From X-bit Labs: http://www.xbitlabs.com/news/cpu/di...ulldozer_Chips_Incoming_Details_Revealed.html
I was just over at Anandtech looking at the bench comparison with calculator in hand. The 2600K and the i7980 are both very close to 20% faster than the 970 and the x6 1100. If this "50% faster than the Phenom II is a true statement, that makes bulldozer 15-17% faster than Intels fastest offering. I think the clock speeds of Ivy just went up.
G, I am really holding off a new build until I can see just what these Bulldozer chips will offer.
And you just may be right about the Ivy clock speeds going up.
By the way, I've read some good things about AMD's latest Fusion offering in their E-350 chip. I'm not a netbook fan but for a person that wants one Lenovo has a nice model utilizing it.
Sounds promising, but then the BD architecture always has.
A couple of things that I would place a question mark beside would be:
"Phenom II-series" covers a quite a range of performance, as it does models. It's probably optimistic to equate the phrase with AMD's top binned CPU in isolation.
Would indicate either:
A very slow ramp (or halo-esque pricing) assuming CPU means CPU and not APU. If 90% of AMD's production is still on 45nm at the end of the year.
A slow road to profitability if CPU actually means CPU/APU. If 90% of the production is still split between Phenom II/Athlon II (basement margins) and Zacate/Llano (low margin which also cannibalizes low-end graphics card sales).
As for Ivy Bridge, I would suggest that Bulldozer's more immediate probable adversary is the Sandy Bridge-EN six and eight cores (Core i7 29xx ?) and the Patsburg (X68/LGA2011) platform-at least for the 8-core BD (I'd assume that 6/4 core would be aimed at SB/IB)...Unless of course AMD are planning on pitting their enthusiast CPU's against Intel's mainstream line in pricing. Hopefully this is the case -speaking from a consumer standpoint, but I would imagine Intel has some wiggle room on 2500K/2600K (and probable future 2700K/2800K) pricing in that eventuality, although I don't think this stance would do anything to raise the public perception of AMD's place in the CPU heirachy.
It will be interesting to see how Intel's known architecture, added to quad channel memory, QPI overclocking and probable (up to) 20Mb L3 cache stacks up to AMD's all new-all-singing-all dancing offerings and their final clocks and turbo profiles....both in performance and overall platform costs.
If G's 15-17% performance advantage holds true, then performance between BD and IB should be fairly similar.
EDIT: Scrub the Ivy Bridge/Sandy Bridge-EN blurb. If the Xbit article is valid , then so too likely is this. AMD don't seem to be looking any further ahead than the present market ( Core i7 970/980X/990X excepted).
Yep, Thats what I was thinking, that IB would be a 10-20% performance boost over SB. AMD seems to be saying the same thing, and not doing their usual lowering of expectations as per the usual for the better part of the decade. It will be exciting if they are actually on to something here.
I thought that as well, but probably just smart marketing. IB is an unknown entity performance wise, going after and referencing the i7's puts things in context for the purchasing public. They have made many statements about BD being much faster than the i7 9xx however. I think we are to extrapolate that out to any newer CPU's they (intel) have coming down the pipeline.
Not too sure about the smart marketing.
We got the same sort of thing with Barcelona, and since it's introduction, AMD's share of the server market has shrunk from ~16% to around 6-8%.
We also recently had the HD 6970 thrashes GTX 580 hyperbole. I can't help but think that leaking these so-called "performance gains" set up AMD-philes for a fall, and while I would contend that the series was a success, the undercurrent flowing through the tech community was/is that of a tarnished release and an unfulfilled promise.
So while I'm optimistic that we get back to the old Athlon/Pentium days of true competition, I can't but help think that AMD is yet again painting itself into a corner to some degree. It's rabbit out of the hat time, and I'd really like to see how AMD can make concrete the claims that they can win in performance, price, performance/$ and performance/mm² while getting significantly fewer dies per wafer, 5+ years of R&D to amortize and the not insignificant matter of wrestling OEM's away from Intel Core i -given that AMD seem to have been less than successful in outselling Intel when the latter had nothing to sell.
Intel's strategy seems to be "walk softly and carry a big stick" of late. Sandy Bridge is, at best, 20% better than Lynnfield but look at the furore the launch caused and the momentum it generated- in no small part from the lack of preconceived performance targets the parts had to hit - indeed, for the enthusiast community, quite the opposite thanks to the news that base clocks were effectively locked. For all the hoopla surrounding it's impending launch, very little information regarding relative performance was "leaked" prior to it's official introduction.
I don't know if it's smart marketing to:
Fire your CEO and two other high ranking technical officers at the same time as you launch product...
Release a card series (HD6970/6950) with spotty driver performance/functionality...
Release a card that isn't covered by warranty if you use it in the manner that it is designed to be used
So I'm reluctant to annoint AMD's marketing as anything other than mediocre based on recent efforts.
well this of course is with the caveat that they have the goods to back it up. Like I said I have noticed that they seem to be making definitive statements and not doing the lowering of expectations parsing. well...I'm hopeful about this release. for one, necessity is the mother of invention, and they have no choice but to succeed and become competitive. If they don't after milking the same architecture for the better part of a decade...they will become a Wiki entry right beside Cyrix/IBM.
No doubt this next six months will determine how AMD fare and are viewed in the heirachy. I doubt whether Zacate/Ontario/Bulldozer/Llano will "fail". I think they can all be termed good or great products, but even great products can sometimes be underwhelming if the expectations are set too high. If the talkingheadspeak says BD is as good as SB/Westmere then the public expectation will automatically be unfulfilled if BD fails to meet any of the parameters. There is also human nature to contend with- If AMD say BD is SB/Westmere's equal then a segment of the potential buying public are going to need BD to exceed Intel's offerings for their enthusiasm to be warranted, since being SB/Westmere's equal has now become the minimum requirement.
I'm pretty sure I could fill a whole page with articles, forum posts and assorted speculation links just concerned with the Xbit info, and virtually all of it now centred around how badly Intel is going to get caned in every metric concerning CPU performance and production. From Cinebench 11.5 to Global Foundries (supposedly) far superior yeilds. Apparently the 10% figure is now being translated to "GloFo's 32nm process has such high yeilds that they get more usable dies per wafer than Intel's 22nm process, so AMD can undercut Intel's pricing"....and this is from mainstream forums, not just the tinfoil hat wearing fantasy island dwellers.
I agree, there is a PR tightrope to be walked here, and it can work to hand AMD their *** in a paper bag if they do it right...unless the BD is so good as to blow SB & IB out of the water...wanna take bets on that ?:haha:
I am going to make a prediction and say that we see AMD television advertising this year! hows that for going out on a limb?
As to whether BD blows SB and IB out of the water- it should be a given that an 8-core would beat a 4-core with HT. So on a benchmark metric BD should come out on top, assuming that BD is running at 3.5GHz base -which is where the 50% increase over Phenom II originates.
How it would contend with a probable 2700K (3.5G) and 2800K (3.6G) might bear closer scrutiny bearing in mind that Intel have left the designations vacant ( LGA 2011 I think starts at i7 2900K). On a level playing field -8 core BD versus 8 core Sandy Bridge-EN I think is a mismatch. I haven't seen anything to indicate that BD is a match for Intel on a core for core, clockspeed for clockspeed CPU.
The other metrics that BD needs to blow SB/IB out of the water are that 4-core BD needs to retail at $210-220 to 330 (2500/2500K & 2600K pricing) -or less- retail and considerably less for OEM's. Assuming Intel don't institute price cuts (not a given at this stage).
If I were taking a wager on whether BD is going to comprehensively beat SB/IB/Westmere then I'd have to hedge my bets:
Performance should be similar (I don't think "blow them out of the water" applies unless you're looking at winning every benchmark, or the majority of benches by say 30-50%)...
Pricing -who knows. Price too high and P67/Z68 looks appealing- much as it does when compared with X58. Price too low and AMD have effectively turned Athlon II/Phenom II stock (and the resell market) into dogfood...
Performance per watt - well BD needs to "beat" SB by 32% outright (125w vs 95w) at the top SKU for that to happen.
Performance per mm² - BD needs to "beat" SB by 39% outright (assuming BD at ~300mm² to SB's 216mm²)
Gross profit margin - Either GloFo's 32nm offers significantly higher yeilds than both Intel's' 32 and 22nm process then it seems extremely unlikely that the larger die BD is going to win here.
Can Bulldozer win in any of those metrics?. No problem. Can it win all of them? No.
One thing is for sure, AMD are more competitive will BD in the lineup - I just wouldn't be hanging the bunting and declaring a winner just yet....which is effectively what AMD is doing now.
That was in jest you know. I don't think there will be a blowout of any sorts (maybe specific catagories) but not an outright beatdown. If for no other reason than these guys are getting very close to the IPC limits with the current architecture type, size reduction limit, and materials. One more gen of CPU's and AMD will be forced to do advertising on TV!
Check my physics theory here, but, I don't actually think that is possible if what I said above is correct.
I will try to find the article I was reading prior to my boot loader meltdown, but I had read that the AMD 32nm yields were phenomenal. Anticipated pricing I am seeing would be consistent with that.
As usual, It will be fun to watch the guys still running P4's and Thunderbirds duke it out!
Add in all the rest too. If it were not for the interest factor in seeing how long it takes AMD to change it's name to Abu dhabi Mubadala Developments (or similar) -a timeframe that probably hinges on AMD's shareprice/market cap/bulldozer/llano numbers- I could care less. I doubt many people could identify the CPU (or probably chipset either) if you sat them down at a computer and let them go about their usual tasks -gaming/encoding/wp/browsing/office productivity etc. I don't see a whole lot of people running a Westmere Xeon (or 2), X58/5520, custom hand-binned DDR3 and PCIe hardware RAID card/Revodrive system- and quite frankly if desktop system performance with the lowest error rate means so much that a CPU with a slightly better Sysmark or SuperPi becomes cause for rejoicing, then thats exactly the system those elated people would be running. Everything else is the best of the worst- my own systems included.
That's great, they all look like my first PSE project.
Yup, obviously mine as well, but the fun is to listen to them tell you why that .00000016 sec makes my system suck! You never hear them say they have the 6 core because they have enthusiast OCD.....like me
The OCD can't be that bad. You're obviously still working on the performance/$ metric, not performance at any cost metric. HA! AND YOU CALL YOURSELF AN ENTHUSIAST!!!
Hey, it's not as if your children cant pay for their own college education you know.
ROFL, just remember....I handed you that one!
My god, I learn so much on this forum just reading you two commenting to each other.