AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

Let me address the last first, this is an AMD thread, not a competition thread so it is off topic in that respect. The originating thought of comparing it to the 8550 was OT, I guess you needed that spelled out. Please stop trolling it with bait posts, that would be nice. just like this topic does not belong in the intel thread either.

No twisting as your answer is ALWAYS get the faster Intel chip. Again stop the thread trolling. Even Intel make different speed CPU's for different systems and the power of them varies.

Lastly, I never try and make anyone look 'Dumb". This is why I try to never edit posts, I allow each others own thoughts to be conveyed intact

Again, to all, please keep the thread on topic.

 

Hey guys!

I havent posted in this thread before but it looks like I MAY push it off topic, if so, please delete!

I was just curious about a scenario I may run in the future.

Since the y410p's are getting pretty cheap on ebay, with my dead y510p I was thinking about selling the parts and buying one. I could sell the 4700mq's from my dead y510p, get the replacement y410p and get maybe a 4810mq to achieve a 3.7-3.8Ghz. This is just context for the system config.

Now to the meat.

I have the Desktop GPU adapter that will allow for Desktop GPU's to function at PCIE 3.0 x8 speeds, due to BIOS limitations only works with AMD GPU's.

I was thinking for the future to get a Vega 56 for this system, do you guys think it would be well balanced with a 4810mq/4900mq ? Would a Vega 64 be balanced or pushing it?

If you guys think I should make a new thread Ill be more than happy to, but I figure most people that keep up with Vega would be here.

Thanks in advance everyone

 

Sure the amd mobiles are behind the intel ones, but the integrated graphics are offering quite a performance boost, especially since there were no improvements on this front by intel for a while now.

Enviado de meu Pixel 2 usando Tapatalk

 

The AMD mobiles look pretty good. it scored 574 in R15 Multi and that is favorable to https://www.notebookcheck.net/Intel-Core-i7-8550U-SoC-Benchmarks-and-Specs.242108.0.html . I would reserve opinions till we have more hardware out there to look at.

 

The AMD mailer that just came in has a list of Cyber Monday / post Black Friday sales:
https://www.amd.com/en/where-to-buy/promotions

 

The one bad thing is they are still pushing the old AMD mobile chips. I understand this is in the much cheaper segments but those old chips are too slow and power hungry. I think it will only serve to give AMD a bad rep trying to sell these off. Even the adamant of AMD fan boys has to admit this.

 

It's only in the $250 range though, at that price range the competition is pretty weak too

There are lots of Ryzen based systems, keep digging through the promotions and they will pop up too.

Then when Dell's linked promotion page is 404'd, search for Ryzen, Dell has a lot of Ryzen desktop builds under a few names. Keep clicking "More Systems" to find them all:
http://pilot.search.dell.com/ryzen

 

I have an Atom nextbook that was $189 a couple of years ago, and it blows away the AMD systems I had compared it too at the time in the thin and light. Agreed none of the systems at this price point are very powerful but at this point every ounce of power counts. I am all for the Ryzen options though.

 

I think that would be good. When putting cards in sli most motherboards default to PCIe 3.0 x8 speeds anyway - and this is the same bandwidth as PCIe 2.0 x16 - so there's not really a bottleneck in terms of that link. For gaming purposes I think a 4810MQ/4900MQ is sufficient for 60fps in almost all situations, so it doesn't matter what video card you pair it with, as long as you don't want 100+fps - it's the fps target that you're aiming for that determines the CPU load, not the GPU you pair it with (...of course your GPU has to have the power to output that chosen fps in the first place though).

 

Why You Should Buy Ryzen RIGHT NOW!!!
Ryzen is having some INSANE pricing deals right now. Ryzen 3, 5, 7 and Threadripper are all a tremendous bargain right now.

 

Thanks! I had a feeling it should be ok but just wanted to get a second opinion

I wonder if we'll get any SFF Vega 56's

 

It was only mentioned and shown briefly, but there was this single fan RX Vega "Nano":
http://www.game-debate.com/news/234...radeon-rx-vega-nano-designed-for-mini-itx-pcs

 

Thanks for sharing!

For the whole adapter and everything I have a plan for modding in the future but requires an SFF GPU since the y410p is smaller and the cooling pad I plan to use with it would likely be smaller as a result.

I dont need bleeding edge performance, thats where I will push the Ranger first, the second machine will just be for MMO's or other secondary desires.

 

SFF?? EDIT, oh yeah, small form factor! They're a bit hot for small I reckon!

 

The desktop equivalent term would basically be "Open Air" for the type of implementation Ill be shooting for. Ill grab a Notepal U2 and let the laptop sit in the hinge and mount the GPU on the opposite side. Only real issue is finding a PSU that will keep the form factor flush, something like those Server PSU's maybe.

I may actually just look into replacing the y510p motherboard but thats a topic for another place I think.

 

Mother of god...those XFX cards have to be the ugliest aftermarket cards I've ever seen.

 

If you are high enough you can see Wall-E's face in XFX one.

 

Cannot... unsee...

 

Awww, they're so ugly, they're cute

The XFX RX 580 is also "cute":

 

If it performs well for the right price I dont really care what it looks like lol

 

AMD Unveils Radeon Software "Adrenalin" Edition, Coming Soon
by Nate Oh on November 30, 2017 9:00 AM EST
https://www.anandtech.com/show/12100/amd-unveils-radeon-software-adrenalin-edition-coming-soon

"For AMD, 2017 has been rather eventful, not just for new hardware releases but also for the curious application of embargoes for product photos and unboxings. But leave it to AMD to pare this trickle marketing down even further. Today, AMD is revealing the name of their annual major driver update arriving in December: Radeon Software "Adrenalin" Edition."

"So there you have it, the name for Radeon Software’s upcoming 2017 major driver update. And in fact, in all their etymological graciousness, AMD even disclosed that it was named after the Adrenalin Rose. What exactly is that flower anyway? Some kind of hybrid rose?

Botanical considerations aside, this is another entry in AMD’s now-annual cadence, joining 2014’s Catalyst Omega, 2015’s Radeon Software Crimson Edition, and 2016’s Radeon Software Crimson ReLive Edition. Even this summer featured a significant update in the form of Radeon Software Crimson ReLive Edition 17.7.2. Suffice to say, we should be thankful for the mildly shorter Radeon Software Adrenalin Edition, moving away from such unfortunate nomenclature like Radeon Pro Software Crimson ReLive Edition Beta for Microsoft Windows 10 Fall Creators Update. Or perhaps RPSCREB4MSW10FCU for short.

Stay tuned. We’ll have more to talk about than just the name soon enough."

Coming December 2017: Radeon™ Software Adrenalin Edition

Heart-pounding action with the Radeon Software Adrenalin Edition
Introducing Radeon Software Adrenalin Edition
Adrenalin Reveal Today, AMD is unveiling the name for the next generation of advanced Radeon Software coming in December 2017.

Named after the vibrant Adrenalin Rose, Radeon Software Adrenalin Edition continues AMD’s commitment to releasing major driver updates annually. The fully redesigned and supercharged Radeon Software Crimson Edition in 2015 received the highest user satisfaction rating of any AMD software ever, while Radeon Software Crimson ReLive Edition continued the 90 percent user satisfaction for 12 straight months in a row.

As a reminder, over the past three years, Radeon Software has delivered to users 70 software releases, launch day support for more than 75 games and over 50 new or enhanced features, with more than 250 million downloads across the globe. Radeon Software continues to lead the way in elevating high-performance gaming and VR experience for gamers, professionals and game developers.

 

Below they talk of the Ryzen 2000 series being based to Zen 1+ and the 3000 series being Zen 2. All to be AM4, now if truly backwards compatible it will be a boon to AMD. It would mean the next time you go to upgrade ot the next gen all you have to do is swap out the CPU. You may not get new motherboard features but you also will have a native flow to the next system, no file imports, reloading programs etc..

https://segmentnext.com/2017/12/01/...-amd-vega-11-amd-vega-supply-and-amd-ryzen-2/

 

Speculation on the future of Threadripper.

It seems AMD has roadmapped Ryzen to the next step being 12nm. This while eventually Epyc will be the first to see 7nm. At present TR has been released with the CCX used in Ryzen chips with no change indicated.

I can see where Ryzen could get 12nm as TR for a while uses up the CCX stock at 14nm. Eventually though I think the TR could see 12nm CCX's but I do not see 7nm, at least not until 7nm makes it to Ryzen.

Furthering the guess work here the 12nm will only get the silicon to usable overclock of 4.3 or 4.4 GHz. AMD really needs to get usable overclocking to 5.0 GHz or above to give real competition to Intel's lineup. If not until 7nm AMD will remain just a budget friendly, alternative, platform. A side note to this as well is who with a 1950x now will spend another $1,000 USD to upgrade to a chip with only 4.4 GHz capability? Different story if the chip can do 5.0 GHz.

 

You are forgetting that 12 nm LP is designed for high performing parts now low performing ones like the currently used 14nm from glofo.
10% increase in clock speed is a baseline. The process also allows for 15% higher chip density as well. So I think we will see higher stock clocks than 10%.

 

There is never a 100% return so 15% density increase sounds like 10% baseline is about right for the speed dincrease. Look at 10nm as it was a density increase as well but in the end there was no increase to he clock or IPC. as they say 'You don't always get what you want'.

 

No, but there is a crucial difference here:
Glofo 14nm process that Ryzen and Vega are currently built on are designed for low clocks.
The 12nm LP process That Ryzen and Vega refreshes will be made on is designed for high performing parts.

Apples and oranges in terms of manuf. process.
Look at TSMC 16nm high perfoming process and how Pascal is able to clock really high before it starts getting inefficient with power draw.
For comparative purposes, Vega is clocked much lower and it still ends up with issues of higher power draw when going over 1450 MhZ for example (nevermind anything higher) - though in that regard, AMD also didn't help matters by overvolting the GPU's.

Yes, Voltage tweaking can help offset power consumption by a lot on Vega and bring it in line with Pascal on stock or moderate overclokcs, but regardless, the fact that the 14nm process Glofo is using was designed for low clocks doesn't help Ryzen break much past the 4 GhZ barrier... and both Polaris & Vega need to be careful on clocks, lest they end up with much higher power draw (just look at custom RX 580's).
Also, Vega benefits far more in performance by having it's HBM running on higher frequencies where 10% increase in HBM requires about 5W extra - which is minimal in comparison to what the clock increases of 10% require in power draw).

Also, when looking at the mobile parts, just have a look at Raven Ridge with Vega iGP, mobile RX 580 (clocked 20% less on the core and consuming about 55% less power than the desktop variant with not too much performance loss either)... that should let you know how much the manuf. process is the bottleneck here for both Ryzen and Vega from being able to clock higher.

 

It would be nice to take clock speed per chip as a comparison but it can't. They are two different designs with their own IPC (simplistic difference(don't shoot me)). each clock for clock will generate a different power and heat requirement even if they were the same process. for actual results it is a wait and see but you will find their own 10% baseline sounds about right, and I am not going to argue their own figures.

 

My point is that you could make the claim that 10% bump is all we will get in performance if Intel was making that transition from current process.
But, Intels' process is already suited for high clocks (hence why you can expect only 10% uplift from clock increases)... AMD's is not (and I'm not saying we will get more than 10%, but there's a good chance we will, because the 12nm LP is no longer limited to low clocks, and AMD apparently wants to retain current TDP targets for Ryzen - how much Vega ends up benefiting is anyone's guess, but I'm reasonably certain the process will drop its power consumption by a large amount because it will enable the chip to run at higher speeds using much less power - this is why I said that AMD would do better to just focus on increasing the HBM frequencies a lot more on 12nm LP and give a moderate boost to the core, because current Vega tweaks show it benefits a lot more from HBM overclocking for a minimum power draw, vs core overclocking which needs a ton more power).

 

10% over your touted 16nm and you expect more over 14nm? Again I am basing this on their figures.

https://www.hpcwire.com/2017/09/24/globalfoundries-puts-wind-amds-sails-12nm-finfet/

 

I'm not sure how much AMD can be trusted on this after they clockblocked Fiji/Fury series by deliberately disabling HBM overclocking through software and vBIOS editing.

 

TSMC's 16nm which is suited for high clock speeds and Glofo 14nm suitable for low clock speeds/low power devices.
You seem to be avoiding/skipping this big detail.

Look at the disparaging differences in power consumption of Pascal (on 16nm) vs Vega (on 14nm)... they are not comparable manuf. processes because (I shall repeat again), TSMC's 16nm is suited for high clocks and high performance devices, whereas Glofo 14nm is suited for low clocks and low power devices.
If you are unaware of the Glofo 14nm low power process and it's capabilities, I suggest you read up on it

AMD will see greater benefits for both GPU's and CPU's coming from their low power 14nm than Nvidia coming from TSMC 16nm for high powered devices, allowing AMD to clock their hardware to much higher levels than before with power consumption being low (or much more comparable to Nvidia's) - you can't do that on an existing 14nm low power process because power consumption goes through the roof if you pass a certain point beyond low clock speeds.

 

I am going by general consensus here. I searched out pretty far and about everyone is estimating the same. this is not something I am just pulling out of nowhere or a smelly area of my person. unless you have some inside info that points otherwise I concede to what most others are saying.

General consensuses is 5.0 GHz will be here with 7nm and Zen2. I would be glad to be wrong here with the rest of the world but I highly doubt it.

 

That was Fiji/Fury series... Vega's HBM 2 is not clock-blocked and can be modified through Wattman.
Yes, there is a certain limit to how high you can overclock before you encounter artifacts and instability, but again, part of the reason is the 14nm low power process which.

Also, AMD will be releasing newer Vega gpu's which have its infinity fabric optimised for gaming (I'm thinking this will happen with the Vega refresh - though if AMD releases a VBIOS which can be flashed directly to do this on existing Vega's, or via drivers remains to be seen).
Not sure how much of an increase we can expect from optimised Infinity Fabric though... but we've seen that vega's compute performance is comparable if not better than 1080ti and Titan X in scientific workloads - especially when undervolted.
Plus, a lot of that pro software is not optimised for Vega (for that matter, neither are games).

Keep in mind that AMD doesn't have a final say on HBM frequencies or supply... the companies manufacturing it do... such as Hynix and Samsung (if I'm not mistaken)... but a 'proper HBM2' should have allowed double bandwidth in comparison to HBM1... and we didn't get that as apparently the 'proper' hbm2 wasn't ready, or there were some issues.

 

The general consensus applies in the case of transitioning from say Intel 14nm (designed for high powered devices) and TSMC 16nm (also designed for high powered devices) to 12nmLP (designed for high powered devices).
It does NOT apply going from Glofo 14nm (designed for low power devices) to 12nm LP.

Apples and oranges in the case of Glofo 14nm and upcoming 12nm LP.

Intel's 14nm and Glofo 14nm are NOT the same.
Glofo 14nm and TSMC's 16nm are NOT the same either (otherwise, AMD would have a process advantage... and it doesn't because the 14nm process is NOT designed for high powered devices... TSMC's is).

I won't be repeating this again.

 

For now. Let's hope it stays that way, unlike what happened to Fiji. Pffft "overclocker's dream" my rear.

 

To be fair, I think Fiji was also made on a Glofo manuf. process suited for lower clocks if I'm not mistaken.
Fiji's power draw was quite high, but it was demonstrated that undervolting greatly reduced its power consumption (by up to 100W) and brought it within 5% of Nvidia in regards to efficiency without losing performance (actually, performance sometimes increased by about 5% due to removal of thermal throttling thanks to undervolting - this was more evident on Polaris though).
So, overclocking it would have been... impractical - you'd have been better off undervolting it.

Actually, AMD can get away with Vega on existing manuf. process... trouble is, they overvolt their chips too much in order to improve yields... but surprisingly, most, if not all Vega chips released in the wild allow for a pretty decent undervolt.
The Vega 56 can be overclocked quite high on the core, and moderately on the HBM (while undervolting both core and HBM substantially), allowing it to reach 1080 level of performance while consuming LESS power than 1080 (which is something even the 1070 has trouble achieving).

One of the other issues with AMD's power draw on existing manuf. process is the number of stream processors.
Look at the amount of stream processors on Vega 56 and Vega 64...
When compared between each other in games on identical clocks, both 56 and 64 ended up with same performance, but Vega 56 still consumed less power than 64.

Obviously the amount of stream processors (compute power) is contributing to power draw - which is a problem in games that don't make any use of it.

Combine than with a factory overvolt and a manuf. process unsuitable for high clocks, and you end up with a power sucking card not suitable for overclocking.

Plus, unlike Nvidia, AMD does not have auto-voltage tuning technology integrated in it's GPU's.
Yields depend on the manuf. process as much as anything else.
Glofo has been having problems in supplying AMD will enough yields for a long time now... which is why we saw those articles that posit AMD might go to TSMC for Navi.

 

Yeah but HBM overclocking on Fiji, not core. If I had to guess, AMD "broke" and never fixed it on purpose for planned obsolescence reasons because a properly tuned Fury X is dangerously close to Vega 56.

 

You guys are rather optimistic about AMD's gpu and CPU.

Threadripper is done in such way that I do not see it as comparable with similar EPYC processors.

I am excited about faster clocked EPYC 2 processors but so far... nothing really else from AMD. Ryzen is good for low power workload but overclock vs overclock, 8700k is better for general purpose compared with R7.

Regardless, EPYC is very well done and thats commendable. Xeon Scalable is comparatively a bit worse to be honest.

To be honest, in the future most thing will hopefully be done on clusters in the cloud. Ideally anyways. Computing is becoming a service and I think thats really good! Imagine renting elastic GPUs where you pay very little on idle and instantly scale up when you are gaming. Thats the future I want to see.

 

Example? Both GP102 and Vega are not good at double precisions....Also, yes, vega isnt optimised on some softwares or even supported. That why people dont buy AMD for this kind of things imo.

If you need type of compute, get/rent P100/V100s. Deep learning often runs on gp102 based servers I believe.

 

Optimism would be misguided... we're talking about the differences and limits between manuf. processes of Intel and AMD which are not the same even though they are technically both 14nm.

Comparing current Glofo 14nm (low power) process on which Ryzen is made to Intel's i7 and i9's on 14nm refined process (for high performing parts) is not exactly a good or fair approach without addressing the differences between manuf. processes (something which none of the reviewers are doing).

The 12nmLP process should equalise for discrepancies in clock rates and allow Ryzen refresh to reach if not surpass Intel early next year on those grounds alone.
For IPC improvements (which at this point are about 5% between AMD and Intel at same clocks) we will need to wait Ryzen 2 most likely.

 

I hope not, I like owning and tweaking the hardware, plus with cloud computing services you're gonna have built in latency when it comes to gaming. (I don't really like renting stuff either, I'd rather own it.)

 

I almost agreed with everything you're saying below!

But the cloud is not 'ideal' for anything we need done locally on massive scales (including 'gaming' workloads).

I can't imagine renting anything with my/my clients' data on it. Either I have full physical control or it's time to look for another job.

 

In your argument, I think you're missing the point that the are not even both technically 14nm (or any other node you wish to use for discussion). These are highly disparate processes for getting to a similar endpoint (a processor).

I don't see anything similar between them when I'm comparing them side by side in my workloads - you can 'feel' the decisions made to get to their own goals - even if I don't have the slightest clue of how they are doing it in practice.

Intel has a certain kind of feel, AMD has their own and the handheld market has their own too. These differences are there for the decisions made on every aspect of bringing usable silicon to market by each of them. None are superior in an absolute sense, but for me (I value snappiness and high productivity equally; when available); Intel is the clear leader by a mile.

The logic you're presenting is good until you throw the above wrench into the mix; the actual practices and methods of bringing a piece of silicon up to sale value is vastly different between any two or more manufacturers - and this difference also applies and compounds with each iteration each manufacturer makes on their respective processes too.

In the end you may be right. However, history has taught me that AMD is the laggy second cousin to Intel and everyone else is a dimwitted relative by marriage (to the same side of the tech pool). Even R7 feels like that to me (1800x O/C'd to just over 4GHz) vs. what Intel can offer.

What I believe is that AMD thinks what it is offering is enough (overall) and will continue to provide that lessor product we all wish more from.

 

AMD’s James Prior talks (Ry)zen 2 and Vega 11
Published: 1st Dec 2017, 12:59 GMT
https://videocardz.com/74260/amds-james-prior-talks-ryzen-2-and-vega-11

Some interesting tidbits have been shared during OverclockersUK live stream with special guest from AMD: James Prior (Senior Product Manager)

AMD Vega 11 is integrated into Raven Ridge APU

The mysterious Vega 11 is not a GPU by itself. It’s a solution for AMD Raven Ridge APUs with 11 Compute Units enabled. James Prior confirmed that Ryzen APUs offer up to 11 Compute Units. So far AMD only released two mobile APU variants, which feature either 8 or 10 CUs (Vega 8/10 Graphics). That said, the chip with 11 Vega Compute Units would be the top tier Raven Ridge APU. No details about desktop APUs have been shared.

AMD RX Vega 56 and 64 to receive an increased supply

It has been confirmed that RX Vega stocks will be increased shortly. This will allow retailers, such as OverclockersUK, to adjust the price accordingly. Our sources have confirmed that AMD is finally supplying partners with Vega chips, which will allow them to introduce custom SKUs in satisfactory number, while reference designs will no longer be produced.

AMD (Ry)zen 2 will use AM4 socket

James Prior reassured that AM4 socket is here to stay (till 2020). The work on Zen 2 has already begun when fundamental parts of Zen 1 were already known. The important thing here is to distinguish Zen 2 from Zen 1 tick-tock process. The upcoming Ryzen 2000 series are likely to use refined Zen+ architecture. A die shrink and architecture optimizations are to be expected. So the Ryzen 2, or more precisely Zen 2 might actually arrive with Ryzen 3000 series, while Ryzen 2000 (or Ryzen 1×50) will use refined Zen1/Zen+ 12nm process instead.

If everything goes according to the plan, forward compatibility for Zen+ and Zen2 will be available with a simple BIOS flash on existing AM4 motherboards.

Questions about HDMI 2.1, Navi architecture and partnership with Intel have also been asked, but the answers were not very satisfying. Watch the full video here: Watch on the web page, or:

https://www.facebook.com/video.php?v=10155853543333349

 

Truth being I can not see upgrading my 1950xthere is nothing I have, other than stress testing benchmarks, that come close to a high stress on the CPU. Maybe by the time an upgrade comes to market something will be out that makes the CPU show some weakness but I somehow doubt it.

 

And what is AMD's anwer to this?
NVIDIA Announces “NVIDIA Titan V" Video Card: GV100 for $3000, On Sale Now

 

Pro GPU on 12nm LP.
AMD will be remaking Vega on 12nm Leading Process too... with likely much better TDP and higher clocks in relation to 14nm low power process like previous Vega.

Pro software showed Vega's (pretty high) performance in relation to Pascal (and that's on an inferior manuf. process)... so, I don't think Vega refresh will have too much trouble keeping up with Volta in that regard but with improved TDP - of course, PRO software needs better optimisations for AMD if you want a level playing field - otherwise, you can't expect for one company to brute force its way and win and then complain on some of its inefficiencies when same roadblocks don't apply to the other company).

 

Leaked NVIDIA TITAN V Benchmarks Show Volta GPU Demolishing All Competitors

"Unigine's Superposition benchmark also yielded some impressive numbers. At stock speeds, the TITAN V scored 5,222 in the 8K preset, and 9,431 in the 1080p Extreme preset. The latter is particularly interesting—famed overclocker Kingpin had previously taken a GeForce GTX 1080 Ti, stripped off the heatsink and bathed the card in liquid nitrogen (LN2), and overclocked it to 2,581MHz, which resulted in a score 8,642 in the 1080p Extreme preset. The TITAN V scored nearly 800 points higher."

"Of course, NVIDIA is not in a rush to bring Volta to the consumer market, as AMD has not fully caught up with Pascal (Vega comes close). The silver lining to that is it gives NVIDIA time to tweak things and flesh out better drivers for when Volta does infiltrate the mainstream gaming sector. Based on what we've seen here, we can hardly wait."


And from the RED camp... AMD quietly made some Radeon RX 560 graphics cards worse

"You have no idea what you’re getting when you buy a Radeon RX 560 unless you’re nerdy enough to be aware of this issue and know the key deep-dive GPU specs to watch out for. And the cut-down models aren’t even cheaper."

"Consumers lose when shenanigans like this occur. No wonder AMD kept the change under wraps. It’s good to see the company moving quickly to respond to the outrage over the discovery—but it would be better if this issue never occurred to begin with. What remains to be seen is how clearly the cut-down cards are described going forward."

 

https://www.tweaktown.com/news/60100/amds-new-ryzen-7-2800x-teased-12c-24t-up-5-1ghz/index.html

Ryzen refresh apparently topping out at 5.1 GhZ with increase in core count.

Sounds rather odd.
I don't think AMD mentioned anything about increasing core count on the refresh.
Then again... it appears to be a rumour.

Not Ryzen 2... that won't be ready until the end of 2018... the Ryzen refresh is being released in early 2018

EDIT:
On another note, the 12nm LP does come with 15% higher density.
I doubt we will see an upgrade in core count until 7nm, but if AMD decided to up their game along with the clocks, then kudos to them.
We won't know conclusively though until we get closer to Ryzen+ release date, but I wouldn't be surprised if we get the same core counts we have now with those clock increases (they do look possible with 12nm LP).

 

Pretty sure thats not true. 6 core CCX would need revamp and upgrade of infinity fabric.