Discussion Forums > Technology
Your view on AMD's Bulldozer
nstgc:
--- Quote from: bloody000 on October 17, 2011, 09:11:42 AM ---
--- Quote from: fohfoh on October 17, 2011, 02:27:27 AM ---Holy fucking spirits batman... iindigo is alive!
--- End quote ---
I thought he got ***** by those tentacle monsters in Nihon.
--- End quote ---
Tentacle monsters don't rape men (well 98% of the time they don't).
ColdFission:
Found an interesting post from HardOCP with the point of view of former AMD engineers:
http://hardforum.com/showpost.php?p=1037482638&postcount=88
And some interesting write-ups from Charlie of S|A:
http://semiaccurate.com/2011/10/17/why-did-bulldozer-underwhelm/
Much of the first half the article explain why AMD needed to hurt more in 2010 to hurt a little less in 2011 by pulling the plug on 45-nm Bulldozer as it was sucking up too many resources to do further designs on 32-nm. The part I quotes was quite interesting to me.
--- Quote ---If you look at some of the details, a few things stand out. First, the cache latencies as measured by Michael Schuette in the Lost Circuits article, are in a word, horrific. L1D caches are 1 cycle slower than Phenom/Stars, and the L2 cache is 25-27 cycles, 10-12 slower than it’s predecessor. It may be twice as large, but it is shared by two cores, and is almost twice as slow. Sharing of resources may have some very nice benefits, but in this case, the downsides are, well, crippling.
It is hard to fathom how such high latencies were tolerated, much less put in to production. While it is unlikely to be possible given the shared front end, a few more transistors burned to split this 2MB L2 cache in to 2 1MB L2 caches would have halved latency. This alone would have a marked increase on performance, and not a trivial one, even a single shared 1MB cache with a 15 cycle latency would likely be an improvement over what we ended up with.
During the press briefing in late August, the question of cache latencies came up, and AMD didn’t give an answer, but said they would get back to us. Given the number of people in the room with business cards that read architect, that lack of an answer did send up serious red flags. The official numbers never came for some reason, and now you know why. Cache latencies are such a massive screw up that it is hard to put in to words. Consider the cache latencies a handful of cuts, not just one.
--- End quote ---
http://semiaccurate.com/2011/10/17/bulldozer-doesnt-have-just-a-single-problem/
Lupin:
--- Quote from: ColdFission on October 17, 2011, 11:36:05 PM ---Found an interesting post from HardOCP with the point of view of former AMD engineers:
http://hardforum.com/showpost.php?p=1037482638&postcount=88
--- End quote ---
Those macforum posts have been used to death. Response from someone still in AMD: http://www.realworldtech.com/forums/index.cfm?action=detail&id=123260&threadid=123240&roomid=2#CURRENT
--- Quote from: ColdFission on October 17, 2011, 11:36:05 PM ---Much of the first half the article explain why AMD needed to hurt more in 2010 to hurt a little less in 2011 by pulling the plug on 45-nm Bulldozer as it was sucking up too many resources to do further designs on 32-nm. The part I quotes was quite interesting to me.
--- Quote ---If you look at some of the details, a few things stand out. First, the cache latencies as measured by Michael Schuette in the Lost Circuits article, are in a word, horrific. L1D caches are 1 cycle slower than Phenom/Stars, and the L2 cache is 25-27 cycles, 10-12 slower than it’s predecessor. It may be twice as large, but it is shared by two cores, and is almost twice as slow. Sharing of resources may have some very nice benefits, but in this case, the downsides are, well, crippling.
It is hard to fathom how such high latencies were tolerated, much less put in to production. While it is unlikely to be possible given the shared front end, a few more transistors burned to split this 2MB L2 cache in to 2 1MB L2 caches would have halved latency. This alone would have a marked increase on performance, and not a trivial one, even a single shared 1MB cache with a 15 cycle latency would likely be an improvement over what we ended up with.
During the press briefing in late August, the question of cache latencies came up, and AMD didn’t give an answer, but said they would get back to us. Given the number of people in the room with business cards that read architect, that lack of an answer did send up serious red flags. The official numbers never came for some reason, and now you know why. Cache latencies are such a massive screw up that it is hard to put in to words. Consider the cache latencies a handful of cuts, not just one.
--- End quote ---
http://semiaccurate.com/2011/10/17/bulldozer-doesnt-have-just-a-single-problem/
--- End quote ---
Follow the thread for that article in S|A's forums.
kitamesume:
what i`m guessing is they'll gonna gamble on the 22nm die shrink with a few more touches to further reduce the power consumption and further increase the clock speed.
i think i got my guess right, right?
kamuixtv99:
--- Quote from: per on October 16, 2011, 11:25:37 AM ---
--- Quote from: kamuixtv99 on October 15, 2011, 12:55:01 PM ---You get what you pay with AMD, They have poor benchmarks but with its price, it's still the 'king' for massive buyers like for schools and internet cafes. These days what you see inside internet cafes are facebook and its lousy games and WoW :P
--- End quote ---
Unfortunately bulldozer is priced higher or the same as the competing intel parts. And the Intel parts use about a third of the power, which saves you a surprising amount of power over the years.
What I am most irritated about is that their server CPU section has been promising us lower power and more performance, the the truth is actually that their old server CPU:s are more efficient, and faster!
However, those are still not competitive with the Xeon CPU:s.
--- End quote ---
You get what you pay with AMD. Yes, that was before Bulldozer. From Athlon/Sempron to X2 I guess. Like those days if you want to do something (video editing, video conversion) other than games then get an Intel by spending $50 to $100 more (In my place $50? hmmm you need to work for 42 hours to earn that.). I only know a little about Bulldozer I'm just getting info from you guys. This is about to see when it reaches to our shores.
--- Quote ---Yes, another one who thinks anyone that doesn't care about computers must be a mindless drone. You are not above "them" just because you care about a few pieces of silicon, kamuixtv99, maybe you will understand that when you find out that you're an "average joe" by shampoo enthusiast standard.
--- End quote ---
Sorry. "average joe" was a wrong term. It's the people who rely on brand name, not sure with the term. Knows a little or none at all on the performance.
Anyways, like what you guys said. AMD is no longer the same as before. A lot of people resigned, change of management and etc.
Good thing I have no plans to upgrade my rig not until something will replace Blu-ray. Yes, Bulldozer may not be the right choice for now.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version