Ryzen 2 slide in KitGuruTech video

[raghu78]

Is a CCX 30mm2..?

If so they can add one within the same die area as Zeppelin thanks to 12nm 15% better density...

https://www.pcper.com/news/Processors/AMD-Details-Zen-ISSCC

Zen CCX is 44 sq mm. btw Pinnacle Ridge on 12LP will use high performance libraries (with higher track height and/or relaxed CPP) which will mean the CCX will be larger. Thats why I do not expect to see more than 8 cores on PR.

 

[ZipSpeed]

How does the old adage go? If it sounds too good to be true... Still, one can dream and in the unlikely event this is true, the 2800X will be replacing my 3770K as my primary rig.

 

[Excessi0n]

I'd believe that for 7nm Zen. 12nm? Not a chance in hell.

 

[Campy]

We can only hope, but it does seem like an unbelievable step forward.

 

[Justinbaileyman]

You sure he isn't mixing up x399/x499 die shrink refreshes??

 

[beginner99]

I'd believe that for 7nm Zen. 12nm? Not a chance in hell.

It's a fake but I agree. Was somewhat believable for 7nm on new "high frequency process" and launch mit 2019. But on 12 nm in 2018? Never ever.

 

[piesquared]

It is more likely fake news than not, but it's not impossible with 12nm. The die size would increase slightly sure but that may be designed into the new WSA. If their next gen GPUs are being fabbed at TSMC then that would leave capacity to fill at GF and the slight increase in die size may be a non factor under the new agreement. There is also nothing forcing AMD to stick to a 95W TDP, they could be 125W TDP chips. New boards may even allow 140W chips for the top performing bins.

 

[raghu78]

It is more likely fake news than not, but it's not impossible with 12nm. The die size would increase slightly sure but that may be designed into the new WSA. If their next gen GPUs are being fabbed at TSMC then that would leave capacity to fill at GF and the slight increase in die size may be a non factor under the new agreement. There is also nothing forcing AMD to stick to a 95W TDP, they could be 125W TDP chips. New boards may even allow 140W chips for the top performing bins.

Ryzen and Vega refresh are both fabbed at GF 12LP.

http://www.tomshardware.com/news/amd-ryzen-vega-12nm-lp-2018,35502.html

At 7nm Navi is fabbed at TSMC and Zen 2 desktop and server (Rome) is fabbed at GF.

https://www.digitimes.com/news/a20171023PD201.html

I am sure the ryzen 2 slide is fake. I think we will see significantly higher clocks and improved memory controller with lower latency in Pinnacle Ridge. PR will be a 8 core die.

 

[piesquared]

Ryzen and Vega refresh are both fabbed at GF 12LP.

http://www.tomshardware.com/news/amd-ryzen-vega-12nm-lp-2018,35502.html

At 7nm Navi is fabbed at TSMC and Zen 2 desktop and server (Rome) is fabbed at GF.

https://www.digitimes.com/news/a20171023PD201.html

I am sure the ryzen 2 slide is fake. I think we will see significantly higher clocks and improved memory controller with lower latency in Pinnacle Ridge. PR will be a 8 core die.

You are probably right. Although a tuned, optimized nd higher clocked 8C/16T Ryzen 1 will be a beast in it's own right.

 

[CatMerc]

I believe the clocks being possible more than I do the core counts. The leaked roadmaps would have said something other than "perf uplift" if there were more cores and the layout changed from Summit.

That said 4.4GHz XFR is a good target to expect.

 

[IEC]

On the one hand, AMD under Lisa Su has shown a penchant for conservative promises... with surprise on the upside. No one expected Ryzen to be such a success (despite its quirks) after the disaster that was the Bulldozer architecture.

On the other hand, those specs are shockingly unbelievable. I think they're more appropriate for a "Zen 2" in 2019 than a 12nm Ryzen refresh.

We'll find out in February, I suppose.

 

[Markfw]

How does the old adage go? If it sounds too good to be true... Still, one can dream and in the unlikely event this is true, the 2800X will be replacing my 3770K as my primary rig.

If true, both my 1700x and 1800x will go bye-bye in favor of the 2800x.

 

[Abwx]

https://www.pcper.com/news/Processors/AMD-Details-Zen-ISSCC

Zen CCX is 44 sq mm. btw Pinnacle Ridge on 12LP will use high performance libraries (with higher track height and/or relaxed CPP) which will mean the CCX will be larger.

GF state 15% better density and 10% better perf in the same slide...

Besides said HP libraries are essentialy memory cells used either for the L1/L2 and for various registers, but adders, multipliers and other ALUs are completely different beasts.

With said 15% density improvement 203mm2 are enough for a 12C.
Even at 220mm2 it would make sense for a server dedicated CPU, after all that would yield 50% more cores for only 16.4% bigger die.

 

[raghu78]

GF state 15% better density and 10% better perf in the same slide...

Besides said HP libraries are essentialy memory cells used either for the L1/L2 and for various registers, but adders, multipliers and other ALUs are completely different beasts.

With said 15% density improvement 203mm2 are enough for a 12C.
Even at 220mm2 it would make sense for a server dedicated CPU, after all that would yield 50% more cores for only 16.4% bigger die.

GF12LP is not an optical shrink. Its a library optimization with some transistor level enhancements.

https://www.anandtech.com/show/11854/globalfoundries-adds-12lp-process-tech-amd-first-customer

"The new 12LP relies on the groundwork set by the 14LPP, but uses 7.5T libraries, which is one of the ways that enables GlobalFoundries to shrink die sizes by increasing transistor density. Since the library contains different elements, IC developers have to “recompile” their designs to take advantage of the process. Meanwhile, since the 12LP and the 14LPP are very similar, for GlobalFoundries’ existing customers migration path to the 12LP is pretty straightforward."

My guess is GF 12LP 7.5T provides performance better than GF 14LPP HD (CPP=78nm ,9T) . For maximum performance 12LP will probably bring a combination of higher track height, relaxed CPP and transistor level enhancements. I am thinking CPP=84nm, 10.5T, taller fins and improved drive current to help AMD hit significantly higher clocks with Pinnacle Ridge.

 

[Abwx]

https://www.anandtech.com/show/11854/globalfoundries-adds-12lp-process-tech-amd-first-customer


.

The new 12LP relies on the groundwork set by the 14LPP, but uses 7.5T libraries, which is one of the ways that enables GlobalFoundries to shrink die sizes by increasing transistor density

They say that transistors density is increased, 7.5T libraries being a mean to do so but it s not the only one at work...

 

[Harmaaviini]

Those prices for 4+ Ghz baseclock 12-cores should have ended any discussion of legitimacy before it even started...

 

[raghu78]

They say that transistors density is increased, 7.5T libraries being a mean to do so but it s not the only one at work...

As far as I know both TSMC 12nm and GF 12nm are not optical shrinks which means Minimum Metal Pitch (MMP) is not changing. I think we will likely see a Contacted Poly Pitch (CPP) of 84nm at 12LP and 7.5T/9T/10.5T options . There are transistor level enhancements according to GF which are enabling higher performance at lower track height compared to 14LPP. I would like to point out 12FDX specs which are no way related to 12LP but could serve to give us some insight on the probable choice of 84nm CPP

https://m.eet.com/content/images/eetimes/1 7 12 14 copared x 800_1505972923.jpg

 

[moonbogg]

So is 12 cores the new standard then? 12 is ridiculous, especially with dual channel ram. It can't be 12 because I made a thread claiming that 8 cores are the new standard.

 

[raghu78]

So is 12 cores the new standard then? 12 is ridiculous, especially with dual channel ram. It can't be 12 because I made a thread claiming that 8 cores are the new standard.

dude that slide is most likely fake. don't get too excited. Its only a couple of months before we know what PR is. imo PR is 8 core with significant clock increase and some tweaks probably like a better IMC with higher DDR4 speeds support and memory latency reduction.

 

[Mopetar]

I'd be happy with those clock speeds at the current core counts and no uplift in IPC. These slides are too good to be true in every conceivable way.

 

[Topweasel]

If true, both my 1700x and 1800x will go bye-bye in favor of the 2800x.

Yeah complete refresh of all my systems if that happened. I mean even in 2019 when Zen 2 hits, if it's up to 12 or 16 cores $200-$400. I will be replacing my 1700, 3930 and 4770(which is in an ITX system, 16c in a ITX would be insane).

 

[Markfw]

Yeah complete refresh of all my systems if that happened. I mean even in 2019 when Zen 2 hits, if it's up to 12 or 16 cores $200-$400. I will be replacing my 1700, 3930 and 4770(which is in an ITX system, 16c in a ITX would be insane).

I just upgraded my 3930k with a E5-2695vs Xeon (6 cores@3.4>14cores@2.5). And I have a E5-2658v3 ES (12 cores/24T) coming for a new box. It was only $190, so I couldn't resist.

 

[StinkyPinky]

I think AMD will be looking to tweak the IPC by a few percent and boosting clocks by 10-15%. I can't see them adding more cores at this point.

 

[IntelUser2000]

I think AMD will be looking to tweak the IPC by a few percent and boosting clocks by 10-15%. I can't see them adding more cores at this point.

The low resolution slide and the earlier AMD roadmap makes this quite suspicious.

That said, let's see what happens.

It's more likely they'll get a 12 core out than increase clocks by 10-15%. 4GHz is at the level where transistor level enhancements may not translate into clock speed increases. It's also where clock related thermal issues increase dramatically.

With Kabylake and Coffeelake Intel increased base clocks by reducing the between it and max achievable overclock. Sandy Bridge at 32nm could do 4.4GHz on air. 5.2GHz or so Coffee/Kaby does today is only 18% increase over that. Transistor drive current wise current 14nm++ transistors are greater than 50% better.

 

[Mopetar]

I think AMD will be looking to tweak the IPC by a few percent and boosting clocks by 10-15%. I can't see them adding more cores at this point.

I’m more pessimistic on clock speed increases and think they should be getting double digit IPC gains. Zen is a new architecture and that usually means more low hanging fruit for the picking.

Think back to Intel moving from Nehalem to SB and SB to IB where they were pulling down large IPC improvements. That seems more realistic to expect unless Zen is already highly optimized or AMD lacks the engineering talent to make a lot of improvements.