AMD's Zen and Intel Kaby lake CPUs are reportedly delayed

[Coaxialgamer]

5 minutes ago, zMeul said:

I know, and I specified LPE

but LPP is an evolution of LPE

And don't forget this is a different process on tsmc's side too

[zMeul]

11 minutes ago, Coaxialgamer said:

And don't forget this is a different process on tsmc's side too

it's still FinFET

 

back in January, there were rumors TSMC had issues with the LPP process; while Samsung announced 2nd gen 14nm FinFET (LPP)

because of that, everyone thought nVidia will 2nd the release of the new generation - as it turned out, quite the contrary; and by the looks of it, nVidia has two Pascal production lines

[Prysin]

1 minute ago, zMeul said:

it's still FinFET

 

back in January, there were rumors TSMC had issues with the LPP process

because of that, everyone thought nVidia will 2nd the release of the new generation - quite the contrary

by this point you are just propaganda trolling

[RagnarokDel]

12 hours ago, annoyingmoments said:

Ehm, it is even worse for AMD then.....but it was already rumoured or half-officialy told by AMD that there are delays.

As for Intel: They bring every year something new, so no big deal. And if they delay it for the sake of to big stock of Haswell and Skylake, then alas, they wait for a good moment to push new technology. (Waiting for AMD?)

Yes, the CEO made it extremely clear that it would be delayed when she said it was on schedule at Computex but let's trust a rumor-mill over the CEO of the corporation creating the product.

[Master Disaster]

12 hours ago, Notional said:

Intel's Skylake CPU's has been very expensive with little added performance/features to show for it, so it would not be a surprise if Intel has built up a big stock they find hard to sell.

But it makes zero sense for AMD to do the same. ZEN with 8 core/16 threads does not clash with any existing product SKU from AMD, thus it cannot cannibalize any other AMD products, but only compute with Intel products. So the argument that stock plays a role for AMD is nonsensical.

Wasn't part of Skylakes huge price because of low stock issues? How could they possibly have gone from not enough stock to too much stock in a few months? 

[Notional]

12 hours ago, zMeul said:

makes sense if you also consider the reports AMD has problems with Polaris clocks

it could be that GloFo has issues with Samsung's 14nm FinFET

There were no reports. There is a moronic rumor claiming Polaris couldn't even do 850mhz, that was heavily debunked few days later, when we saw Polaris running over 1200Mhz. Considering AMD's architectures are based on high performance in the architecture, and not just high hz speeds, that translates poorly to more performance, it sounds pretty good.

10 hours ago, Curufinwe_wins said:

Also, in fairness to Zmeul, it isn't obvious 1200MHz means they aren't having clock speed issues.

 

I mean consider this: Suppose Nvidia was showing off Pascal with 1500-1700 speed clocks. In hindsight that would be clear evidence of issues with clock speed (since low 2000s are normal), but without knowledge of Pascal, you would NEVER have arbitrarily expected clock speeds that high (at the uppermost levels of Maxwell OCs) to be indicative of a POOR chip.

As stated zMeul's rumor FUD was that Polaris couldn't even handle 850mhz. What you do have to bear in mind, is that Paxwell, sorry Pascal's architecture is optimized for high frequency. That is very much possible, but also comes at a cost. Hz on a chip is similar to RPM's on an engine. More is better on the same engine to some extent, but a 10.000 rpm 2 cylinder engine will not outperform a 5.000 rpm 12 cylinder diesel engine. NVidia's architectures are more like gazoline engines and AMD's more like Diesel.

 

Also do remember Polaris is still the mid to low end chips. We have no clue if Vega will run at higher frequencies.

5 hours ago, zMeul said:

GloFo licensed the 14nm process from Samsung: http://www.globalfoundries.com/newsroom/press-releases/2014/04/17/samsung-and-globalfoundries-forge-strategic-collaboration-to-deliver-multi-sourced-offering-of-14nm-finfet-semiconductor-technology

 

if you recall, Apple sourced custom chips on LPE process from both Samsung (14nm) and TSMC (16nm) - those from Samsung were hotter

I'm sure that one CHINESE test showing TSMC being 20% more efficient and cooler is SUPER LEGIT.

 

Now back to reality, where 14nm FF LPE from Samsung is better than TSMC's 16nm FF (non +):

http://www.tomshardware.com/news/iphone-6s-a9-samsung-vs-tsmc,30306.html

 

So let's look shall we?

 

Hotter?

Nope, Samsung's process node is cooler.

 

Is it more power consuming?

Quote

In this case, we see Samsung's 14nm FinFET process has a slim but noticeable advantage, lasting 10.8 percent longer (16 minutes) than the iPhone 6s Plus with the TSMC made A9.

Nope not that either.

How about worse performance?

Quote

As expected, there's no discernible peak performance difference between the two different A9 models. All of the CPU, GPU, and system performance scores show less than a 2 percent difference, which lies within the margin of error for these tests. 

Nope, 3/3 strikes. You're out. Too bad you blacklisted me for constantly debunking your FUD propaganda. You might learn something, but I'm pretty sure you don't want to learn.

 

Samsung's process uses less power, runs cooler and has same performance as TSMC's. Albeit the difference is miniscule between Samsung's LPE and TSMC's FF non plus.

 

3 hours ago, leadeater said:

At least HE didn't drop it .

Well...

https://zippy.gfycat.com/BraveWhichDolphin.webm 

[Trixanity]

7 hours ago, zMeul said:

let me remind you that nVidia shower 2+Ghz on stage and no one was able to achieve that with the FE cards, only with the AIB custom boards

 

if AMD didn't had, and does not have, problems with Polaris, where are the cards? they showed Polaris 10 ever since January

If Nvidia aren't having problems, where are the 1080s? I can't buy one. And if I could find one, it would be priced $100-$150 above what it should cost. 

 

Oh wait, maybe AMD aren't rushing out to paper launch their cards with a stock availability of about 500 cards worldwide for a $100 mark-up. 

 

Maybe - just maybe - they're making sure availability will be in abundance, so you can actually get a card for $199 on launch day. 

[79wjd]

@Notional every other report on tsmc vs Samsung a9 gives tsmc the advantage. But no one has large enough sample sizes to mean anything at all.

[Notional]

5 minutes ago, djdwosk97 said:

@Notional every other report on tsmc vs Samsung a9 gives tsmc the advantage. But no one has large enough sample sizes to mean anything at all.

The only test I've seen that shows that result, is the chinese test. Every other site that claims the same, is based on that one chinese test. If you have any other test, I would be interested in seeing it

[79wjd]

23 minutes ago, Notional said:

The only test I've seen that shows that result, is the chinese test. Every other site that claims the same, is based on that one chinese test. If you have any other test, I would be interested in seeing it

I never bothered looking into the sources (or their validity) for all the reports I saw since I didn't care enough, but iirc Ars did their own testing.

 

http://arstechnica.com/apple/2015/10/samsung-vs-tsmc-comparing-the-battery-life-of-two-apple-a9s/

 

John Morrison and Austin: http://www.cultofmac.com/391574/tsmc-a9-chip-gains/

[lukart]

For sure AMD will have the Zen out by the end of the year, now... will it sell in volume? no.

[Notional]

36 minutes ago, djdwosk97 said:

I never bothered looking into the sources (or their validity) for all the reports I saw since I didn't care enough, but iirc Ars did their own testing.

 

http://arstechnica.com/apple/2015/10/samsung-vs-tsmc-comparing-the-battery-life-of-two-apple-a9s/

 

John Morrison and Austin: http://www.cultofmac.com/391574/tsmc-a9-chip-gains/

It's somewhat of a conundrum for sure. Arstechnica only saw a big difference in Cinebench 3, where as Tomshardware more in depth cinebench 3 results did not show that. Toms hardware did mention difference in ram nand between the two. Ars did not cover that possibility. Ars did however also state that these phones have many sources of components, so it's always difficult to compare fully, as there are many differences other than just SOC.

 

However none of this really relevates to AMD and NVidia, as AMD uses LPP, not LPE, and NVidia uses FF+, not FF.

 

At the bottom you can see the difference between TSMC's 16FF and FF+. The latter is smaller, but still bigger than Samsung's LPE. LPP should be a little smaller, or at least higher performing.

[Coaxialgamer]

26 minutes ago, Notional said:

It's somewhat of a conundrum for sure. Arstechnica only saw a big difference in Cinebench 3, where as Tomshardware more in depth cinebench 3 results did not show that. Toms hardware did mention difference in ram nand between the two. Ars did not cover that possibility. Ars did however also state that these phones have many sources of components, so it's always difficult to compare fully, as there are many differences other than just SOC.

 

However none of this really relevates to AMD and NVidia, as AMD uses LPP, not LPE, and NVidia uses FF+, not FF.

At the bottom you can see the difference between TSMC's 16FF and FF+. The latter is smaller, but still bigger than Samsung's LPE. LPP should be a little smaller, or at least higher performing.

i have to say , i'm rather impressed at how good intel's 14nm is compared to the competition.

[Notional]

10 minutes ago, Coaxialgamer said:

i have to say , i'm rather impressed at how good intel's 14nm is compared to the competition.

Indeed. Their pitches are smaller, so it's more compact. Both GloFo and TSMC crashed and burned on 20nm, because they used planar technology. This is why we got stuck with 28nm cards from AMD and NVidia for 5 years! 

Intel used something called trigate. Afaik, that is a technology similar to FinFets. So they got a large head start.

[Coaxialgamer]

14 minutes ago, Notional said:

Indeed. Their pitches are smaller, so it's more compact. Both GloFo and TSMC crashed and burned on 20nm, because they used planar technology. This is why we got stuck with 28nm cards from AMD and NVidia for 5 years! 

Intel used something called trigate. Afaik, that is a technology similar to FinFets. So they got a large head start.

trigate IS finfet . The only reason it isn't called  that is because of patents ....

[patrickjp93]

21 minutes ago, Notional said:

Indeed. Their pitches are smaller, so it's more compact. Both GloFo and TSMC crashed and burned on 20nm, because they used planar technology. This is why we got stuck with 28nm cards from AMD and NVidia for 5 years! 

Intel used something called trigate. Afaik, that is a technology similar to FinFets. So they got a large head start.

Tri-Gate is using 2+ fins on the gate instead of 1. It's just an evolution of FinFET.

 

7 minutes ago, Coaxialgamer said:

trigate IS finfet . The only reason it isn't called  that is because of patents ....

Intel refers to them as FinFET as well, but the Tri-Gate thing was its own invention so...

[Curufinwe_wins]

14 hours ago, zMeul said:

let me remind you that nVidia shower 2+Ghz on stage and no one was able to achieve that with the FE cards, only with the AIB custom boards

 

if AMD didn't had, and does not have, problems with Polaris, where are the cards? they showed Polaris 10 ever since January

Not quite true... 2GHz+ is easily attainable with reference pcb's under water or with the fan speeds at 100%. Just 100% is insane.

 

The low temps though were obviously just due to the gpu rendering not working the gpu as hard as gaming does.

 

And sure, it is annoying Polaris was showed off so long ago and still hasn't launched.

[dalekphalm]

8 hours ago, Notional said:

-snip-

https://zippy.gfycat.com/BraveWhichDolphin.webm 

I must admit, Linus' eyesight lines up PERFECTLY with that laser beam... he totally did it

[zMeul]

51 minutes ago, Curufinwe_wins said:

Not quite true... 2GHz+ is easily attainable with reference pcb's under water or with the fan speeds at 100%. Just 100% is insane.

 

The low temps though were obviously just due to the gpu rendering not working the gpu as hard as gaming does.

 

And sure, it is annoying Polaris was showed off so long ago and still hasn't launched.

I saw quite a few reviews but none would make 2Ghz+ stable, even on CLC

link?!

[Coaxialgamer]

8 minutes ago, zMeul said:

I saw quite a few reviews but none would make 2Ghz+ stable, even on CLC

link?!

http://videocardz.com/60923/galax-overclocks-gtx-1080-to-2-2-ghz-on-air-2-5-ghz-with-ln2

 

 

[zMeul]

3 minutes ago, Coaxialgamer said:

http://videocardz.com/60923/galax-overclocks-gtx-1080-to-2-2-ghz-on-air-2-5-ghz-with-ln2

^ not what I was talking about

was talking about reference PCBs

[DevilishBooster]

23 hours ago, dalekphalm said:

I do imagine that once Zen hits, we're going to see the Bulldozer based AMD chips being sold off at literally bargain bin pricing. They're still relatively expensive to manufacture - although since the Node is so damn old, AMD have likely optimized the fuck out of it. Never the less, AND will want to get rid of as much older gen stuff as humanly possible in anticipation of the eventual release of their lower end Zen CPU's and APU's.

I'm secretly sitting over here hoping that bulldozer prices plummet after Zen is released because then I'm going to snatch up a bunch of FX 6300's (or maybe even higher end) for my Folding @ Home Beowulf cluster instead of buying motherboards and RAM for my small collection of PGA-478 and LGA-775 CPUs.  *fingers crossed*

[DevilishBooster]

22 hours ago, zMeul said:

makes sense if you also consider the reports AMD has problems with Polaris clocks

it could be that GloFo has issues with Samsung's 14nm FinFET

Aaaaaaaaaand there's zMeul!!! LMAO!! It actually took you longer to comment than I expected.  

[shdowhunt60]

10 minutes ago, DevilishBooster said:

I'm secretly sitting over here hoping that bulldozer prices plummet after Zen is released because then I'm going to snatch up a bunch of FX 6300's (or maybe even higher end) for my Folding @ Home Beowulf cluster instead of buying motherboards and RAM for my small collection of PGA-478 and LGA-775 CPUs.  *fingers crossed*

There's a shit ton of Sandy Bridge Xeons out there for, like $70. Shame the mobo's are so expensive.

[DevilishBooster]

1 minute ago, shdowhunt60 said:

There's a shit ton of Sandy Bridge Xeons out there for, like $70. Shame the mobo's are so expensive.

Trust me, I already looked into that and the mobo prices are what stopped me. I have soooooooooo many projects lined up and I can never really do any of them because I'm broke as fuck, even with a full time job.