12-28-2016, 11:52 AM
Thanks for all the links SteelCrysis.
I am really am surprised by the info. Zen seems better than I would have guessed. I am though, kind of afraid to let this info excite me. Its supposed to be an engineering chip that is clocked at 3.1 ghz with very little boost. This would mean very substantial ipc gains. There is a lot left to ponder. Like, what is the clock speed characteristics and limitations of their 14nm node. As this is a very huge element in how Zen may compete.
The data presented is very positive. And the gap in IPC appears to be closing. What a great thing that would be. I must still admit that i am not without reservations. I am fearful of hype being built up. These are limited test and its all unofficial to boot. I would hate to see hype once again come back to bite another one of AMDs big and long awaited releases. So i choose to try to contain all of that until there are some real and official data.
See, even if these are real benches in a legitimate environment (where nothing is artificially gimped or scales purposefully tipped), there are things that must be considered. Will the new chip be consistently capable of distributing load across cores in all cases? Will there be stalls and wasted cycles? How about prefetching, cache, and branch prediction?
Its one thing to do well in canned multicore test or a very predicable data crunch that deals with blocks of data that break down evenly in a perfectly predetermined manner. But to make use of many cores is much more difficult in random work loads.
There are so many areas where intel has been making improvements that dont shine so much in multicore benchmarks dealing with large chunks of static data that is evenly broken down and totally predicable. With such a small set of benchmarks here and the lack of specific results, its just so limited and impossible to see any detail. And that is one of my larger concerns here. Why is that so? Why do we have these results and why are they so vague? Jumbled together to offer no resolution. Just some averages of averages
i dont know.
So, yeah. I am truly starting to feel excited for zen and Amd. But, i feel have to keep my excitement down. I guess it wont be a bad thing to hold back my expectations. If its great, i will still have to wait. We all have to.
But the seeds are now planted. And its gonna spread all over. RyZen now has bigger shoes to fill. People are now declaring Zen to no be a Sandy or Ivy like IPC. Its now being touted as better than Haswell on many forums. And its gonna have to meet that mark. Which i truly hope its at least close..cause thats more than i ever expected. if its better, bring it
I am really am surprised by the info. Zen seems better than I would have guessed. I am though, kind of afraid to let this info excite me. Its supposed to be an engineering chip that is clocked at 3.1 ghz with very little boost. This would mean very substantial ipc gains. There is a lot left to ponder. Like, what is the clock speed characteristics and limitations of their 14nm node. As this is a very huge element in how Zen may compete.
The data presented is very positive. And the gap in IPC appears to be closing. What a great thing that would be. I must still admit that i am not without reservations. I am fearful of hype being built up. These are limited test and its all unofficial to boot. I would hate to see hype once again come back to bite another one of AMDs big and long awaited releases. So i choose to try to contain all of that until there are some real and official data.
See, even if these are real benches in a legitimate environment (where nothing is artificially gimped or scales purposefully tipped), there are things that must be considered. Will the new chip be consistently capable of distributing load across cores in all cases? Will there be stalls and wasted cycles? How about prefetching, cache, and branch prediction?
Its one thing to do well in canned multicore test or a very predicable data crunch that deals with blocks of data that break down evenly in a perfectly predetermined manner. But to make use of many cores is much more difficult in random work loads.
There are so many areas where intel has been making improvements that dont shine so much in multicore benchmarks dealing with large chunks of static data that is evenly broken down and totally predicable. With such a small set of benchmarks here and the lack of specific results, its just so limited and impossible to see any detail. And that is one of my larger concerns here. Why is that so? Why do we have these results and why are they so vague? Jumbled together to offer no resolution. Just some averages of averages
i dont know.
So, yeah. I am truly starting to feel excited for zen and Amd. But, i feel have to keep my excitement down. I guess it wont be a bad thing to hold back my expectations. If its great, i will still have to wait. We all have to.
But the seeds are now planted. And its gonna spread all over. RyZen now has bigger shoes to fill. People are now declaring Zen to no be a Sandy or Ivy like IPC. Its now being touted as better than Haswell on many forums. And its gonna have to meet that mark. Which i truly hope its at least close..cause thats more than i ever expected. if its better, bring it