Will AMD deliver with Pinnacle Ridge? Or will it be Vega 2.0 - Electric bonglo?

Will AMD deliver with Pinnacle Ridge? Or will it be Vega 2.0 - Electric bonglo?
Has anyone other than currycraft report on this?

Other urls found in this thread:

youtube.com/watch?v=HVtojNukkA0
twitter.com/SFWRedditImages

It's Ryzen on higher clocks and some uarch tweaks, aka what Intel used to call a tick (Ivy/Broadwell)

it is not possible to not deliver, its basically Zen with bug fixes and 12nm process for higher clocks
will buy for sure, time to replace my aging FX

Are IPC increases possible?

No, its just a refresh.

yes, of course

Yes, even on the same uarch you can extract higher IPC with smart fiddling with the caches, TLBs, branch predictor, etc.

A simple term is setting your disk block size to lets say 4MB instead of 512KB, no real change needed but it still improves performance.

thank god someone with the knowledge showed up, or I might've been bashed for saying that there will be increase

youtube.com/watch?v=HVtojNukkA0

Pinnacle Ridge is not Zen v2. It's more like a new stepping.

The improvement from Pinnacle Ridge should be about as big as the improvement Kaby Lake made over Skylake.
So basically, you'll maybe get a couple of percent IPC increase, mostly you'll just get a few hundred MHz higher clock speeds because of the improved manufacturing process.

Zen v2 is where IPC is going to increase significantly, as well as use 7nm which is much better than their 14nm/12nm process.

Zen2, what clock rates can we expect? 5ghz? I was gonna buy ryzen but I thought I'd waiting till zen2

The 7nm process is targeting 5GHz stock.

The current 14nm process has a >3GHz (up to 3.5GHz) target. And AMD is pushing it further than its target.

Impossible to say what to expect. Global Foundries is aiming to deliver a process capable of 5ghz, and AMD certainly needs that to compete with intel. Certainly north of 4ghz should be all but guaranteed. North of 4.5ghz possible.

Funny to think that 32nm was able to hit 4ghz.

Cool, but wheres Raven Ridge?

>Navi
>Watch out!

FX chips still hold the highest clock speed records.

Yeah, well 14nm was originally made for mobile CPU's. Low power consumption, but lower clock speeds.

Wouldn't expect more than 1-3% if that, just like with Intel tick. It's going to be clocked up to 5Ghz though according to rumors.

This refresh is not Zen2.

why should they? clock on clock ryzen is faster than intel ipc wise

what ryzen doesnt have is enough OC headroom to battle intel

pinnacle ridge will be the zen+ on 12nm
the true zen2 on 7nm is named nowdays matisse or mattise whatever and its not going to be here untill 2019..

Hey!

It was 8ghz or something close to

Not exclusively for. Lower power draw is a side effect of a more efficient manufacturing process, higher clocks come as the process matures. Its gonna be a while till we see smaller that 7nm, and I'm supriwed that were still using x86 honestly. Can't wait for proper emulation

>Can't wait for proper emulation
Like the one in Itanium?
That ended well...

Itanium could have worked if people suffered through it for half a decade. Realistically you can engineer anything to the point of "good enough."
There isn't any concrete reason for us to hold on to X86 apart from existing software. If a new arch can emulate this well enough during a transitional period where new native software is phased in then X86 would soon be a distant memory. Itanium ultimately failed because it set out to do something while giving up too much. AMD came along with the AMD64 extension and that didn't require the same sacrifices, so that won out in the end. The same situation could play out again today with solid emulation.

What should we use instead of x86? PowerISA is only usable for building ovens, ARM is locked down turd, RISC-V is useless outside of IoT shit. That leaves us with?

The thing is who remembers itanium?

How is arm locked down, especially anymore so than x86? Heck x86 can't even be made without having a licence and perpetuates a duopoly. Riscv could easily be extended again, though arm did it well enough imo

Thats up in the air.
Maybe some switch on event optoelectronic DARPA tech. Maybe some other radically different non Von Neumann arch.
I think any conventional core arch we have in the market today will reach more or less the same plateau for IPC and total perf/watt. Put enough money into ARMv8 or some future derivative and we'll have a slightly different X86. Change the names of the vendors but the situation will be the same.

Unless we look at something still in the realm of white paper theory we're stuck in the realm of needing to increase single core perf/clock. We aren't slowing down in this area because the actual execution ALUs are being bottlenecked, we're slowing down because actually fetching, decoding, and scheduling the instructions to feed them is immensely hard. We can build these super wide cores and ALUs go unused so often and so consistently that we can support virtual threading on them ala SMT. How are we going to give the front end of a core 30%, 50%, 100% more throughput? Thats the billion dollar question.

Why shouldn't they? It's a first generation core, it has at least another 20-30% headroom to grow.

matisse will be 5ghz, not zen refresh

it has at least 200% if they get THIS design up to 5nm
the problem will remain tho if amd go from 2 to 4 and 6 per ccx then im afraid that we are going to end up having a monster cpu even at low end but without proper support to utilise all of that

Zen's time ends in 2020/21, AMD said 4 years for consumer Zen products, it's one of the release day interviews, they said AM4 will be supported for 4 years and after that they move to a new arch.

>Or will it be Vega 2.0 - Electric bonglo
AMD's CPU and GPU divisons are entirely seperate, any comparison here is fairly pointless.