When And took off, it TOOK OFF!

Its really not that surprising that they would want a Zen2 Threadripper on 7nm out there, as Ryzen 3000 would otherwise already eat into the lower-end parts of the previous Threadripper lineup, if the rumors of its specs are true.

For me it was obvious.... AMD will release it as soon as they can. they want to keep the momentum of Ryzen going and strike now when Intel doesn't have an answer. Intel can quickly catch up with their massive R&D and AMD knows that.

Looking forward to upgrade my TR trusty 1950x 🙂

HardwareCaps:

For me it was obvious.... AMD will release it as soon as they can. they want to keep the momentum of Ryzen going and strike now when Intel doesn't have an answer. Intel can quickly catch up with their massive R&D and AMD knows that.

The only part I don't agree with is the word "quickly". Intel is struggling to keep up with their own plans that had little to do with AMD, to the point that they're contracting TSMC. Intel has more than enough money to make a major competitive product but they don't seem to be able to get it out fast enough.

schmidtbag:

The only part I don't agree with is the word "quickly". Intel is struggling to keep up with their own plans that had little to do with AMD, to the point that they're contracting TSMC. Intel has more than enough money to make a major competitive product but they don't seem to be able to get it out fast enough.

Intels problem isn't actual processor R&D, that is probably done for two or three future generations at this point. Their main problem has been process technology, ie. 10nm. If that worked, Ice Lake would've been out years ago. This is a topic AMD isn't dealing with at all, since they just outsource that to TSMC, so its not really anything comparable. Its also not really something you can make go much faster with more money, since this stuff just takes forever (and already costs billions). Once 10nm is actually out and stable, I fully expect Intel to "strike" far faster. And with any luck their next process won't suffer from the same delays, with the lessons they learned from 10nm. So perhaps they are finally over the "hump" soon when 10nm launches in mainstream, and quicker progress will resume, now that they actually have competition again. On top of all that, Intel has gone on record to say that they'll be designing future CPU cores less dependent on the process architecture, so even if a future process is delayed again, hopefully the next core architecture won't be directly married to it, and they can release it on a different node if need be. All that said, AMD is still "catching up" even with Zen2, the only thing AMD has going for them is throwing cores at people, which at least to me is getting old beyond a certain count in mainstream/enthusiast.

nevcairiel:

Intels problem isn't actual processor R&D, that is probably done for two or three future generations at this point. Their main problem has been process technology, ie. 10nm. If that worked, Ice Lake would've been out years ago. This is a topic AMD isn't dealing with at all, since they just outsource that to TSMC, so its not really anything comparable. Its also not really something you can make go much faster with more money, since this stuff just takes forever (and already costs billions). Once 10nm is actually out and stable, I fully expect Intel to "strike" far faster. And with any luck their next process won't suffer from the same delays, with the lessons they learned from 10nm. So perhaps they are finally over the "hump" soon when 10nm launches in mainstream, and progress will resume, now that they actually have competition again.

I agree; that's pretty much what I meant, I just didn't phrase myself very well.

On top of all that, Intel has gone on record to say that they'll be designing future CPU cores less dependent on the process architecture, so even if a future process is delayed again, hopefully the next core architecture won't be directly married to it, and they can release it on a different node if need be.

I'm curious to see how that will turn out. The benefit you mentioned is a good one but node sizes aren't changing all that often anymore. I think the level of optimization Intel does with their nodes is what gets them such high clock speeds.

schmidtbag:

The only part I don't agree with is the word "quickly". Intel is struggling to keep up with their own plans that had little to do with AMD, to the point that they're contracting TSMC. Intel has more than enough money to make a major competitive product but they don't seem to be able to get it out fast enough.

well "quickly" can mean quite a few different things. I definitely think that Intel will stop slacking and will get the 10nm process running by the end of the year. I think in 2020 we will see a good Intel product. 2019 is the time for AMD to shine. because Intel literally has their own fabs they are a lot more flexible in terms of changing processes and speeding things up. for AMD/Nvidia they basically rely on TSMC/GF if they miss their targets/goals or choose to focus on a different node/market(GF dropping 7nm plans, TSMC prioritizing mobile). there isn't much to be done.

32c64t TR 3950x? What can Intel do?

Undying:

32c64t TR 3950x? What can Intel do?

Probably have another record year 😱

Noisiv:

Probably have another record year 😱

Lol yeah - they made personnel changes that will yield improvements a year or so from now but in the meantime their massive clout & entrenchment in the industry gives them so much time to maneuver. It doesn't help that AMD is having a hell of at time getting these products into OEM channels. We'll probably find out that Intel is just paying companies to avoid AMD again. The last time it cost them ~1.3B fine for 90% marketshare. Basically a bargain deal even if the fine was 5x as much.

Its good we get great CPU´s from AMD and hopefully good priced but current GPU´s are not kind of hardware i want to pair with it.

"strike now when Intel doesn't have an answer. Intel can quickly catch up with their massive R&D and AMD knows that." keep dreaming... not if have the fundamentally wrong archiecture for the current playing field. That takes many years - irrespective of budget, & same goes for serious GPUs. Intel should have adopted MCM ages ago - even their engineers said so, but they pretty much said openly that desktop was dying and they had sexier markets to chase - so they just used their long amortised monolithic architecture as a gravy train, and got sucker punched by Zen MCM. As we see, the 8700k pretty much maxed out the architecture at 6c/12T - its successors have been a yawn which add downsides in equal measure with upsides - anything ot get max ipc on a few cores - to yield a good sound bite for their corner case advantage over MCM. MCM give plenty of cores, but has inherent lag, & effect on IPC. For anything threaded, intel get shredded.

nevcairiel:

...All that said, AMD is still "catching up" even with Zen2, the only thing AMD has going for them is throwing cores at people, which at least to me is getting old beyond a certain count in mainstream/enthusiast.

I'd suggest that with the IPC and additional architectural changes for Zen2 that have been well described, and for some time, the 'old' 'throwing more cores' is almost certainly not the only positive change we'll see. Intel, of course, did have an IPC advantage for some time - and an advantage that was eagerly waved time and time again. Providing sources are correct (we all hope), it seems that this advantage will, at the very least and finally, be nullified.

nevcairiel:

All that said, AMD is still "catching up" even with Zen2, the only thing AMD has going for them is throwing cores at people, which at least to me is getting old beyond a certain count in mainstream/enthusiast.

Speak for yourself. I could always use more cores - I'm 100% maxed out! 😉

Denial:

The last time it cost them ~1.3B fine for 90% marketshare. Basically a bargain deal even if the fine was 5x as much.

It would be funny to see them get 1.5 billion fines from the USA, Canada, the EU, China, Japan, India, Brazil, Australia, etc. Then it would start to affect even Intel.

All I want is - a nice 10-15% IPC increase. - a decent clockspeed increase to around 4GHz base and 4.5GHz boost (sorry I don't believe the rumours about 5GHz) - a few more cores and threads - a 105w TDP (95w would be better obviously but with the added above extras I don't see this being viable). The rumoured 3800X seems like a steal, with 12 cores 24 threads but I seriously don't believe the rumoured 5GHz clock speeds. I don't think 7nm is there just yet. If they can bring 10-15% IPC improvement at the same clock speeds whilst bumping the clocks up from 4.3GHz on the 2700X to 4.5GHz on the 3800X then I will be happy. We all know that they over engineered the X370 boards with most having ridiculous VRM's and mosfets that are way over powered for the first gen Ryzen's. So this to me tells, me that even the X370 boards will for the most part be fully supported. I mean X470 over X370 doesn't give you much if anything at all.

the siege has to be relentless!

CPC_RedDawn:

All I want is - a decent clockspeed increase to around 4GHz base and 4.5GHz boost (sorry I don't believe the rumours about 5GHz)

That's about what I hope they'll do, a 10% increase in frequency while using no more power than Zen+. 10% would get us to 4.7-4.8 GHz. Another thing that will be interesting to see is what will be the minimum configuration of Zen2. It seems like yields would have to be poor to cut down Ryzen 3000 to 4c/8t. Could we see the lowest Zen2 sku#s at 6c/6t?

All that said, AMD is still "catching up" even with Zen2, the only thing AMD has going for them is throwing cores at people, which at least to me is getting old beyond a certain count in mainstream/enthusiast. And who could have imagined the adding cores can actually increase performance...?