.
I hope AMD bring it with the 8XXX or 9XXX series and embarrass Nvidia's overpriced offerings
Blimey what happened here? LOL.
My point was that if these leaked specs are anything to go by then the new performance level set by titan could possibly mean the potential new 7950 is no longer 5% behind the potential new 7970. I've not even mentioned titan pricing or anything so please direct your fanboy talk elsewhere.
I didn't really blame anyone, my point was titan has set a new performance level that might affect AMD's card structure and those leaked specs would indicate that. Nothing more, nothing less.
Blimey what happened here? LOL.
My point was that if these leaked specs are anything to go by then the new performance level set by titan could possibly mean the potential new 7950 is no longer 5% behind the potential new 7970. I've not even mentioned titan pricing or anything so please direct your fanboy talk elsewhere.
Fanboy
You're the one blaming NVIDIA/Titan for... what, exactly? AMD are in control of their own products, research and development and pricing. I focused on pricing but that's merely one angle of a larger point.
Edit: dodgy auto correct on phone
Last edited:
Its not really a case of blaming so i don't know why you're getting so defensive over it. If the leaked specs are true then the only reason the 8970 will be seeing such a large performance increase is to keep up with the titan id imagine. I don't see the problem in suggesting that. Blaming Nvidia or titan pricing is not the aim of my post, but its safe to assume that titan has forced the hand somewhat.
HaHa Scotty! Note to self in future don't even bother to mention the N word or the T word on this forum.
a) pointing out a company being autonomous is not defending, it's fact and just how it is...
b) Titan is nothing to do with what AMD do; their card structure is completely their own choosing
I know its their own choice but however you want to dress it up the titan performance level has forced the hand. If titan hadn't of come along then this wouldn't of happened and AMD would not need to have made such a large 'potential' improvement to the 8970. Some might even argue this is a good thing, but that will depend upon the potential cost at the time.
This is just my personal view and this does not translate to I blame Nvidia, Titan owners and Nvidia pricing in general so please don't make it out like that, there's really no need.
I look back at your quoted post and i see lines such as 'lol that's rubbish' and 'incoherently blaming NVIDIA and Titan because it doesn't fit into your own brand preference'. Please stop spouting such crap and read what i actually said. I think my point is a fair one and if you don't like it then fine but please don't try and bait me over it.
Calm down.
It hasn't forced anybody's hand. AMD are free to do what they like with their line up and their pricing. So it is "lol rubbish". Just because it's your personal view doesn't mean it can't be wrong nor questioned. Opinions can be wrong when they're formed on questionable thought processes. As I have said, pricing was just one angle of a larger point. It isn't difficult to comprehend this.
It wasn't just me who pointed out it was "blaming", eh?
. Perhaps the issue was your original post and the way it was formed. Looks like you agree as it was now edited 
. It really isn't very complicated: AMD are in control of their own destiny be that line up, pricing etc. To suggest that the Titan is going to affect AMDs own line up is quite an incoherent point to make.
Last edited:
Titan will affect it because its the fastest card and you can be sure that AMD will want that crown back at some point. I'd even speculate that's why the new cards were held back, likely the 8970 would have been slower than the titan but if these new specs are true then it should be at least as fast or if not faster. I also don't think it will be priced anywhere near a titan and knowing AMD id expect it to be priced around a 780GTX or slightly cheaper. Just speculation on my part but that would explain why AMD's cards were about to launch, then titan rumours surfaced, then AMD cards are withdrawn and 7xxx stays strong throughout 2013.
Its going to be interesting to see what AMD's response is. Will they introduce a new card or just completely shuffle the performance round of their existing lineup.
Caporegime
- Joined
- 18 Sep 2009
- Posts
- 30,522
- Location
- Dormanstown.
All they can do is launch 28nm until 20nm's capable/ready.
It makes more sense releasing 28nm than waiting for 20nm, unless 20nm isn't far off, but then Nvidia would know this too, yet they've just launched a 28nm generation.
At any rate, who knows what AMD will do.
But I'm getting bored of my 7970, and Nvidia's 780 I refuse to buy.
Last edited:
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
This simply isn't how the semi conductor industry works, IF and I mean IF that 8970 spec came out in Q3, it would have taped out, without question BEFORE Titan launched, and it would have been in the design stage well over a year ago, its likely the design work for it would have started more like 2-3 years ago. You can't see a competing chip, decide you want something bigger and have it designed, tested, taped out then in production in a few months, these are products that take 6-8 months from tape out to launch and 1-2 years in design before that.
AS for the specs, they ARE bogus, AMD would not release a xx70 on one chip design, then a xx50 version as a different chip, leaving zero room for the salvaged parts. a 420mm chip vs a 270mm chip being that close in performance.. IF AMD stripped out compute, its possible, and for 99% of people compute is pointless then its possible, if not, then they wouldn't be that close, but frankly the specs don't stack up. There would be a 8950 based on the salvaged 420mm core, because a 420mm core WILL have lower yields than a 270mm2 core, which has two salvaged parts, yet they've left zero room in the line up for the salvaged 420mm parts.... simple answer, nope.
While there are arguments IF they have different architectures for drastically different numbers of rops/tmu's based on compute needs, in general 50% more rops for less than 30% increase in shaders.
While Bonair supposedly has 7Cu's per ACE(or half of the gpu) its, odd, its likely it really does have 8 in(with 4cu's per array, a number AMD has used for ages, makes sense, is used in most of their products unless it needs even less than that(mobile Jaguar/bobcat parts). Bonair(7790) already doubled the ace's in a low end gpu from 1-2, the 7970 and 7870 only had 2 ace's, for similar shader counts going to 4 would be a huge step up, going to 8 sounds incredibly unlikely. Anyway as such the 28CU's for a "full" 8950 seems an incredibly unlikely number, as does 36 for the 8970(although IF again the architecture was different enough for the two cards that might be possible).
3 geometry engines and 8 ace's in a 8970, but 2 geometry engines in a 8950 but in a smaller part, still with 8 ace's. This is simply a no go, in general there is one geometry engine per "front end" part of the core. So if each part of the front end was split into two haves you might get 1 geometry engine and 4 ace's in the 8950, fine, but then how would the 8970 work. 2 geometry engines in one half of the front end, one in the other... would never happen. Front end split into three for a more compute heavy part, possible, but how would 8 Ace's split evenly across a three way front end, it wouldn't.
Then IF the 8970 was a more compute heavy part which is the only way some of those numbers make sense, then why would a less compute heavy part have as many ACE's as the high end part... they wouldn't. ACE's are the "command" processor for a gpu, so you might theoretically want a lot more ACE's to handle increased workload, different types of workload and handle more threads more efficiently. None of the numbers stack up at all.
If these are part of the same architecture, then you would have the same CU/TMU/ROP/Geometry/ACE ratio's in most of the cards. if they are different architectures(possible) then the high end would have more geometry/ace's/rops per shader than the "gaming" lower end chips, which those specs aren't consistent with.
Basically I see nothing in those numbers that makes the slightest bit of sense at all. Even if they did an almost identical to Nvidia thing and the 8970 is actually a salvaged part itself and a "real" full fat 420mm core is saved for professional and actually has more CU's/shaders, etc. That would infact be more likely based on these specs, but the specs still don't make sense, geometry vs ace's don't add up, high end vs midrange ace's don't add up, bandwidth/shader ratio vs high end doesn't add up.
Last edited:
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,248
- Location
- Billericay, UK
Some shady character called Gibbo who claims to work at some online computer parts store says they will lunching new cards starting in June.