The architecture, IS finished and WAS finished. The clock speeds and all of that all already all decided, that's all needed to have been finished so that Zen can launch soon.
Don't be dense, it was a single benchmark and if you don't think it was cherry picked, your out of your mind. Wait for the full benchmarks before we run around calling Zen the savior of AMD. Let me remind you that even the promised 40% IPC bump over the 8350 class parts doesn't put Zen much farther along than Haswell-esq IPC if I remember right. There's a lot more to the story than a single canned benchmark...
Also, they won't be hitting 4.0Ghz out of the box on the 8 core 16 thread part. The manufacturing process they're using isn't conductive to those kinds of clock speeds, and the Zen core is a much more complex core than a Vishera core. Its a lot harder to hit those high clock speeds on a core with the size and complexity that Zen demands. They'd be doing more than cherry picking to have to bin for 4.0Ghz stock clocked, 8 core, Zen chip.
Double Also, if you didn't get that I was joking around and making fun of the hype train, please, reread.
I'm not really sure i buy the OS only tweaks. I'd be more willing to believe that amd has some new fangled version of UEFI for messin with OC's and what not.
I am digging the $300 price point because if it can hang with the 350-420 buck intels at that price then that would definitely change things for the better. Now, I said hang with, not out perform completely.
well.. theres always the thing where 'performance' is entirely relative, some people talk about performance meaning how many instructions it can execute as a whole(the entire chip) under real world scenarios vs what it can do running some piece of garbage software thats not designed to go fast at all really
if we're talking about the first one, assuming amd isnt lying regarding the blender test which intel almost always does very well at(rendering in blender in general) then it would be competitive in 16v16threads vs 5960x/6900k assuming similar clocks. in this area the 7600k wont even compete with a 6core zen source: anandtech cinebench r15 multithread
i5 thrown in just for kicks, but mainly talking about the nehalem i7 990x, being five gens older(although die shrunk) sure the 6700k wins here but its only in the clock speed, but the 990 oc's alot more than the 6700 will hitting about 4.3-4.6 from 3.46, instead of going 4>4.5-4.8
basically caby will maybe be competitive with a sandy 6 core for the i7 quad
in the second kind of speed, even a 4770/4790k will be faster since those are faster than 6900/6950x no big surprise, sure theres some workloads which are difficult to make single threaded but as is proved by this and the previous consoles generation shows many of them seem to be able to make use of more than 4 threads, on pc the artificial limit of 4 threads is probably just laziness/what the steam survey says the most people have although, have been those leaks about intel pushing 6 cores with coffeelake i think, unless they jack price up finally just give people more hardware for the money
Now you are being misleading . It is a 40% IPC boost not over the 8350, but the Excavator core which is 10% higher IPC than the 8350 Vishera core. So your ptremise is flawed.
Not true. Clock speed and clock/voltage ratio is the last thing that will be known. In August Lisa Su couldn't know final clocks for release date because node is in maturing process. If the leak is correct, node matured for 400mhz from august to october/november
And there's this from Samsung (not sure is it applied to Zen already)
The company said that the new process will deliver higher performance than the third-gen 14LPC without affecting the power requirements. The fourth-gen 14 nm process technology is suited for high-performance and compute-intensive applications.
They were STILL testing the Engineering Sample when that was being demonstrated, that wasn't the complete version. The final clock speed still isn't decided.
Again, this was a comparison with Instructions Per Clock (That was the POINT of comparing them at the SAME clock speed), on a program called Blender, which BTW, is a very common 3D modeling program that anybody can use and is a decent program (is counter-intuitive though), unlike the case with the Polaris benchmarks where they picked games I am probably never going to play and most people aren't.
Also, no, they promised 40% greater than Excavator (i.e. FX 8370), which was already 10% above Vishera (FX 8350) and even further ahead of Piledriver and Bulldozer where AMD started to FLOP on their CPU design.
actually 8370 is still piledriver, but excavator is 10-15~ depending on exact example over pile driver, only excavator chips are the carrizo apus, steam roller was kaveri and godavari, but those only had slight ipc but mostly was gains in less penalty for using both threads in a module, and pile driver had slight ipc over bulldozer but also with better clocking(probably from yields)
nope, nor a steam roller chip, was some talk(maybe just rumors cant remember at this point lol) about the silicon then, not being upto snuff, them being on glofo 28nm something basically that the silicon just wasnt gonna work for high power, which.. dunno, but basically kaveri clocks lower than richland(which was piledriver/trinity refresh), and carrizo clocks even lower than kaveri but if the silicon really was just not good for high power/high clocking usage i guess would make sense to just optimize shit out of it for lower power,
which the carrizo laptop chips actually looked decently good, like 2.7ghz turbo 3.5 35w laptop, then oems fuck them again and dont ship anything cept 15w version.(and in barely any choices which all still sucked) but basically ipc went up pretty decent amount, clock speed went up, shaders went up like 20% clocks for those went up, and the power gating is better. for the 450-550$ like lower end i3 type notebooks would be a good option for poor kids or whatever can type whatever school work and play fallout 4 at 720p
Only cpu on desktop (for now) with excavator cores is athlon 845 on fm2+ socket (which is probbably higher binned and overclocked laptop version). Other excavators will come to AM4 socket as budget option (still current gen - 28nm, CMT) and will be called bristol ridge. That cpu's are finished long time ago (in summer or autumn) but AMD allegedly postponed it release because they need to work more on AM4 platform. They allegedly don't care about Bristol ridge success too much (current gen and lower tier product after all), Zen and having a good AM4 platform is priority so bristol for custom builds will wait. These earlier relesed OEM PCs with bristol on AM4 platform probably won't see Zen support
I'm more hyped about the 490 Vega card being 12TFLOP, the more I hear about Zen the more worried I am that its either going have major issues or cost insanely too much. Its getting to the point of 'too good to be true, then likely isn't' stage!
yep, but it clocks low/is locked i think, personally would guess actually a chance glofos 28nm just sucks for high power lol.
but would have guessed definitely a chance zen wouldnt clock as high, being steamroller and then excavator both successively clocked lower, if they had carried over some of the stuff that increased the ipc(branch prediction whatever other stuff) was why those had clocked lower.
12tflops seems pretty resonable, being they already had 8.9 or whatever, with arch improvements over polaris even, probably 15-25% higher clocks. 1000max for the fury nano at 8tflop, 1050 for the furyx, hitting maybe 1150~1300 for new vega card, especially considering the reports of newer 480's the higher binned pro parts power draw being equally impressive(as the reports). for my extrapolation would guess 40~45% pretty doable in gaming performance even with the same 4096 shaders considering the significantly poor utilization in dx11 the fury(s) usually had, the improved geometry processing that will probbaly be present like is on the 480 etc.
as far as zen, honestly think even if amd delivered literally, exactly a 5960x for say 400$ soo many youtube shill/people in general would shit on them for not even being able to compete with the 7600k fails again, zens the bulldozer etc. to them the 6700k is only faster than the 6600k because of the clock speed they care jackall about max performance just the utilization in the current game titles.
only way i'd see them having a hope of not getting shit on is if they had a quad with smt at 4~4.2ghz for like 150$ where it would be just barelly slower than an i5 and unlocked with more threads
still dunno lol :p. the rumors of there being a specially binned one, if that one was ~450, the non super bin could be like 350-400$. just makes sense to do it when they get better reviews(since they always look at the flagship) but can still ship the lower fully functioning 8cores for more than a 6 or 4 core
thats still an i7 for price of an i5, an i7 six core for the price of i7 quad and i7 8 core for price of i7 6 core. for the low end still have the carrizo apus, and raven ridge apus may follow similar pricing to those dunno
Still allot of rumors that i cannot realy get my head arround. BUt 8 core 16 threads zen chip that performs close to a 6900K for $300,- is most likely not gonne happen. They might undercut Intels prices a bit, but i cannot believe that they would undercut it that much. That would not make any sense to me personally. Because AMD has to make money pretty badly out of this. But if the rumor about the pricings turns out to be true. Then i could imagin that the performance of Zen might be lacking.
