PD based 8350 is just spanking the poor 1100T in 95% of the workloads(often by a very large margin) and OCs much better at the same time. Yet still we see people saying it's crap and even recommending old Phenom IIs over 8300/6300 (ignorance or they are just stubborn?).
Maybe, but this is equivalent to saying that when those first P4 chips came out, Intel should have scrapped the whole line and started over.
4FP units (more flexible though) Vs 6FPunits and still better throughput (much better if FMA is used). Slimmer integer core but still providing equal if not better performance (due to better branch prediction,scheduling,prefetchers etc.). PD concept is fine and AMD is just refining it more going to SR and EX.33% more cores and an entire process node advantage when it comes to clockspeeds and power consumption...one would hope an 8-core 32nm CPU could best its older 6-core 45nm sibling.
Doesn't mean the 8350 is superior, just given advantages that the thuban core was never afforded.
None of that hypothetical stuff is relevant though, that is fair to say. All that matters is what you (or I) can buy right now. And my 8350 is definitely a performance beast compared to a 1100T! Spent less on it too What's not to like?
4FP units (more flexible though) Vs 6FPunits and still better throughput (much better if FMA is used). Slimmer integer core but still providing equal if not better performance (due to better branch prediction,scheduling,prefetchers etc.). PD concept is fine and AMD is just refining it more going to SR and EX.
That they wouldn't have half or more of the company telling them they have an opportunity is the WTF part. Before reading that article, and doing some more checking to make sure it wasn't just that writers spin, I just assumed they had plans to have jaguar in dense server plans. Wouldn't they have those plans going all the way back to the first canceled Brazos refresh? Apparently they didn't and that's sad.
They should have, and eventually they did. Both Willamette and Prescott were "unmitigated failures".
The PIII would have been better, but still not enough to best AMD's Athlon XP and eventual K8-based X2.
Very similar to the K10 -> Bulldozer situation IMO.
Intel had enough cash, and time, to correct their error and bring Core2 to the market. I don't know if AMD has that kind of time or money to correct their bulldozer mistake.
Refining it is fine, that is the less expensive option to be sure.
But will refining it ever get them close to being on-par with Intel at whatever time they release their respective refinements?
Given the node disparity, is it realistic to expect to be at par with Intel at all?
Consequently they aren't all too keen on x86 either I take it, as in they see ARM giving more perf/watt than x86 as it stands today, but if they are able to tweak Cortex-A57 just enough to waltz Intel then we might see a stronger resurgence in the server market for AMD !
The Bulldozer (the products) is a mistake because they were released prematurely. The Bulldozer micro-architecture is not a mistake, its the best micro-architecture AMD had to this day and they have to keep pushing it all the way it can take.
It was delayed an entire node.
That's also a fallacy. Even if we grant that its the best micro-arch they've ever had, it doesn't validate its existence. Its still a failure if AMD could have made an even better design if they had gone a different route.
33% more cores and an entire process node advantage when it comes to clockspeeds and power consumption...one would hope an 8-core 32nm CPU could best its older 6-core 45nm sibling.
Doesn't mean the 8350 is superior, just given advantages that the thuban core was never afforded.
It was afforded that advantage in Llano.
Llano vs Trinity is what really rubbed in the uarch advantages(or lack there of). 4 32nm Star cores vs 2 32nm Piledriver Modules.
The Bulldozer micro-architecture is not a mistake, its the best micro-architecture AMD had to this day and they have to keep pushing it all the way it can take.
How do you determine if a micro-arch is good or not?
I realize this is a very open-ended question and can afford simple answers such as: "it does <list of things> better than <stuff>". But that's like saying car A is better than car B because it has better MPG or better handling or stuff like that. What I'm interested in is: what makes it better? Why is it better?
Those benchmarks are hardly indicative- Trinity in general outperformed Llano quite handily.
http://www.anandtech.com/bench/Product/399?vs=675
(Yes yes, that is the A8-3850 instead of the A8-3870K, but that is what there are benchmarks for in the database- and there is a mere 100MHz clock speed difference between them, anyway!) There are still a few cases where Llano wins out, but in general Trinity is a clear win.
Correct me if I am wrong, but wasn't Llano fabricated on an immature 32nm Glofo process, which badly affected yields and maybe performance? While when trinity was launched, the process should have been much more mature, so performance advantages may just reflect that trinity was built on a more mature process.
It was delayed an entire node.
That's also a fallacy. Even if we grant that its the best micro-arch they've ever had, it doesn't validate its existence. Its still a failure if AMD could have made an even better design if they had gone a different route.
And still it was an unfinished product, look how much better the Vishera is with just minor tweaks.
Dont Judge the architecture by the end-product, if you had the Vishera in 2011(against SandyBridge) then nobody would be talking about a failure today.
The performance is exceptional at things the users of the chip won't do. Seriously, there are parts of the server market where OpenCL performance will be very nice to have, but the seamicro customers are not them. What they want is something that runs PHP, Python and Ruby well at low cost, and hell will freeze over before OpenCL will either accelerate any of those tasks or the customers will expend the engineering resources to use OpenCL directly. (The decision to base your business on scripting languages is based on the idea of reducing programming costs. OpenCL costs a lot to program for.)
They should have, and eventually they did. Both Willamette and Prescott were "unmitigated failures".
The PIII would have been better, but still not enough to best AMD's Athlon XP and eventual K8-based X2.
Very similar to the K10 -> Bulldozer situation IMO.
Intel had enough cash, and time, to correct their error and bring Core2 to the market. I don't know if AMD has that kind of time or money to correct their bulldozer mistake.
The Bulldozer (the products) is a mistake because they were released prematurely. The Bulldozer micro-architecture is not a mistake, its the best micro-architecture AMD had to this day and they have to keep pushing it all the way it can take.
AMD Defense Force, Go!
It was release prematurely after what, 6 years in development?
Even AMD says it was a failure, and you disagree?