Caporegime
Oh, I know that they'll sell, maybe not as much to the enthusiasts compared to Phenom II's however.
-
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.
AMD Bulldozer Finally!
- Thread starter AceTK
- Start date
Associate
- Joined
- 13 Nov 2007
- Posts
- 1,483
If they did that significantly I'd probably buy one ... my 920 rig is heading for home server duty and looking for new kit to replace it.
I think he means produce them faster, not up the clock speed
Yeah the performance is good enough that the uninformed (of which there are a huge amount) will lap up the 6 and 8 core models.
Yes...sorry...I was too ambiguous.
They need to manufacture them at a faster rate, to increase supply.
Lack of supply is their main problem at present, because there are clearly plenty of people willing to buy them, at current (inflated) prices.
Soldato
to manufacture them faster, they need to stop messing around with the insane die-size, make a one with two modules and no pointless four Hyper Transport connects and they'll get much better yields, plus better thermals and power consumption one would imagine.
Edit: also don't know if anyone is interested by the transistor count was wrong all along, the transistor count for Bulldozer (8***) is 1.2B rather than the insane 2.0B that was around. so in that respect it seems like much more competitive vs. Intel processors, since a lot of the anti-Bulldozer arguments were based around transistor counts and die-size, still got no idea why the die-size is so big, could be something to do with Global Foundries issues?
Edit: also don't know if anyone is interested by the transistor count was wrong all along, the transistor count for Bulldozer (8***) is 1.2B rather than the insane 2.0B that was around. so in that respect it seems like much more competitive vs. Intel processors, since a lot of the anti-Bulldozer arguments were based around transistor counts and die-size, still got no idea why the die-size is so big, could be something to do with Global Foundries issues?
Last edited:
AMD Shifts Course from Intel Rivalry
http://www.pcworld.com/article/245365/amd_shifts_course_from_intel_rivalry.html#tk.nl_dnx_h_crawl
http://www.pcworld.com/article/245365/amd_shifts_course_from_intel_rivalry.html#tk.nl_dnx_h_crawl
That's still roughly equal with Intel's much faster (but more expensive) Gulftown series.
So, just to clarify, are you suggesting that AMD go back to the drawing board and redesign the CPU? IF this is the case, surely this would be an absolute (financial) disaster as they will have to create an all new CPU, then run simulations, then instruct the CPU factories of the new CPU design. To do all this takes a lot of time and certainly is not a quick fix. It may end up taking them another year to make all these modifications, by which time, Intel will have brought out Ivybridge. By this time, AMD will not longer be competing against Sandybridge the even faster Ivybridge.
The way I see it, is that AMD have come too far now. They have to persist with their new CPU design and see it through.
Soldato
no, not at all! the design by its very nature is modular, the whole idea is so one can add and subtract modules without any real effort, i.e. no need to go back to the drawing board. so why they don't show off the merits of their M-SPACE modular design by simply removing two of the modules and making a two module chip instead of using all four module (some with one disabled, some with two) to improve their margins makes no sense.
also 1.2B transistors isn't too bad, sure its still a load but then Bulldozer with its complicated yet simplified design was always going to be quite heavy on the transistor count. but the count is still not as bad as one would think, it has eight integer cores, four complicated floating-point units and an absolute ton of cache, it only has ~300 million more than Thuban, so it hasn't worked out too badly in that respect.
the four Hyper Transport connects have no use in the desktop space either, so if the design is indeed modular, then take them off and stop adding pointless things to an already large die, as far as I know those connects are more for the server side of things rather than desktop.
the four Hyper Transport connects have no use in the desktop space either, so if the design is indeed modular, then take them off and stop adding pointless things to an already large die, as far as I know those connects are more for the server side of things rather than desktop.
But then surly the argument comes in where it would cost more to manufacturer two different products.
I am sure it is cheaper for them to just disable modules than it is to produce a different CPU without the module.
I agree that by doing it the way that they are doing it now, that they are wasting resources. at the end of the day though it all comes down in what is cheaper for them to produce
Soldato
Has any one tested Win 8 gaming performance with a Bulldozer?
Is that even possible at this early stage?
Is that even possible at this early stage?
Caporegime
If you use the Windows 8 preview I suppose you could.
I was interested to note that the 2 billion transistor figure for the 8xxx Bulldozer line was out by 800m. It's apparently 1.2billion ... so it's an awful lot more 'efficient' than many feared. With future revisions and GF (hopefully) sorting out the process tech, it may well prove to be a very sound architecture in the long term.
Caporegime
What'd you mean by more efficient?
Performance per transistor, yes, we can no longer say "2 BILLION TRANSISTORS FOR WHAT?"
But it now means that the die isn't all that small, being less dense than SB, and only amplifies the heat/power concerns.
I know AMD are getting peeved with GF, some article about them taking their APU's to TMSC, who knows if that's true like.
Although, I'm still sceptical, they've ONLY now came out with saying it's not 2 Billion transistors?
Last edited:
Soldato
surely the lower transistor density further points toward problems on Global Foundries side of things? know AMD do some stupid things sometimes, and god they do, but to design a new chip from the ground up to have some of the problems it has whilst keeping in mind AMD actually know a thing or two about processor design, as much as people choose to forget.
lots of things about Bulldozer don't make much sense for it to be AMD problem, performance wise its not at all bad, not great at gaming but equally not terrible in the same respect that Phenom II isn't terrible for gaming. but the low transistor density and high die size, thermal problems and power consumption all point toward Global Foundries more than AMD themselves.
with regards to removing pointless component and modules from the die, for the sake of argument lets assume that each 4*** die would be less than half the size of a 8*** die perhaps, how is it not in their best interests to make a dedicated 4*** line rather than use crippled 8*** series? perhaps AMD should consider a bold move, keep 8*** production and move 4*** series to TMSC instead?
lots of things about Bulldozer don't make much sense for it to be AMD problem, performance wise its not at all bad, not great at gaming but equally not terrible in the same respect that Phenom II isn't terrible for gaming. but the low transistor density and high die size, thermal problems and power consumption all point toward Global Foundries more than AMD themselves.
with regards to removing pointless component and modules from the die, for the sake of argument lets assume that each 4*** die would be less than half the size of a 8*** die perhaps, how is it not in their best interests to make a dedicated 4*** line rather than use crippled 8*** series? perhaps AMD should consider a bold move, keep 8*** production and move 4*** series to TMSC instead?
They're already supply constrained enough as it is .... making several separate production lines which are unnecessary would pretty much put them out of the desktop processor market, at the moment.
It's something they could consider once the Gulf money for GF eventuates in more production facilities.
TSMC don't have any suitable processes for the bigger desktop CPUs ... besides, they have horrific problems with sub 40nm themselves, and will have zero spare capacity for the foreseeable future.
It's something they could consider once the Gulf money for GF eventuates in more production facilities.
TSMC don't have any suitable processes for the bigger desktop CPUs ... besides, they have horrific problems with sub 40nm themselves, and will have zero spare capacity for the foreseeable future.
Soldato
been Advanced Micro Device' problem for many years now, no matter how competitive their processors are they can never get enough of them into the market to knock Intel off their perch. their best shot was back in K8 days when they had Prescott by the balls, but the fact that everyone knows who Intel are and almost nobody seems to know who AMD are emphasises the problem even more, asked ten random friends the other day 'know who AMD are...?' one person knew...ONE! and they all knew what Intel is, that speaks volumes about AMD's problem!
lack of supply and masses and masses of Intel buyers make it a frankly impossible situation for AMD to prevail in! doesn't matter if Bulldozer B3 comes along and knocks seven kinds of crap out of Sandy Bridge, it won't make a bit of difference in the grand scheme of things! for example Trinity will be better than Intel, no doubts about that in its target market, but will it make a difference? and will it be plagued by AMD supply problems?
lack of supply and masses and masses of Intel buyers make it a frankly impossible situation for AMD to prevail in! doesn't matter if Bulldozer B3 comes along and knocks seven kinds of crap out of Sandy Bridge, it won't make a bit of difference in the grand scheme of things! for example Trinity will be better than Intel, no doubts about that in its target market, but will it make a difference? and will it be plagued by AMD supply problems?