What about Zen+? What do we know about them? Any feature you're

Potentially 3% IPC increase and 400-500 mhz clockspeed boost. Screencap this post.

>>59449038
try 10%

>>59449001
Too much stuff to fix and include to get a well educated number, I'd wager some average 8-15% IPC and higher clocks.

IT'S SHIT AMD IS JUST SHIT

>>59449038
close

>>59449152
closer

>>59449374
just about right

>>59451337 (You) (You) (You)

>>59449152
>why didn't AMD motherboard manufacturers fix it?
intel is cucking people with 1-5% increases
bretty sure amd wants in on that milking phormula

pretty sure the real gains frmo zen+ will be from mobo manufacturers releasing hardware that isn't cucking amd buyers like they currently are.

pinnacle ridge when?

>>59449038

amd cpus perform way better with higher clocks, if 100% is 3.7ghz then expect a 500mhz increase to give 20% increase in performance.

thats the reason i think next gen wont clock any higher, and at the moment software is amds biggest enemy since its proven that optimized software can take way better advantage of the zen cpus

>>59451621
The entire issue about low clocks on Zen is from 14nm LPP, which is, to be straight to the point, a mobile process that isn't meant to clock high.

If AMD wants higher clockspeeds, it'll have to use a HP process or whatever Samsung/GloFo call it, but that would lower Fmax/Vmin scaling down and the chips wouldn't be as efficient as they are now at lower clocks.

>>59449001
when we can except second generation of zen?

>>59452343
They said 2018 in 2015 but it seems implausible. I think they could come out in 2019 2Q

>>59452343
should be late 2018

>>59449001
don't expect too many changes

>>59451490
AMD can't start cucking people until they get more market share

intel is still better lmao amdrones

>>59449001

Needs a faster CCX interlink and higher clocks.

File: AMlosi.jpg (259KB, 670x768px) Image search: [Google]

259KB, 670x768px

>>59449001

Well, we are waiting

>lol just wait for the next AMD product

AMDlets

when will they learn?

>>59455816
this

>>59456204
People can actually look forward to new products from AMD. If it was Intel they'd just get the same product every year.

>>59449001
I would like for them to already be moving to 6 or 8 cores per CCX for Zen+...
but pretty sure that's not happening until 2019 or 2020, if even then.

More realistically:
-Improvements to the interconnect to reduce latency and increase bandwidth.
-Decoupling some things from the CCX interconnect
-MMU improvements
-More vertical L3 cache to pack it tighter
-~10-15% higher stock clocks and overclock headroom both with a different manufacturing process for desktop than server.

They might make the FPU a bit larger and improve AVX2 performance, but I doubt that really. I feel like the cores themselves will stay exactly the same for a while.

I'm also somewhat doubting they're produce desktop and server parts on a different process. I think they'll force us to deal with this to increase margins.

>>59449001
It's free real estate

>>59452201
LPU, a newer Samsung process, is supposed to clock 10% higher at a given vcore with no draw backs.

Question is: Has GloFo licensed LPU from Samsung like they have LPP?

>>59454711
>>59454742
Everything I've seen suggests Q1 2018 for 2nd generation of Ryzen.

AMD really seems to want to do yearly updates now that they have something worth updating.
With Bulldozer they were stuck on 32nm and 28nm due to really stupid contractual agreements that the former CEO signed with GloFo.

>>59449001
So if I'm understanding this chart correctly, the roadmap dictates that the future chips will be more powerful than the current chips.

Fantastic.

>>59456246
"A socket a year keeps the goyim in fear."

>>59452201
Not even true breh. Ryzen doesn't clock crazy high because it's a jack-of-all-trades architecture with the server market as its primary target. Seriously think about how impressive it is that AMD has one single architecture-nay, one single cpu design-that is compelling enough to fill an entire product portfolio. From a design and production standpoint, Ryzen is really fucking neat.

Obviously such a route has its drawbacks. The most obvious being underwhelming clockspeeds and serious growing pains as they sort out all the unpredictable issues and bugs moving forward. It's a bit disheartening to see just how troublesome the launch has been, but it's to be expected when their resources are so limited.

>why didn't they fix this shit before launch?
Ryzen is already 5+ years in the making. They've done the best they could, but another daunting question all those years has been "when the hell is AMD going to launch a new CPU already!?"

I don't doubt we'll see updates or even full refreshes between now and Zen+. But before anything else, AMD needs capital, and without a doubt wins in the server market will bring plenty for them to invest with.

File: 1481484192769.jpg (36KB, 604x604px) Image search: [Google]

36KB, 604x604px

>>59457785
kek

>>59457788
It has nothing to do with the architecture.

It's the manufacturing process that has ideal Vcore/Fmax ratios at lower frequencies.

>>59457620
The good news about a single process is they can focus their attention on an essentially unified product line. Improvements across the board will trickle down to all users the way they have with GCN. AMD has show what making small changes to a scalable architecture can mean for longevity in that area, and Zen should show similar results.

A great sign is their increased efforts in working more directly with 3rd parties, though obviously they're still quite limited in scale. With present resources they have to decide carefully which areas to focus most on, lest they bury themselves in work they can't hope to finish effectively and efficiently. For now we'll see them focus on the server markets first (once naples goes live, that is) meaning limited support for consumers, but again the nature of Zen means improvements should trickle down.

>>59457899
An architecture can be made to work on a few various processes. Especially ones as similar as LPP and LPU.