Excavator on SOI has been confirmed to be >4 GHz on the highest 65W SKU.Have we even confirmed that there will be a high-frequency version of the Excavator cores? Excavator is meant to be made using High Density Libraries- these are intended to significantly improve power consumption, but at a cost.
Depending on what AMD did with the "Excavator" architecture. AMD could be able to improve performance with little effort.
The changes with Steamroller's front-end allows for an evolution from MCMT. This evolution is specifically called Clustered SMT.
MCMT;
http://i.imgur.com/v3CTSnd.png
1st Gen CSMT;
http://i.imgur.com/uRI6Byp.png
(FDL -> Forwarding Logic, this removes dependencies between clusters sharing a module. This will improve performance when running SPMD in regards to integer. The FPU is already SMT, so the improvement in the FPU will need to come from AVX2 and FP256.)
Excavator on SOI has been confirmed to be >4 GHz on the highest 65W SKU.
Excavator on SOI has been confirmed to be >4 GHz on the highest 65W SKU.
Come on guys stop with the AMD vs Intel crap.. both are great Cpu's and this isn't 2000-2001 any more and anything with 4 core is more then enough for 90% of users. We all know 860k-960k is no where near a 2500K and its not meant to be.Its like comparing a Dodge Caravan to a Lamborghini..Both vehicles will get you to where you need to go but the more expensive one will get you there a little bit faster.Not every road is gonna let you go 120Mph though just like your not going to be able to tell the difference in every program between either cpu..Maybe in gaming or video encoding but thats about it..
They are capable of introducing both forms with little effort without causing a huge change. It is essential adding a macro/micro-op cache level of difficulty.Clustered SMT is far from "little effort".
Where?
AMD Leads on Linkedin.Where, precisely?
Come on guys stop with the AMD vs Intel crap.. both are great Cpu's and this isn't 2000-2001 any more and anything with 4 core is more then enough for 90% of users. We all know 860k-960k is no where near a 2500K and its not meant to be.Its like comparing a Dodge Caravan to a Lamborghini..Both vehicles will get you to where you need to go but the more expensive one will get you there a little bit faster.Not every road is gonna let you go 120Mph though just like your not going to be able to tell the difference in every program between either cpu..Maybe in gaming or video encoding but thats about it..
The athlon K versions usually have the cache chopped. So I doubt it'll get remotely close as its a neutered chip to begin with
They are capable of introducing both forms with little effort without causing a huge change. It is essential adding a macro/micro-op cache level of difficulty.
Front-end level because of the separate decodes.
Execution level because of the proximity of the cores.
For execution level;
What AMD would be adding is a Forwarding Retire Queue or upgrading the current Retire to be able to have Forwarded instructions. Then, Forwarding Logic to scan the Retire Queue for dependencies between both cores. If there is a dependency, the given instruction is shifted to another core, where it will be resolved quicker.
LinksAMD Leads on Linkedin.
Excavator can be on only two nodes after 28-nm SHP; 28-nm SOI or 20-nm LPM/SOC. Puma+ being on a 28-nm node is hinting at 28-nm SOI.
I seem to remember you saying your overclocked 760K would match or even beat an i5 of recent vintage. When in all reality there hasn't been an i5 made that isn't more than a match for a 760K stock or otherwise.
Lies? how stupid can you even be. I got double the FPS of my old Phenom II X2 with my previous intel dual core pentium G haswell which is still unplayable since the old phenom II X2 is around 9 FPS min in many many situations. Totally unplayable
Maybe do some research next time before looking like an idiot making claims about my "lies"
dual cores STINK in these 64 player maps with high tech Engines like frostbite 2 and 3. Mantle tries to help but its still pretty bad, the haswell pentium G3220 was trash which is why I got rid of it. Or wait you must be one of those idiots who look at Single Player benchmarks ROFL
Just imagine 3 years later after the release of battlefield 3 and people are still this retarded to pointing to single player benchmarks and saying ow look BF3 runs smooth on dual cores guess thats that.
Ow wait nobody buys these games for single player neither do they play single player.
Find another way to make your point without calling people idiots.
esquared
Anandtech Forum Director
What you guys think will happen here with excavator in terms of single and multi threaded CPU performance?
AMD likely to reach sandy bridge or ivy bridge performance finally?
Have we even confirmed that there will be a high-frequency version of the Excavator cores? Excavator is meant to be made using High Density Libraries- these are intended to significantly improve power consumption, but at a cost:
http://www.anandtech.com/show/6201/amd-details-its-3rd-gen-steamroller-architecture/2
So an Excavator core may not be able to clock as high as a Steamroller core. I expect that we will see Kaveri stick around on the desktop.
XV is on the 28nm process as SR and Puma+. We have an official AMD slide from last week.
I don't know about anyone else but I play Battlefield 3 on a G3220 with a 560 ti on 64 player maps, 1080p@high and my fps stays within the 50-80 range, absolutely playable for me.
They are using HDL that is affecting the frequency scaling drastically. XV core would be lucky to clock anywhere near SR no matter what process changes they implemented.No, that says it's on a 28nm node. Not the same process.
They are using HDL that is affecting the frequency scaling drastically. XV core would be lucky to clock anywhere near SR no matter what process changes they implemented.
Using this 28nm SHP process, which employs bulk silicon, forced AMD to sacrifice some CPU frequency at the high-end of the TDP scale, but with better power characteristics at the sweet spot.
XV is on the 28nm process as SR and Puma+. We have an official AMD slide from last week.
I seem to remember you saying your overclocked 760K would match or even beat an i5 of recent vintage. When in all reality there hasn't been an i5 made that isn't more than a match for a 760K stock or otherwise.
Considering how much AMD is hurting for cash, they'll likely charge more if it performs better, rather than forcing Intel to lower their prices to compete.
$190 7850K's, anyone?
This 100%.
Intel has a lot of budge room on its margins and binning while AMD doesn't.
Every single intel consumer chip ships with a fully functional igp with the full number of EUs. Quads with core defects are discarded. Only 'perfect' chips are sold.