|
Because it was faster from where I was on their site at the time, and they both show 2600k not even close to being obliterated.
|
Yeah, the 2600k is just a beast.
Glad I got my 2500k.  It is interesting if you are following on xtreme systerms, someone did a proof of concept that you can get 20-30% performance boost by disabling one of the cores in each cluster to turn it into a much faster quad core over just disabling 2 clusters. He hasn't posted any gaming benchmarks yet but it is definitely interesting as you can clock higher and can remove the penalty for shared cache...
Hey, with enough tinkering and a hypothetical load balancing/throttling/core toggling software patch, it can beat Thuban! xD
Definitely a wacky chip but interesting...
|
It seems for 90% of people, the i5-2500k is better bang for buck. Better in single threaded apps, overclocks better.
Shame for AMD...love them D:
|
Here's something interesting:
http://www.xbitlabs.com/news/cpu/display/20111013232215_Ex_AMD_Engineer_Explains_Bulldozer_Fiasco.html
http://www.insideris.com/amd-spreads-propaganda-ex-employee-speaks-out/
Apparently, this ex-AMD engineer claims that part of Bulldozer's failure relates change to the design process of chips by AMD. Traditionally, AMD and Intel handcrafted their chips, but at a certain point (not mentioned), AMD decided to switch to an automated process of designing their chips. He claims that automation increases the speed of the design process (and yet Bulldozer was delayed for half a year), while its inefficiencies leads to a 20% bigger die area and 20% less performance. This explains why a 2 billion transistor CPU can't beat an older X6 1100T which has much less transistors (~900 million) on most tests. And the relatively large size of the chip contributed to its high cost.
|
On October 14 2011 22:49 android_245 wrote:Here's something interesting: http://www.xbitlabs.com/news/cpu/display/20111013232215_Ex_AMD_Engineer_Explains_Bulldozer_Fiasco.htmlhttp://www.insideris.com/amd-spreads-propaganda-ex-employee-speaks-out/Apparently, this ex-AMD engineer claims that part of Bulldozer's failure relates change to the design process of chips by AMD. Traditionally, AMD and Intel handcrafted their chips, but at a certain point (not mentioned), AMD decided to switch to an automated process of designing their chips. He claims that automation increases the speed of the design process (and yet Bulldozer was delayed for half a year), while its inefficiencies leads to a 20% bigger die area and 20% less performance. This explains why a 2 billion transistor CPU can't beat an older X6 1100T which has much less transistors (~900 million) on most tests. And the relatively large size of the chip contributed to its high cost.
I have a feeling we'll see a lot of articles feeling out who the public will accept as a scapegoat before we hear about someone getting fired. First of many.
I still think it should be the marketing guys who overhyped it, and whoever set standard pricing.
Not saying this is impossible or anything, just looking at how the world tends to work.
|
On October 14 2011 22:49 android_245 wrote:Here's something interesting: http://www.xbitlabs.com/news/cpu/display/20111013232215_Ex_AMD_Engineer_Explains_Bulldozer_Fiasco.htmlhttp://www.insideris.com/amd-spreads-propaganda-ex-employee-speaks-out/Apparently, this ex-AMD engineer claims that part of Bulldozer's failure relates change to the design process of chips by AMD. Traditionally, AMD and Intel handcrafted their chips, but at a certain point (not mentioned), AMD decided to switch to an automated process of designing their chips. He claims that automation increases the speed of the design process (and yet Bulldozer was delayed for half a year), while its inefficiencies leads to a 20% bigger die area and 20% less performance. This explains why a 2 billion transistor CPU can't beat an older X6 1100T which has much less transistors (~900 million) on most tests. And the relatively large size of the chip contributed to its high cost.
Yeah this was discussed months ago on MR too (where he was made a post with that kind of information). I'm lowering engineering expectations for the next 18-24 months, and even then, it's probably still too soon to expect any serious deviations from the engineering design decisions AMD made on BD 
Not really a fun time for builders, cuz there's nothing to debate on lol.
|
On October 15 2011 01:46 mav451 wrote:...I'm lowering engineering expectations for the next 18-24 months, and even then, it's probably still too soon to expect any serious deviations from the engineering design decisions AMD made on BD ...
Well w e did just get a new CEO this quarter... Maybe something will... is... changing?
|
You need to be realistic about the timetable to go from engineering design, validation, to a shipping product.
|
|
|
What is this, the youtube thread? That's at least the second time this has been posted here, and this time without even gracing us with a 1-liner.
Christ on a stick.
|
On October 14 2011 22:49 android_245 wrote:Here's something interesting: http://www.xbitlabs.com/news/cpu/display/20111013232215_Ex_AMD_Engineer_Explains_Bulldozer_Fiasco.htmlhttp://www.insideris.com/amd-spreads-propaganda-ex-employee-speaks-out/Apparently, this ex-AMD engineer claims that part of Bulldozer's failure relates change to the design process of chips by AMD. Traditionally, AMD and Intel handcrafted their chips, but at a certain point (not mentioned), AMD decided to switch to an automated process of designing their chips. He claims that automation increases the speed of the design process (and yet Bulldozer was delayed for half a year), while its inefficiencies leads to a 20% bigger die area and 20% less performance. This explains why a 2 billion transistor CPU can't beat an older X6 1100T which has much less transistors (~900 million) on most tests. And the relatively large size of the chip contributed to its high cost.
I realize I'm late quoting this and I'm the 3rd to do so, but I get the feeling that many don't quite get what's going on here. Actually I'm not so sure myself since I've never worked with VHDL / Verilog / whatever or any transistor design tools or digital logic design beyond the baby stuff.
A processor is logically just an insanely complicated digital logic circuit made up of transistors with various properties, in various arrangements, hooked up and laid out in a certain exact pattern. The key idea is that there are multiple ways to build something to do the exact same thing.
All the functional blocks like adders, registers, cache, etc. are made up of transistors. There should be tools for creating blocks as well as laying out parts of the design and figuring out how to connect them together.
For best electrical characteristics (maybe this may effect possible clock speeds and power consumption), fewer gates to pass through so less delays, using the fewest amount of transistors, and so on, it's best to optimize the layout and construction of elements by hand. No automatic tools are perfect.
It's very roughly like writing C code, compiling, and then doing hand optimization of the resulting assembly code to make it a bit faster. And mostly C is just used for computational efficiency to begin with.
Obviously you still have to do a lot of stuff manually and plan things out, even if you're using some automatic tools to help with some of the design. There's just the question of how much skimping is going on, and how much difference that's really making.
I think the comment may be as much a critique of the culture shift as much as of the technical merits of taking such shortcuts.
|
Maybe that ex-AMD engineer is trying to make the process sound bad because he got fired from a job doing the basic circuit design?
|
|
|
On October 15 2011 06:59 JingleHell wrote:
Maybe that ex-AMD engineer is trying to make the process sound bad because he got fired from a job doing the basic circuit design?
Nah, a chip that uses DOUBLE the number of transistors of the i7 should be able to get somewhat close in performance. But instead it's far, far behind and just uses a shitload more power. Sounds exactly like the type of crap that automated tools generate.
I sincerely hope AMD's new CEO kicks this practice out the door, because if they don't recover quickly, AMD is gonna be on its way out, which is bad for consumers... Intel has already shown its willingness to squeeze competitors out of the market and charge exorbitant prices. Imagine what it will be like if they have a true monopoly?
|
I know how they can save this situation:
AMD, the new Mac!
|
On October 15 2011 08:32 ddrddrddrddr wrote:
I know how they can save this situation:
AMD, the new Mac!
Kinda hard to market something as supposedly elegant and user-friendly when it's plugged into a motherboard 24/7.
|
|
|
AMD always have the best code names!
"QUADFATHER"
"Interlagos cores"
"Thuban"
That's sadly about all they've got going for them in my book though...
|
How are you gonna post quadfather without an accompanying image...jeez 
+ Show Spoiler +
|
It's interesting, if you want to buy a new socket motherboard just for the sake of getting a slightly faster athlon without L3 cache.
Any sane person would still go with a sandy bridge if they have to change motherboards.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/GcOiB.jpg)
