5nm is possible on siliconI thought 7nm is not possible on silicon? Are AMD using something else for 7nm?
5nm is possible on siliconI thought 7nm is not possible on silicon? Are AMD using something else for 7nm?
Yes perhaps, but they would have been better off doing that then the flawed idea that was bulldozer, the concept was interesting but severely flawed, as ST performance was about equal to phenom 2, the module was massive vs intels HT core, so much so they marketed it as 1 module = 2cores, when in originality CMT was supposed to be a direct rival to HT (1 module =1 intel core), just more efficient.
That would be 2-3 years. Not 4 though. I see better Ryzen on the horizon (lol). I may wait for improved Ryzen but I'm afraid of price hikes.actually I think it goes like this:
2018: Ryzen+ on 14nm
2019/2020: Ryzen 2 on 7nm.
Yeah CMT vs. SMT was really a none issue with the exception about how its implementation destroyed SC performance. BD was never going to do well if they couldn't get close to the I series processors in performance. I mean when people could get better performance out of i3 than their highest speed 8c beast in 90% of the use cases. There was no chance.AMD just might have done better marketing a whole module as "1 core with super HT". It would still leave the issue of single thread performance on the table of course.
I am not sure that logically follows. Faildozer didn't do them any favors.
Problem is a module was huge, intel would have had more real cores vs better SMT with intel as well as better ST perf.AMD just might have done better marketing a whole module as "1 core with super HT". It would still leave the issue of single thread performance on the table of course.
That's what they did. Amd started planning zen a couple of months after bulldozer.What I don't get is, okay, Bulldozer happened and it was a disaster, even AMD knows this.
But then why stick with the plan to release things like Steamroller and Excavator? 6 months before release, AMD would have known that Bulldozer was a dud. But why continue development on the idea? Why not abandon it, when it was clear that it was not working?
Why did it take Jim Keller to force AMD to admit that Bulldozer was a load of sheet? I mean, surely they already knew that, so why not start designing a competent CPU architecture as soon as Bulldozer came out? They could have released something a lot better than Steamroller and Excavator if they had.
It ran so hot that it forced AMD to develop heat and power usage optimizing technologies that they can deploy in Zen. If Bulldozer had run cool and power efficient, AMD would not have needed to develop advanced power saving technology.
Because they had no other solution. They delayed BD enough and even pushed back the other two for for more tweaks. But you can't just snap a finger and have a new CPU architecture. People who think that have problems imagining the development process. They hired back Heller and started developing a new core immediately. They wouldn't have had any chance of catching Intel if they pushed this out sooner. They would be completely dead if they just stopped working on BD.What I don't get is, okay, Bulldozer happened and it was a disaster, even AMD knows this.
But then why stick with the plan to release things like Steamroller and Excavator? 6 months before release, AMD would have known that Bulldozer was a dud. But why continue development on the idea? Why not abandon it, when it was clear that it was not working?
Why did it take Jim Keller to force AMD to admit that Bulldozer was a load of sheet? I mean, surely they already knew that, so why not start designing a competent CPU architecture as soon as Bulldozer came out? They could have released something a lot better than Steamroller and Excavator if they had.
They replaced the youngish guy with the punchable face?
Heh, works thanks. Don't know why I didn't try that.
Problem is a module was huge, intel would have had more real cores vs better SMT with intel as well as better ST perf.
It was a futuristic fancy idea for a MT future utopia with the gpu taking the FP slack, but the whole module concept was flawed and not just the shoddy execution, thats why we are not seeing a brand new CMT design on 14nm, the concept just Wouldn't work even in todays heavily MT environment and DX12 future.
It would have been interesting if they had designed zen with CMT ground up, same idea just with super wide modules with massive execution resources per core/thread, a brainiac design rather than a speed demon.
Saying that, its easy to judge in hindsight, credit to AMD for sticking their neck out and trying something new, incredibly brave with their resources, its better to try and risking falling than not try at all, afterall the unexpected upside to the failed bulldozer was AMD gained valuable MT expertise which have enabled them to overtake the masters in SMT technology - no mean feat considering the 15 year experience with HT and exponential resources intel has.
Not to mention AMD has had to use some innovative backs to the wall engineering to improve bulldozers efficiency through the iterations, essentially polishing a large steaming hot turd into something that turned into very power efficient with excavator.
AMD has carried this expertise and experience into zen and now have produced a perf/watt perf/mm2 and MT marvel, even on an inferior less dense and immature process, miracle!
All clouds have a silver lining ay?
Yeah, the Phenom II processors were much much better and they were overall competitive. They could have revamped the architecture and get more out of it, but I guess they just made the wrong decision and by the time they realized how colossal of a mistake they made it was too late to turn back.
Select the posts you want, scroll down to the compose box and click on "insert quotes."I have never gotten + Quote to work on this forum. I only realized yesterday i can hit reply mulitple times to get all the original posts into my reply.
Thanks guys i am learning a lot from this thread.
Node names and feature sizes are more or less completely decoupled at this point. A "7nm" process is more a collection of fabrication features than any reference to what's happening on a real chip.I thought 7nm is not possible on silicon? Are AMD using something else for 7nm?
Problem is a module was huge, intel would have had more real cores vs better SMT with intel as well as better ST perf.
It was a futuristic fancy idea for a MT future utopia with the gpu taking the FP slack, but the whole module concept was flawed and not just the shoddy execution, thats why we are not seeing a brand new CMT design on 14nm, the concept just Wouldn't work even in todays heavily MT environment and DX12 future.
I first mistakenly read this as "a garbage company called Microsoft" which gave me a good chuckle.IBM's first choice, but couldn't reach an agreement with Gary Kildall. And when he missed a meeting for negotiations IBM said 'screw it', and contracted a garage company called Microsoft to make what was essentially a clone of CP/M
I think Steamroller would have been the better starting point as an 8 core flagship what if scenario. Excavator had such a heavy focus on power management and lower watt mobile applications. I think it would have been designed very differently if Steamroller was the starting point and they were a little more competitive. But I'd still like to know how an 8 core XV with L3 would perform even now.Now if XV had been the starting point for the Bulldozer architecture, who knows how it would have played out...?
They would have been better improving phenom 2, imagine how many hundreds of millions were spent on bulldozer and derivatives?
Imagine what phenom could have been had they spent half of that dev cost on phenom 3, SMT, better caches, wider core, new process etc etc.
Would have been better imo.
Select the posts you want, scroll down to the compose box and click on "insert quotes."
“Designing microprocessors is like playing Russian roulette. You put a gun to your head, pull the trigger, and find out four years later if you blew your brains out."
Yes your right, there was another project cancelled, also i think i read somewhere that bulldozer was an old/long running project, something like 8 years or something, someone told me yesterday at SA forum that bulldozer originally was planned to have another alu per thread but was decided it was not needed!I can't imagine what Phenom could have been like and neither can you. Maybe the design was tapped out and they couldn't get any more out of it. Obviously AMD didn't think it could be improved enough.
Where you around at the time? AMD cancelled several projects between Phenom and Bulldozer, one of which may have been Phenom III.
They did provide some evidence, although nothing finger print like.Don't listen to what you read on SA forums. Excavator was the end of the line, it was a filler CPU until Ryzen. There was not a plan to upgrade Excavator.
Excavator had such a heavy focus on power management and lower watt mobile applications. I think it would have been designed very differently if Steamroller was the starting point and they were a little more competitive. But I'd still like to know how an 8 core XV with L3 would perform even now.