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

That's good to know.
But, since you already flashed the GPU to V64 you've already achieved great efficiency.
Also, you can do something similar as the germans did here (since your GPU undervolts quite well) :
https://www.hardwareluxx.de/index.p...vega-56-und-vega-64-im-undervolting-test.html

It outperforms GTX 1080 while drawing less power than 1080.
Which is pretty great.

 

Perfect, I love that article. They are great at hardwareluxx.de

I did just that, I am outperforming the stock v64 (which I cannot see any reason to purchase unless you really want the liquid edition) and the stock 1080!

What a great card, absolutely ridiculous. Although, the fan sounds like a jet-engine.

 

Top 5 Best AMD X470 Motherboards
Hardware Unboxed
Published on May 9, 2018


What Is In AMD's Combat Crate? Worth it?!
Timmy Joe PC Tech
Published on May 8, 2018
AMD sent over their Combat Crate, a new PC gaming bundle that includes an MSI Tomahawk motherboard, Ryzen 5 1600 and a MSI Armor RX 580 8gb. This turns out to be a great value to gamers and you should really have a look if you are considering buying a ryzen system. There is even a Ryzen 7 version. There will be a build and performance data up coming with all of these components so SUBSCRIBE!

 

Couple new benchmarks of the TR build. Enjoy!


http://browser.geekbench.com/geekbench3/8610593
https://browser.geekbench.com/v4/cpu/8226873
https://browser.geekbench.com/v4/cpu/8227136

 

Getting ready with your TR 1.0 benchmarks, to compare to your TR 2.0 results?

 

I wish. I won't have the money for an upgrade this year, unfortunately. Instead, my head was clear and just had physical pains today on one side, so wanted to go back and redo some benches, including getting some mods right on win 10 to get my scores up. I already can do with 4050MHz what most others take 4100-4250MHz to do with TR. But since I played with getting 3600 fully stable and slightly tighter on timings than last time, I figured why not do another run through on some of them. Evidently my win 10 tweaks did wonders for my GB4 multi-threaded score. I take issue that they practically rigged the multi-threaded to Intel's benefit on GB4 by changing it to "workloads more likely for people's use of their computer." In other words, they changed it in such a way that faster speed helps with MT and it isn't as heavy workloads unlike with some of how the GB3 multi-threaded looks. It also made my wprime score jump for 32m by a fair amount (2.153 seconds). But, I do also need a baseline for when I tear down my loop for maintenance, which might lead to lapping and going with liquid metal this round!

Edit: Also, with HPET on, I was getting a reading of nearly 4080 on the core speed. Don't know what happened there, but know that is something I saw that raised an eyebrow. But, it is what it is.

 

@ajc9988 I reckon you can squeeze a good amount out of your memory if you play around a bit more.

 

just a question of whether or not the mobo and the IMC of the 1950x can take much more, especially with the cpu clocks balls to the wall

 

It's just the time consuming nature of memory OCing and verifying stability. I was running CL14-17-17 at 3733 on my Intel platform. But, I must note that even though these are Trident Z 4133 memory modules (2 kits of 2x8GB), they are the 19-21-21 modules, not the 19-19-19 modules. Because of that, I have the 14-17-17 timings instead of 14-15-15 or 14-14-14. I tried doing 14-15-15, but I couldn't get my sub-timings as tight as with the 14-17-17. Same with the 16-16-16 timings I tried. I will attempt to go back and try tightening things from here eventually, just that I've found a sweet spot on sub-timings, resistances, and voltages at the moment with full HCI stability to 1,000%.

Very true. I need to get the waterblock pulled off and look at the surface contact to see if I can eek out just a bit more from this chip as well.

Here is the image of my timings. I used the looser of the tRFC, but the tighter didn't give much benefit and hurt performance in some tests (over-tightening can be as harmful as loose timings to performance). Now, I previously used 14-18-19 for timings and actually got slightly better performance on some things, but they were not fully stable as these timings are (for an idea, I finish TM5 in about 5 minutes or less):

 

5 minutes for tm5! dayum! mine takes like 10-11 min

 

If you need a screen shot, I can give one. Fastest I've ever run it is around 4:49, I think. Those secondary timings really speed things up!

Sent from my SM-G900P using Tapatalk

 

Edit: remember that CL14@3600 is 7.84ns and a clock cycle time of 0.56ns. So these sticks are moving!

 

this is quad vs dual channel for ya clocks and timings take the backseat on this one

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

In case you haven't seen my UMA and NUMA AIDA64 scores:

Edit: the lower scoring one is not the most recent for timings, I'll run it again.

 

@Johnksss - Catching up to your 7960X score! Just 514 points less on GB3 using 4.08GHz and ram at 3600@CL14-17-17-34 (note: I have it set to 4.05, but when I turn on HPET, it reads 4080):

 

Very nice!
Too bad I only ran that bench twice. I'm pretty sure i could have done better.

Seems like AMD does well in that bench, but most of the rest...not so much.

 

I agree with you on all counts! AMD does real well with that bench, but it is a pain trying to get reasonably good scores on MANY benches out there (I've been trying).

 

Can you guys try running Geekbench in realtime priority?

 

Already done.

Sent from my SM-G900P using Tapatalk

 

Okay, what is this? I think you are missing a few key parts to this scenario that show everything people would want to see. LOL

 

http://forum.notebookreview.com/thr...ga-polaris-gpus.799348/page-457#post-10726619
This post one page back gives the rundown on timings, etc. Ryzen Timing Checker has what the system is using, which certain things are changed from the timing calculator.

 

That screen shot should look like those then.

 

I know. I just ran it quickly and cropped to that since I already provided it previously, but I should have had them all together so no need to reference other posts. (was a bit lazy)

 

Now that I can respect!!! As lazy as I am. haha

Nothing special. Was just curious if it would complete with no errors.

 

Still has me beat handily! Now, the IMC being less on Zen I hope is fixed by 7nm. If they are able to reach about 5GHz with that gen on the GloFo/IBM process (which is what they are targeting for Server/HPC products at GloFo), things could get interesting next year. I know they have memory partners working with AMD at their campus in Texas, but I don't know the specifics and they have primarily been doing it, allegedly, for the DDR5 change coming. Although it would make sense to design it during the major revision in die shrink, it may not be ready until 2020-21 for integration, which means after they are talking about EUV being used for the lithography.

I'm a little sad (and happy) that Intel's 10nm is shaping up like it is at the moment. Whiskey and Cascade are more 14nm+++ designs, which are good, but are also reaching likely the limit at 14nm. We'll see what Ice can do probably next year (I see cannon being skipped for IceLake, if possible).

 

Intel probably want to get rid off the 14nm stock with unlikely permanent spectre fix and their 10nm will have in-silicon fix and ergo on new architecture the speedup will be immense thanks to AMD and possibly Jim Keller.
I still want AMD Chips to obliterate Intel's offerings.

 

Jim Keller played a role in development of Zen architecture, but he wasn't solely responsible for it. Two others were the lead developers of Zen uarch if I'm not mistaken... so I don't know how much Intel will benefit from having Keller around.
At any rate, it took years for AMD to release Zen since they hired Keller... it might take Intel half the time with their resources... but nothing is certain as they are having 10nm issues.

 

I think they have 5nm in their labs and Keller is working on that when it hit Moore's Law limit.
Intel is simply milking customers when AMD has used 12nm in their Ryzen.
You'll see that Intel claiming 7nm spot in next year. A direct jump I'd say.

 

So, going to respond to all of this together: 1) Jim Keller is there to get problems worked out that can help increase silicon yield and 2) work on the new CPU architecture which will not be released until 2021 at the earliest. It doesn't matter how much money you throw at the problem, some things cannot be done quicker. One of those is designing the successor to the iCore line. Not only that, you cannot move forward with 5nm and lower with the densities and yields Intel targets without EUV. Earliest EUV products will hit the shelves from mass production currently seem targeted at 2020. Intel has already bought quite a few EUV machines, but they didn't have the other components needed to perform on this lithography ready. In other words, that money has been sitting, tied up in hardware for production that they cannot use at the moment. 2019 is the earliest this will be available to market participants, which is why 2020 is EUV chips. Meanwhile, Intel has been hiring for the past year or two for their new replacement architecture. That means work is already underway. Keller being hired now and overseeing the entire silicon development division, which includes the new uarch, says that he will finish it up, but that his influence on it will not be seen until 2021-22.

Now, I have to ask what you are basing the 5nm thought on? Intel has not suggested skipping 7nm and with densities, it seems unlikely. Intel is working on 5nm, but it is theory and research related to quantum tunneling and material sciences for continuing building chips of that type and to have it ready for their fabs when the time comes. But, originally, Cannon, Ice, and Tiger were 10nm, 10nm with the reintroduction of FIVR, then the node shrink to 7nm with FIVR. Then, Intel hit that wall. Tiger got changed to 10nm. Now, with the extension of 14nm and ****ty yields on cannon, it is likely that it will be skipped for Ice Lake next year. That places Tiger around 2020, still on 10nm. To then go to 5nm would be such a huge cost, I find it dubious. Their 7nm will be about as dense as TSMC and GloFo's 5nm process anyways. So the process advancements should be fairly predictable if EUV finally gets on course.

 

https://www.asrock.com/support/index.asp?cat=bBIOS
@TANWare - looks like a new beta bios with the AGESA 1.0.0.6. Came out two days ago. I'll likely try it sometime over the next week.

 

Radeon RX 580, Worth Buying in 2018? Asrock's New Phantom Gaming RX 580
Hardware Unboxed
Published on May 14, 2018

 

The 2.0 older bios for me seems more stable.

 

@ajc9988 bruh we miss yew come talk about more Ryzen stuff

 

The older BIOS is more stable. I was able to get 3600MHz stable on the ram, just with looser timings. The real problem is the second I tried to overclock the core multiplier, it started kicking thread exception errors in TM5! I reported that immediately to Asrock. But, from what I've seen, the BIOS revisions using AGESA 1.0.0.6 for all manufacturers have errors of some form. Some it is memory stability and timings. One reported that the clocks got locked at lower C-states. It is pretty brutal at the moment, if being honest. I'll likely go back to 2.0 later today and reload my awesome settings then. I also wonder if I should try doing a bios flashback to load the new bios rather than through the bios, just to make sure it loaded properly as I've had issues with it not fully clearing prior bios editions and having issues. So I guess do the flashback and test again, then go back to 2.0.

Yeah, I've been missing ya'll as well. I only started getting back on NBR, but was gone from everywhere for awhile. I'll hit up there sometime today.

 

The beta did not last 24 hours here. The CPU overclock was not steady at all. Live and learn as they say.

 

Ryzen + 4266 Mhz Memory DESTROYS Cinebench R15
Timmy Joe PC Tech
Published on May 11, 2018
Using this Ryzen 2700x, Gigabyte Aorus Gaming 7 X470 and some trick G.Skill Trident Z Samsung B-die memory:
http://hwbot.org/submission/3852501_

 

AMD Introduces Ryzen™ PRO Processors with Radeon™ Vega Graphics Highlights
AMD
Published on May 14, 2018
At a press event at AMD headquarters in Santa Clara, California on May 10th 2018, AMD proudly launched new processors designed for enterprise-class desktops and notebooks – the AMD Ryzen™ PRO processor with Radeon™ Vega Graphics. The event was attended by several top customers and executives, press, analysts, and key worldwide partners including all three top tier commercial PC OEM vendors: Dell, HP, and Lenovo. The launch reaffirms the key factors crucial for commercial customers and are accentuated with Ryzen PRO processors are security, reliability and performance.
See how AMD Ryzen™ PRO makes a difference: www.amd.com/ryzenpro


AMD Challenges Intel With Massive Partnerships, Vega 56 Nano?!
Gamer Meld
Published on May 15, 2018
AMD is bringing a lot to the professional market. Also, 2nd Gen Threadripper is coming soon, and Vega 56 Nano Edition?!


This is EPYC™
AMD
Published on May 14, 2018
Say goodbye to antiquated theories, old science, and staid technology. Say hello to AMD EPYC™ - the server processor of today built for the datacenter of the future. AMD EPYC™ processors deliver more performance, more advanced security features, and more value than Intel Xeon.
Find out more at: https://www.amd.com/en/products/epyc-...

 

Hmmm, somehow missed this gem...

Why would ANYONE Delid THREADRIPPER? - Vanity IHS Installation
Linus Tech Tips
Published on Dec 2, 2017
Linus wants to delid... Threadripper? Let's just hope he doesn't drop it...

 

AMD shares rise after analyst upgrades chipmaker due to Intel’s manufacturing technology delay

• Susquehanna raises its rating to neutral from negative and increases its price target for AMD shares, citing the company's improving server chip sales.
• "We believe Intel's [10nm process] delay will help to maintain/improve AMD's competitiveness for their next generation of Epyc and Ryzen products," the firm's analyst writes.

Tae Kim | @firstadopter, Published 17 Hours Ago Updated 10 Hours Ago
https://www.cnbc.com/2018/05/16/amd...t-upgrades-it-on-intels-technology-delay.html

" Intel's problems in moving to its next generation chip manufacturing technology may be a boon for AMD's sales, according to one Wall Street firm.

Susquehanna raised its rating to neutral from negative and increased its price target for AMD shares, citing the company's improving server chip sales.

Intel revealed on its Apr. 26 earnings conference call it delayed volume production under its 10-nanometer chip manufacturing process to next year. Conversely, AMD said it plans to start next generation 7-nanometer chip production in late 2018 on its call.

One nanometer equals one billionth of a meter. Smaller nanometer chip-making technologies allows companies to create faster more power efficient chips.

"We believe Intel's delay will help to maintain/improve AMD's competitiveness for their next generation of Epyc and Ryzen products," analyst Christopher Rolland said in a note to clients Wednesday. "A more competitive process technology may increase the likelihood of share gains versus Intel over the next few years, a potential game changer … We also expect additional share gains to come from the ramp of Ryzen Mobile, and AMD's server product Epyc."

AMD's stock closed up 3 percent Wednesday.

The analyst increased his price target for AMD shares to $11 from $8, representing 12 percent downside from Tuesday's close.

Rolland also back-pedaled on his negative thesis that a new cryptocurrency mining chip from China would hurt AMD's sales. He downgraded the chipmaker's stock in March, predicting Bitmain's ASIC [application-specific integrated circuit] for mining ethereum would take share from the company.

"The Ethereum ASIC we previewed from our Asia trip [earlier this year] held 3x performance improvements, but fresh price hikes have destroyed its value proposition," he said. "Initial specs for the $800 E3 rig offered a >3x improvement over premium GPUs. However, in the weeks following, Bitmain discretely increased prices to $2,150 (memory costs?), cutting price/performance by two-thirds… a non-starter."

Cryptocurrency miners use graphics cards based on AMD's and Nvidia's chips to "mine" new coins such as ethereum, which can then be sold or held for future appreciation. Digital currency ethereum is up more than 600 percent over the past 12 months, according to Coinbase data.
— CNBC's Michael Bloom contributed to this story."

 

I suddenly got rich.

Man the market is so weird.. nothing happens when AMD posts good results, but when the analysts praise them, boom goes he dynamite!

 

In fact, this was one of the most negative analysts out there. Basically, he is compounding old information into his analysis, then changes his price points accordingly.

We live in a supposed efficient market. But, this analyst is compounding months old, nearly half a year old, data to come to this conclusion that should have been obvious for the whispers and talk at the international transistor meeting in December. Once information is public, it is assumed the price impact is had within 3 days of the release of that information. This is an example cutting against that due to bias in the analysts. They cannot see what is in front of their face (largely because quite a few are **** at their jobs and have their heads up their own asses). They may be busy analyzing 50 companies at a time, often within the same industry, but when you have two primary x86 producers and it takes you 3-6 months to compound information into your analysis and recommendations, it is telling. Not only that, if you look at hedge funds, etc., along with other money managers, they are not worth what you are paying them. You have the instance of a cat throwing up a toy and picking companies beating returns of seasoned analysts and a student in that comparison. You have Warren Buffet proving an index fund costs less on fees and outperforms actively managed accounts, etc. Stuff like this is why that happens, because they don't understand what the news is when it happens until they are told in very simplistic terms, like Intel announcing when the release of 10nm will be, instead of the component pieces over those months, talking about their densities on 10nm being crap, them not able to get yields on the node, only releasing 2 two-core cpus for mobile and laptop, etc. It was obvious to anyone with a brain, which is why I said over 6 months ago that cannon was likely abandoned with the announcement of an 8-core 14nm+++ this summer shortly after the covfefe release, needless to say folding in information on the security front starting in January. The market is not efficient and does not reflect real value of the company.

 

MSI X470 Gaming M7 AC Motherboard Review + Linux Test
Level1Techs
Published on May 17, 2018
No ECC Memory compatibility on this board, unfortunately!

 

Ryzen 7 2700X Review - Part 3: Memory Scaling + 5 Kits Tested
Hardware Numb3rs
Published on May 19, 2018
Third part of the review, I tested the memory scaling with 5 Kits, in 24 configurations, Productivity and Gaming

Spoiler: Memory Tuning Software, Test Software, and Components Used

Ryzen Dram Calculator by 1usmus:
http://www.overclock.net/forum/13-amd...

Ramtest:
https://www.karhusoftware.com/ramtest/

Components used:

AMD Ryzen 2700X
Intel core i7 8700K
ASUS Maximus X Z370
Gigabyte Aorus Gaming 7 X470
Gigabyte AB350N

G.Skill Flare X 3200 C14 2x8GB
G.Skill Trident Z 4500 C19 2x8GB
Trident Z 3600 C16 2x8GB
HyperX Predator 3000 C15 2x4GB
Corsair Vengeance LPX 3200 C16 2x8GB

Samsung 960 EVO 512GB

Seasonic Prime Gold 850W
Full EK Custom Loop
EVGA GTX1080

warcrab cyber 10 hours ago
good work dude, Flare x seems like the best way to go for ryzen, you can achieve 3400mhz with a little extra voltage and manual configuration of the timings. the 4500mhz kit seems like a waste of money since ryzen is limited to 3400mhz and may go to 3600mhz with loose timings. The sweet spot for ryzen is 3400mhz with tight timings.

Hardware Numb3rs 9 hours ago
warcrab cyber yes, the Flare X on my Apex and 8700K can boot in Windows at 4500MHz C19, not stable but a hell of a kit, some cherry picked 2700X can reach 3600mhz on ram but it’s hard, mine start to be unstable at 3533Mhz, 3400 C14 it’s the sweet spot before diminishing return

Marek Nowakowski 1 day ago
Awesome video! finally, someone did more than the obvious tests.

Hardware Numb3rs 1 day ago
Thanks !

 

2nd Gen Ryzen vs. Skylake X, Intel’s HEDT Lead Is Shrinking Fast
Hardware Unboxed
Published on May 19, 2018

 

AMD To Snuggle 20% Of CPU Market Share In 2018 With AMD Ryzen, Will Take 40% Market Share In Coming Years
https://segmentnext.com/2018/05/19/cpu-market-amd-ryzen/

 

Taichi X470 - Unboxing, Linux Test, OC Test, Zen+ 4 stick memory speed tests
Level1Techs
Published on May 20, 2018
On the x470 motherboard reviews I'm trying to make a point of testing performance and capabilities of "fully populated" memory speeds in a worst case (mixed memory types) config. It can make a big difference in performance if the timings don't line up just right for memory in both dual and quad stick configurations.

Spoiler: Details...

R7 2700 Max all-core Speed: 4.2ghz all core (best voltage 1.30v/soc 1.05v) Max single-core turbo 4.250 ghz.

R7 2700X Max all-core speed: 4.150ghz (best voltage 1.4250 v/stock soc), 2700x best single-core Max turbo speed (4.525ghz) w/AIO cooler (1.53v).

On Power Delivery: I think I said something like "14-ish phase" design -- it's a 12+4 phase design. (6+2 w/doublers all around -- I thought the SOC lacked doublers. They're there. Don't need them, but they're there.). This is really, really overkill and mostly not an important point on Ryzen boards thus with the current CPU lineup.

Is the 2700 a better all-core overclocker than the 2700x because those parts are lower-leakage (so less wattage)? Or is it because I lost the silicon lottery on the 2700x? Perhaps a bit of both?

Will do a separate video on this similar to the "What is SenseMI doing for me?" because those are related.

Note at one point I might have said Quad Channel when referring to memory -- if that's the case I meant Quad Stick config. Lol oops. Only X299/X399 is quad channel. I know that, just misspoke.

I think it was cut accidentally, one thing I mentioned but would like to re-emphasize is that I would have liked to see this board was a USB bios flash or roust bios recovery option. We learned how important that is on the 1st gen ryzen boards, but it isn't here on this board yet. Something to consider if you're thinking you might swap around your CPUs a lot.

UEFI tour link goes here: I'll try to remember to add it when it shows up...

 

AMD StoreMI Test : Game Loading Time
PLAYWARES TV
Published on May 21, 2018
AMD StoreMI Test : Game Loading Time

 

Do you get why this is so interesting? There are multiple parts to it. First is the TSMC vs GloFo fight. The second is the underlying process technology: 1) which materials will be used for the shrink to solve the quantum tunneling problem if maintaining silicon substrates, and 2) transistor shape (finFET vs IBM's GAA).

The first is the most obvious. TSMC is already in 7nm volume production. GloFo is right behind them. GloFo plans on using 7nm for second half 2018 volume production, then 7nm+ with compatibility with EUV in 2020 available products (theoretically volume 2H 2019). TSMC has 5nm planned for 2019 and 3nm for 2022. If GloFo isn't liking the transistor efficiency or energy efficiency with the die shrink, especially since they have some of those familiar with IBM's 5nm and GAA design (more in a minute), they may try to get to 3nm first, that way to steal away clients from TSMC as well as keeping a lead with Intel, which should be to 7nm around 2020-21. If Intel, with Keller, works on getting the 5nm going instead due to getting their butts beat with their 10nm process, then the 3nm process would be what holds them at bay. Samsung is looking at 8nm next year and 6nm following that the same year or in 2020. In fact, Samsung has laid out their process roadmap, showing that they are waiting on EUV availability, then will abandon finFET after 5nm as it no longer is effective, instead going to GAA. https://www.tomshardware.com/news/samsung-4nm-foundry-roadmap-revealed,34515.html . Because of this, you see that the three largest fabs, thanks to the GloFo/IBM fab merger, are now on track to keep barrelling ahead with little known on Intel's side.

Now, the materials used to create the transistors sub-5nm creates a huge upside opportunity. These are III-V materials, if I recall correctly. Investing in those now, before the uptick in a couple years, means you get in before the demand soars due to switching over. But there are a couple proposed materials to be used. For that reason, you want to be careful and keep an eye on it, while either investing in companies with mines obtaining the raw materials for multiple potential candidates for use or companies that will be refining it into multiple III-V materials and supplying the larger players in the industry. This is a mandatory industry change, so in that manner, due to the problems related to quantum tunneling, etc., it makes this a safer bet. But, remember, investing too early can be as detrimental due to tying up capital or to unforeseen changes, as buying too late.

On the transistor shape, as mentioned above, finFET is reaching its limits. Gate All Around (GAA) was IBM's solution for this, presented last year or in 2016. IBM showed off a 7nm transistor in 2015 and that will now be GloFo's 7nm going into volume production in the second half of this year (licensing agreement). We have Samsung already confirming they have licensed the GAA transistor type from IBM, likely in part due to the IBM/Samsung/GloFo recent workings together, which seems to have been fruitful for all involved. This means that TSMC and Intel may have to license GAA moving forward or create their own solution. Considering Intel's been working on the problems of Quantum tunneling and Moore's law for awhile, but haven't announced anything, and is having a problem at 10nm fabrication, I'm betting a license or falling further behind some of the competitors (I expect to have 7nm EPYC meet or exceed Intel's Cascade and potentially Ice lake offerings). Remember, Intel had a patent related to finFET, then sat on the finFET transistor until a couple years before the expiration of the patent, which is what gave a large acceleration on miniaturization to a degree. Now, we have IBM creating its successor, but filing patents on certain applications just last month. In other words, the get licensing fees until the 1nm use of graphene or similar 2d materials (they are looking at multiple 2D materials, and graphene still has production issues to overcome, although they have developed a way to inject a band gap, making it usable and providing near 1THz speeds, which is a mind-boggling jump). So, with 3nm being suggest instead of 5nm at GloFo, and AMD's close work with GloFo, and the fifth Zen iteration being around 2021 (a delay to 2022 shouldn't be a killer if using 3nm) currently in design (which considering AMD said they were leap frogging, meaning one team designing the shrink, one doing the refinement, then leap frogging, this means they would have the Zen+ people now working on the Zen 3 7nm+ design. But, if the Zen 2 7nm is still in design, who is working on the new Zen architecture being referred to as Zen 5? I'll note that with the new node shrink and potential use of GAA, any company that will use GAA would need as much head time as possible considering this is a complete change, and since Zen as we know it had an expiry of 2020, and 3-4 years being common for development and implementation of new architectures or processes or transistors, and GAA being announced by IBM in June of 2017, if work started immediately on incorporating it (or fairly quickly), then 2021-22 perfectly fits the use and adoption of it, and them mentioning that it is in development is not surprising considering timelines for adoption, regardless if the main teams are on it yet or not (open question)). So, there is some good things coming down the chute! ;-)

Edit: @jaybee83 - You'll like this look ahead.

 

indeed i do its incredible to see how fast AMD is already snatching away market share from Intel nowadays still, im a bit skeptical as to the generalized naming scheme of upcoming process nodes, its been shown before that intel based x nm processes actually have higher densities than competitor´s x-y nm nodes. then again, why wouldnt intel just adjust to that and tell their marketing department to lower x to undercut x-y?

interesting proposition tho, concerning those new anti-quantum-tunneling 2D materials now just have to have enough "gambling" money on the side for such investments

 

Don't think Intel is on the lead with manufacturing process since other fabs caught up.

IBM developed and demonstrated 7nm in 2015… and this high performing process is what glofo is using for Zen2.

Whereas amd will use TSMC 7nm (also made for high performance if I'm not mistaken) for vega and navi