Basically everything like expected from previous leaks but I'm surprised about the driver updates coming directly from AMD. That is a big issue for current gen Ryzen-Laptop users...
AMD put extra firmware into their code so that they would be able to do a more direct "to the lineup" if required for certain update paths, this is why Zen 1 was able to "drop in" to a 3xx or 4xx board and also got a bit of an enhancement when the 2xxx chips launched some of this helped (sometimes a fair amount) the 1xxx chips to go higher speed or higher dram speeds/stability.
This. I have a Lenovo Ryzen 2700U notebook... And the driver updates have been bullshit thus far... So this is very much a welcomed change for the better.
Actually interested to see how the 3750H stacks up against my 2700u, especially if it supports memory faster than 2400mhz in dual channel... I might. Might... Be tempted to upgrade if it gives me that extra "kick" in my games. - AMD might also be able to fix the high idle power consumption too?
Had a 2in1 with the 9220 last year for a brief period... Was capable of playing skyrim at 720P@30fps, so it's fine for older/simpler games, but the display was almost unusable... And the EMMC drive made it feel like a slug so returned it to Dell for a full refund.
"2700u, especially if it supports memory faster than 2400mhz in dual channel" - it already does. I have 3 2700U laptops running 16GB of 2666mhz ddr4. These are Elitebooks 755G5 which came with 1x8GB 2400mhz, take that out and replace it with 2x8GB 2666mhz and voilá, it works.
Officially AMD only supports 2400Mhz on Ryzen mobile. - So anyone who runs out and grabs 2666mhz memory after reading your post is likely going to be gambling as there is zero guarantee the BIOS will upclock to 2666mhz.
I have the Lenovo E585 myself... Doubt it will support anything more than 2400mhz... And I am not about to sink a few hundred bucks to find out.
True. It is not a safe bet. But there are loads of users finding out that it does work. My post is merely an indication for those who are willing to bet.
And, if you stick those dimms in a laptop limited to 2400, it will work at 2400 speed most of the times. The risk is small. Those 2666 dimms I have running in the EB755G5 will run at 2400 speed in the EB650G4. I tried it, of course.
Why would you be surprised about AMD providing an official Ryzen Mobile driver ? it was clearly necessary, after OEMs proved incapable and unwilling to respect and support their clients ....
I kind of wish desktops could be "from the source" type driver updates that do not rely on MSFT way of forcing certain drivers that "break things"
They should work with MSFT, Intel, Nvidia etc to make sure that the makers of the hardware that require specific BIOS or Driver sets to function "best" for whatever reason are left the #$%#$% alone.
Actually, I thought all the leaks said 7nm EVERYTHING here. We get crap 28nm (in 2019? LOL, 7nm with far more perf would have gotten higher ASP's), and 12nm Ryzen chips. ER, UH, not what I expected. I expected GPU info at 7nm with roadmaps/launch dates, and 7nm CPU launches TODAY. What leaks were you reading?
AMD is bound by contract to buy X wafers from Global Foundries.
If you dont know what that is in short in 2009 AMD signed an agreement when it spun off its fabs to form a new fabrication company called Global Foundries that sets fixed pricing as well as volumes of silicon wafers AMD has to buy from Global Foundries.
It in part nearly bankruped AMD beause they could not source cheaper wafers OR reduce wafer purchases when Bulldozer failed as hard as it did. They had to purchase wafters, used or not! This was a huge problem when Global Foundries COULD NOT scale below 28 NM. Eventually cancelling their own in house project and instead purchase/licence the tech for 14nm FF from Samsung, which was a low power process and truly not suitable for what their primary customer of AMD wanted, a High Power process for their new Microprocessors.
This is why Ryzen first gen hit a wall at 4ghz. It's not heat, it's not voltage, it isnt the archetecture. It's fundenmentally the node its fabbed on. the 14NM (or the improved 12Nm) process that GF is using is fundementally a Low Power node from Samsung, it isnt designed for efficiency or to be able to hit high clock speeds. What it IS designed to do is be VERY VERY efficient at lower power constraints. which makes it ideal for mobile parts that dont need as high of a clock speed where power is the big limiting factor.
While the Wafer Supply Agreement has been ammended multiple times in part to keep AMD from going under, the 2016 Ammendment also allowed AMD to source some wafers from other sources. The 6th ammendment done in 2016 it is currently set to expire in 2020,
What that means for AMD is while the ammendments have allowed them to change their cost and source wafers outside of Global Foundries (It's the entire reason they have 7nm chips sourced from TSMC. Global Foundries cancelled their 7nm node last year) Even though they are having to pay Global Foundries for using TSMC wafers!
It doesn't remove the requirement to buy wafers from Global Foundries completely though, they still MUST purchase or recieve a certain inventory of wafers. What do you do if you MUST buy a raw material by contract? You turn it into something you can sell. Ideally for a profit, but even for a loss would be better than just absorbing the cost and getting nothing out of it. Assuming the final sale reduces the amount lost.
Ergo, this is another reason these APU's are on Zen+ 12nm process rather than 7. They probably could put them on 7, but they NEED to use wafers from Global Foundries to reduce their losses. As far as why they are making 28 NM chips. Well they are better than the competitions (intel's) at this power envelope, and they probably still have more wafers they are contractually obligated to buy. Even if they still lose some money on the end product. It's still something better than just eating the full cost of the wafer with nothing to show for it. Some revenue to reduce the loss is better than a full and total loss.
It is an odd move, since it is the laptop which would benefit the most from a lower power envelope. Seems AMD feels confident about these parts being competitive enough to win some designs.
Because they want to bring back their server CPU market. This is the reason their desktop parts are also on their latest node which will happen again this year at 7nm
I don’t know that I’d read much into the mobile announcement. Raven Ridge came 6 months after desktop Zen, and these are the 12nm versions. With the extra lead time designing an APU, they likely wanted to stick to the older architecture and just get the die shrink done. The big problem I see is that the mobile series will be a generation behind despite being in the 3xxx lineup.
Yes, that much extra work for OEMs too - designing for a new platform, power and cooling changes, connectivity. Well, looking forward to the presentation by AMD!
Oh and since AMD definitely lags behind Intel in laptops, a 7nm surprise would have been awesome. They do know what they're doing I guess,so yeah, let there be competition.
Most likely they want to fulfill wafer orders from GloFo so that they are not penalized further due to their existing agreement. They could have skipped 12nm and Zen+ and go directly to 7nm and Zen 2 in the mobile space, but that would leave GloFo holding their... proverbial instruments. With the GPUs also soon moving to 7nm they could only order I/O dies from them, and I don't think the revenue will be huge from them.
On the other hand it is GloFo that does not fulfill their end of the agreement : they decided to delay 7nm indefinitely, but they surely cannot ask AMD to do the same. Does their wafer supply agreement with AMD have any clauses about newer nodes and failure to deliver them? I have no idea, but I believe there should be.
No one that has knowledge of AMD mobile (laptops) expected 7nm mobile APU this year. AMD is always one year behind on mobile chips. In 2020 it's likely we get 7nm AMD mobile. I only wonder when i can find an AMD 3000 mobile in a local store, often it takes months after AMD releases a mobile chip until you can buy them in local stores. There are the few presented models that you can order online and it then takes a long time before someone can find them in local stores, maybe i can find them locally at the second half of the summer. I really hope that AMD mobile 3000 hits stores faster this time around.
@tekniknord ... AMD is "one year behind" who exactly ?! :) Itself ? Because when compared to Intel mobile CPUs, AMD is one or two years ahead in every metric except maybe idle power consumption ....
AMD mobile is lagging one year after their desktop product releases. I didn't mention Intel at all because my answer was related to AMD desktop releases vs mobile. I guess i wasn't clear enough.
AMD's 7nm mobile chips in 2020 could be a massive improvement over today's announcement. It could double core counts to up to 8. Power efficiency increase could result in lower TDPs, needed for stuff like fanless Chromebooks. Along with Navi graphics, maybe it will have hardware decoding for the AV1 codec.
its basically almost no contest market segment there right now... AMD cant just poof a part for it...so they repurpose an existing part to fit that segment .... remember for chromebooks .... the part should be low TDP as well as "cheap" enough for it... or else it defeats the purpose. ... current crop of products are low on power ..but being on cutting edge tech ..they are quite costly to offer for such a segment.
Actually Zen is NOT good for low power systems. The cores are efficient, but Infinity Matrix eats so much energy that it's almost impossible to design something usable below 15W. That's why AMD is reviving older tech.
Zen was a server product from the start - perfect for 150W+ systems. Consumer desktops are already pushing this architecture out of comfort zone. To be competitive in laptops, AMD would have to design purpose-built CPUs and the simple fact is: they can't afford that right now.
Yeah.... But they could do a more simplified Zen apu that would work fine at this power level. Single CCX that has some lower powered GPU and simple lower power connection to it.
Generally though they do not seem interested in mobile...could be a lot of interest in well though out lower power Zen derivatives. I think AMD and Intel are slowly allowing too much to get lost to ARM as far as the broader computing market. Sub 10w stuff should be on newer even if not latest process at least and use best core IP levels they can. Stuff like excavator against more upto date ARM chromebooks at similar price just makes AMD look that much worse.
My guess would be that the difference between a 35 watt part, and a 15 watt part, with identical specs, is that the 15 watt part is set for more aggressive power management. So it will still yield less sustained performance, although peak performance will be identical, because it won't be able to keep everything active all the time.
Precisely this. Hopefully the 35W sustained performance gain will be good enough on mobile this time around, maybe like the 35W desktop AM4 parts. That would make for some really great gaming laptops.
Cinebench R15 nT ("CPU"): Core i5-8250U vs. Ryzen™ 5 3500U: 524 vs. 651 (24%/1.24X faster for AMD) Core i7-8565U vs. Ryzen™ 7 3700U: 619 vs. 688 (11%/1.11X faster for AMD) 3DMark® Time Spy ("GPU"): Core i5-8250U vs. Ryzen™ 5 3500U: 399 vs. 907 (127%/2.27X faster for AMD) Core i7-8565U vs. Ryzen™ 7 3700U: 444 vs. 967 (118%/2.18X faster for AMD)
8565U is a 2018 model and 8250U was launched in 2017. Intel will soon release 9th gen models and that's what these Ryzens will have to compete against.
Also, why would anyone care whether AMD managed to squeeze more power from an ULV chip (even less so about GPU performance)? Previous gen already had plenty. The big question here is: did they improve efficiency and power management. Laptops built around 2500U and 2700U have atrocious battery life.
esting by AMD performance labs as of 12/4/2018. “Battery life” defined as hours of continuous usage before the system automatically shuts down due to depleted battery. Video playback tested according to Microsoft WER methodology, while “general usage” is tested via MobileMark 14. Results presented in hours, in order of: 1st Gen AMD Ryzen™ 7 2700U Mobile Processor (100%) vs. 2nd Gen AMD Ryzen™ 7 3700U Mobile Processor. General Usage: Ryzen™ 7 2700U: 8.1 hours vs. Ryzen™ 7 3700U: 12.3 hours (51% longer) Video Playback: Ryzen™ 7 2700U: 6.9 hours vs. Ryzen™ 7 3700U: 10 hours (40% longer) Ryzen™ 7 2700U Test System: Lenovo IdeaPad 530s, Ryzen™ 7 2700U, 2x4GB DDR4-2400, Radeon™ Vega 10 Graphics (driver 23.20.768.0), 1920x1080 AUO 403D 13.9” panel, 512GB Toshiba KBG30ZMT512G SSD, 45Whr battery, 150 nits brightness, Windows® 10 x64 RS4. Ryzen™ 7 3700U Test System: AMD Reference Motherboard, AMD Ryzen™ 7 3700U, 2x4GB DDR4-2400, Radeon™ Vega 10 Graphics (driver 23.20.768.0), AUO B140HAN05.4 14” panel, 256GB WD Black WD256G1XOC SSD, 50Whr battery, 150 nits brightness, Windows® 10 x64 RS5. Results may vary with drivers and configuration. RVM-164
I think this test is a little unfair - since GPU on AMD is more of discrete GPU until like iGPUs - maybe the test should be against say 8705G in Dell XPS 15 2in1
Paring these APUs with a discrete GPU like the 560x is completely pointless. When will OEMs learn to simply let the APU shine on its own instead of crippling it with worthless extra silicon?
Hopefully that AMD driver fix works, forcing OEMs to do driver updates was an incredibly dumb thing for AMD to do in the first place that killed a LOT of enthusiasm for the platform once the poor performance was shown off, and there was no reason for it other then AMD being lazy. Hopefully their updated drivers will fix a lot of the longstanding mobile issues.
Not sure if it was forced, but typically I was under the impression that as a silicon vendor, you may prefer OEMs to handle releases since every mobile platform is technically different and OEMs would "be the best equipped" to do their own custom internal testing to catch bugs that a general release could cause on their particular systems. Obviously we understand now that in practice, OEMs dont do anything at all here, even up to a year later (I'm looking at you HP), so thank God AMD is going to do a general drive release. I gather tech-savvy users will take the risk for better perf and power management and fix their own issues, and the layman won't even notice. Win-win.
Will AMD have to be careful with all the different laptop chassis cooling abilities out there when they are sending drivers directly to laptops and skipping the OEM?
Why won't AMD get it... and just make what most mainstream users want and need. A 4/4 or 6/6 4ghz cpu with a vega 10 or beefier gpu, with gddr or hbm. Sigh. One day... one day.
For AMD, APUs are low margin products. Zen CPU dies of the same size sell for ~€200 on average, and go for both desktop and servers. APUs, on the other hand, sell for ~€100 on desktop and mobile- need to be pushed into Intel dominated mobile market, with volume being not high. And mobile market has a lot of inertia, so it is not easy to enter. So, it makes sense for AMD to do it slower. They still have a lot to fix- like drivers, switchable graphics support, adding LPDDR support needed for small/high end devices, making video playback more efficient, etc. And, like I said, bigger/faster money is in Zen dies- which already had a 12nm refresh and now get Zen2 + 7nm. Lower margin product- APUs- will follow the path of 12nm refresh of same silicon, that offers a better product at minimal investment, and only then get whole new CPU and iGPU design at 7nm. Probably again for a new product (7nm) and then a refresh cycle (7nm+ or 5nm), which is a tick-tock strategy of all AMD products.
AMD have unfortunately followed Microsoft Operating System (OS) support blindly for at least two generations, even while Linux was captiring the Server, Data cwnter and Cloud computing Market left to Intel for plundering. Even ARM has started serious entry into the Server market.
Now the company is looking at Chromebooks, at a time when seveal Chromebook manufacturers are eyeing ARM based systems.
By the way, more than 52% of all school systems in USA are deploying Chromebooks, and Amazon has announced that for the last three Thanksgiving through Christmas shopping seasons, they have sold more Chromebooks than "every brand and model of (Intel) Windows based laptops combined"
While I sincerely hope that AMD can be financially successful in competing with Intel, these always late and always choosing the wrong OS “exclusively” practices has severely hurt the company, thus pushing them into a perpetual catch-up mode.
7nm mobile APUs might be great, especially if they were to have LPDDR4X or 5 support. That would allow AMD to enter high-end ultrathin market, and offer decent graphics performance. But these 12nm chips are also very good, and only need good laptop designs. And this Asus TUF actually looks well built (XFastest review).
" Starting in Q1, any laptop with a Ryzen Mobile CPU that uses integrated graphics will be supported direct from AMD with a single driver package to cover all devices. "
This is really good news. Laptops--I'm looking at you specifically, Dell--have traditionally been hamstrung by vendor insistence on not letting stock AMD drivers work with their products, leading to massively-outdated drivers being the only ones available.
Chase the RICH AMD, not the POOR. Ask NV/Intel for advice on your next product announcements & perhaps more importantly, HOW TO PRICE THEM to make NET INCOME. IE, don't start a price war from day one launches. Nobody else wants one, so not sure why you keep starting them with EVERY launch. If your next gpu performs like a GTX 1080 (rumor 7nm small aiming here), then for god's sake, CHARGE 1080 PRICING (not $249 rumor? LOL, no net income), especially if you are LOWER watts while doing it. YOU WIN. YOU CHARGE! You lose, you are bargain bin, and so is your NET INCOME. It is that simple. Aim HIGH, or DIE.
Your cpu's seem to be heading in the right direction, I only hope you don't charge like your in 2nd place for the next 12-18 months pissing away your 7nm vs. 14nm Intel competition lead. Intel's tech is not screwed because of a failed launch or failed 10nm. IF they are being truthful (surely as they have the money to work on multiple things at once), 7nm will make 10nm dead ASAP and is ON schedule unlike 10nm 2yrs-3yrs late. If you don't make hay while the sun shines for the next ~18 months, how will you combat Intel R&D for the next round? Again, see Intel/NV launches. You aren't in gpu (until 7nm maybe) above $350 really now, so NV stacks pricing ON TOP of old gen (perf of 2060 is worth it, just saying, price going up with perf). Intel loses ability to keep up with 14nm and thus screws poor people and moves production to RICH SERVER people. Both WISE decisions, unlike you chasing console (poor PC people?...LOL), apu, and er, chromebooks with 28nm chips. 720 low quality? ROFLMAO. 720p should not be uttered in 2019, and at 7nm you'd be talking 1080p gaming (with your gpu) and ALL DAY battery life like Intel's coming stuff. Also note those chips are in $1600 laptops, while you're aiming at what, $250-300 chromebooks? This HP starts at $269...LOL. Who's margins and income do you think are BETTER here people? Hint, it isn't AMD.
Quit trying to make small cheap chips too. Small=2nd rate. Larger (NV reticle limit, Apple large socs etc)=NET INCOME because you tend to WIN at least for a while until the other guy comes. You go small you get 7nm gpu@GTX 1080 (maybe...LOL) performance. Large=RTX with new features to brag about, DLSS+RT and 60-92% better than 1060 for 2060 depending on tech you turn on per game. https://www.youtube.com/watch?v=9mxjV3cuB-c The 2060 die looks like it's about 2x AMD 7nm gpus coming (low 200's). Thus, POWER gamers buy in droves, vs. a YAWN over your poor people ~230mm^2 that can't keep up with old 1080ti. That is NOT a good launch vehicle for new tech, NV way is FAR better for the bottom line (ie, NET INCOME). Do you want to be BILL GATES, or the HOMELESS guy? MSFT gets most of it's money from ENTERPRISE, meaning RICH folks. You get your money from POOR people mostly, and your NET INCOME shows it Q after Q, year after year.
Also, market share means NOTHING if you don't have decent margins on them. I'd rather own 10-15% of the market and 50% of the NET INCOME (see apple iphones), than 85-90% of the market and the other 50% NET being split over the REST of the market, meaning nobody can afford to keep up with Apple R&D at these rates. The same story can be said about Intel/NV vs. AMD. At least you're chasing server/HEDT with cpu, but why not GPU KINGS too? You will continue to fail if you aim low. Poor people complain about $10-50 on a gpu, while RICH people just keep buying titans yearly without blinking. If you're making $200k+ a year a titan card isn't even a Christmas gift, it's F5 Olympics at launch no matter what day of the year that comes. See the difference? Don't bother with poor people until the rich can't afford to buy more ;) When a rich product is stuck on the shelf, THEN and ONLY then, do you make poor gpu/cpu stuff :) PERIOD. You are not in business to be our friend, you are in business to make NET INCOME. PERIOD.
Your going small on gpu, so NV just can keep 12nm going and drop the price at most, then hit reticle limit again Q1/20 with 7nm and raise prices back up. Your strategy is helping the competition, rather than KILLING their HIGH margin cash cows (IE, gpus above $300, consumer/pro/server here). You won't even nick NV profits with cards under $250 and their R&D train will just keep going with 4B+ a year NET.
Get smarter AMD, chase the RICH first and poor LAST. They are laughing at your NET INCOME and so is Intel. R&D at Nvidia is now 1.8B, while AMD 1.1B and they make a lot more stuff. Why is this so hard for AMD to figure out. Chasing poor, just keeps you POOR.
BTW that AMD number for R&D has finally eclipsed 2014 data and BARELY (1.074 back then, 1.1 now). Meanwhile NV on the up train for years chasing RICH. OH, and you can help the poor if you're RICH (donate all you want Intel/NV with 60% margins), but you can't do that if you're POOR living paycheck to paycheck (AMD with 30-35% margins)...LOL.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
70 Comments
Back to Article
Chaitanya - Sunday, January 6, 2019 - link
Are these new CPUs built on zen+ or Zen2 architecture?rascalion - Sunday, January 6, 2019 - link
2nd paragraph in. Zen+brakdoo - Sunday, January 6, 2019 - link
Basically everything like expected from previous leaks but I'm surprised about the driver updates coming directly from AMD. That is a big issue for current gen Ryzen-Laptop users...Qasar - Sunday, January 6, 2019 - link
i wonder if this will include driver updates for pre Zen amd apus as well.................brunis.dk - Monday, January 7, 2019 - link
I wonder if your next post will have more dots at the end........................................Dragonstongue - Tuesday, January 8, 2019 - link
Likely not pre Zen stuff.AMD put extra firmware into their code so that they would be able to do a more direct "to the lineup"
if required for certain update paths, this is why Zen 1 was able to "drop in" to a 3xx or 4xx board and also got a bit of an enhancement when the 2xxx chips launched some of this helped (sometimes a fair amount) the 1xxx chips to go higher speed or higher dram speeds/stability.
StevoLincolnite - Sunday, January 6, 2019 - link
This. I have a Lenovo Ryzen 2700U notebook... And the driver updates have been bullshit thus far... So this is very much a welcomed change for the better.Actually interested to see how the 3750H stacks up against my 2700u, especially if it supports memory faster than 2400mhz in dual channel... I might. Might... Be tempted to upgrade if it gives me that extra "kick" in my games. - AMD might also be able to fix the high idle power consumption too?
Had a 2in1 with the 9220 last year for a brief period... Was capable of playing skyrim at 720P@30fps, so it's fine for older/simpler games, but the display was almost unusable... And the EMMC drive made it feel like a slug so returned it to Dell for a full refund.
Opencg - Sunday, January 6, 2019 - link
I agree laptop manufactuers suck at posting driver updates. At least most of them dovelanapontinha - Monday, January 7, 2019 - link
"2700u, especially if it supports memory faster than 2400mhz in dual channel" - it already does. I have 3 2700U laptops running 16GB of 2666mhz ddr4. These are Elitebooks 755G5 which came with 1x8GB 2400mhz, take that out and replace it with 2x8GB 2666mhz and voilá, it works.StevoLincolnite - Monday, January 7, 2019 - link
Officially AMD only supports 2400Mhz on Ryzen mobile. - So anyone who runs out and grabs 2666mhz memory after reading your post is likely going to be gambling as there is zero guarantee the BIOS will upclock to 2666mhz.I have the Lenovo E585 myself... Doubt it will support anything more than 2400mhz... And I am not about to sink a few hundred bucks to find out.
velanapontinha - Tuesday, January 15, 2019 - link
True. It is not a safe bet. But there are loads of users finding out that it does work.My post is merely an indication for those who are willing to bet.
And, if you stick those dimms in a laptop limited to 2400, it will work at 2400 speed most of the times. The risk is small. Those 2666 dimms I have running in the EB755G5 will run at 2400 speed in the EB650G4. I tried it, of course.
IGTrading - Monday, January 7, 2019 - link
Why would you be surprised about AMD providing an official Ryzen Mobile driver ? it was clearly necessary, after OEMs proved incapable and unwilling to respect and support their clients ....1_rick - Monday, January 7, 2019 - link
Well, Dell in particular has been blocking video drivers on their laptops as far back as 2005.Dragonstongue - Monday, January 7, 2019 - link
I kind of wish desktops could be "from the source" type driver updates that do not relyon MSFT way of forcing certain drivers that "break things"
They should work with MSFT, Intel, Nvidia etc to make sure that the makers of the hardware
that require specific BIOS or Driver sets to function "best" for whatever reason are left the
#$%#$% alone.
Am I right ^.^
TheJian - Tuesday, January 8, 2019 - link
Actually, I thought all the leaks said 7nm EVERYTHING here. We get crap 28nm (in 2019? LOL, 7nm with far more perf would have gotten higher ASP's), and 12nm Ryzen chips. ER, UH, not what I expected. I expected GPU info at 7nm with roadmaps/launch dates, and 7nm CPU launches TODAY. What leaks were you reading?Sindalis - Sunday, January 13, 2019 - link
Why? Simple. the WSA.AMD is bound by contract to buy X wafers from Global Foundries.
If you dont know what that is in short in 2009 AMD signed an agreement when it spun off its fabs to form a new fabrication company called Global Foundries that sets fixed pricing as well as volumes of silicon wafers AMD has to buy from Global Foundries.
It in part nearly bankruped AMD beause they could not source cheaper wafers OR reduce wafer purchases when Bulldozer failed as hard as it did. They had to purchase wafters, used or not! This was a huge problem when Global Foundries COULD NOT scale below 28 NM. Eventually cancelling their own in house project and instead purchase/licence the tech for 14nm FF from Samsung, which was a low power process and truly not suitable for what their primary customer of AMD wanted, a High Power process for their new Microprocessors.
This is why Ryzen first gen hit a wall at 4ghz. It's not heat, it's not voltage, it isnt the archetecture. It's fundenmentally the node its fabbed on. the 14NM (or the improved 12Nm) process that GF is using is fundementally a Low Power node from Samsung, it isnt designed for efficiency or to be able to hit high clock speeds. What it IS designed to do is be VERY VERY efficient at lower power constraints. which makes it ideal for mobile parts that dont need as high of a clock speed where power is the big limiting factor.
While the Wafer Supply Agreement has been ammended multiple times in part to keep AMD from going under, the 2016 Ammendment also allowed AMD to source some wafers from other sources. The 6th ammendment done in 2016 it is currently set to expire in 2020,
What that means for AMD is while the ammendments have allowed them to change their cost and source wafers outside of Global Foundries (It's the entire reason they have 7nm chips sourced from TSMC. Global Foundries cancelled their 7nm node last year) Even though they are having to pay Global Foundries for using TSMC wafers!
It doesn't remove the requirement to buy wafers from Global Foundries completely though, they still MUST purchase or recieve a certain inventory of wafers. What do you do if you MUST buy a raw material by contract? You turn it into something you can sell. Ideally for a profit, but even for a loss would be better than just absorbing the cost and getting nothing out of it. Assuming the final sale reduces the amount lost.
Ergo, this is another reason these APU's are on Zen+ 12nm process rather than 7. They probably could put them on 7, but they NEED to use wafers from Global Foundries to reduce their losses. As far as why they are making 28 NM chips. Well they are better than the competitions (intel's) at this power envelope, and they probably still have more wafers they are contractually obligated to buy. Even if they still lose some money on the end product. It's still something better than just eating the full cost of the wafer with nothing to show for it. Some revenue to reduce the loss is better than a full and total loss.
Teckk - Sunday, January 6, 2019 - link
Interesting. So, we're not going to see any 7nm processors for laptops in 2019 from AMD?brakdoo - Sunday, January 6, 2019 - link
At least not until Navi is ready. Might see something at the end of the year.Cellar Door - Sunday, January 6, 2019 - link
It is an odd move, since it is the laptop which would benefit the most from a lower power envelope. Seems AMD feels confident about these parts being competitive enough to win some designs.Exodite - Sunday, January 6, 2019 - link
Zen+ is already very efficient at lower clocks though, the real struggle was scaling it up to 95W using the 12nm process.With that in mind I imagine the performance delta of the 7nm desktop chips will be much larger than it would be for mobile parts.
In that light sticking with 12nm for mobile makes sense (plus whatever fab deal are already there I suppose) but obviously I'm only speculating. :)
zodiacfml - Sunday, January 6, 2019 - link
Because they want to bring back their server CPU market. This is the reason their desktop parts are also on their latest node which will happen again this year at 7nmMonkeyPaw - Sunday, January 6, 2019 - link
I don’t know that I’d read much into the mobile announcement. Raven Ridge came 6 months after desktop Zen, and these are the 12nm versions. With the extra lead time designing an APU, they likely wanted to stick to the older architecture and just get the die shrink done. The big problem I see is that the mobile series will be a generation behind despite being in the 3xxx lineup.Teckk - Sunday, January 6, 2019 - link
Yes, that much extra work for OEMs too - designing for a new platform, power and cooling changes, connectivity. Well, looking forward to the presentation by AMD!Teckk - Sunday, January 6, 2019 - link
Oh and since AMD definitely lags behind Intel in laptops, a 7nm surprise would have been awesome. They do know what they're doing I guess,so yeah, let there be competition.Santoval - Sunday, January 6, 2019 - link
Most likely they want to fulfill wafer orders from GloFo so that they are not penalized further due to their existing agreement. They could have skipped 12nm and Zen+ and go directly to 7nm and Zen 2 in the mobile space, but that would leave GloFo holding their... proverbial instruments. With the GPUs also soon moving to 7nm they could only order I/O dies from them, and I don't think the revenue will be huge from them.On the other hand it is GloFo that does not fulfill their end of the agreement : they decided to delay 7nm indefinitely, but they surely cannot ask AMD to do the same. Does their wafer supply agreement with AMD have any clauses about newer nodes and failure to deliver them? I have no idea, but I believe there should be.
tekniknord - Sunday, January 6, 2019 - link
No one that has knowledge of AMD mobile (laptops) expected 7nm mobile APU this year.AMD is always one year behind on mobile chips.
In 2020 it's likely we get 7nm AMD mobile.
I only wonder when i can find an AMD 3000 mobile in a local store, often it takes months after AMD releases a mobile chip until you can buy them in local stores.
There are the few presented models that you can order online and it then takes a long time before someone can find them in local stores, maybe i can find them locally at the second half of the summer.
I really hope that AMD mobile 3000 hits stores faster this time around.
IGTrading - Monday, January 7, 2019 - link
@tekniknord ... AMD is "one year behind" who exactly ?! :) Itself ? Because when compared to Intel mobile CPUs, AMD is one or two years ahead in every metric except maybe idle power consumption ....tekniknord - Monday, January 7, 2019 - link
AMD mobile is lagging one year after their desktop product releases.I didn't mention Intel at all because my answer was related to AMD desktop releases vs mobile.
I guess i wasn't clear enough.
nandnandnand - Monday, January 7, 2019 - link
AMD's 7nm mobile chips in 2020 could be a massive improvement over today's announcement. It could double core counts to up to 8. Power efficiency increase could result in lower TDPs, needed for stuff like fanless Chromebooks. Along with Navi graphics, maybe it will have hardware decoding for the AV1 codec.Skip these 12nm mobile chips.
jasaero - Sunday, January 6, 2019 - link
How do you say something is underserved and then offer excavator 28nm crap to help?wolfesteinabhi - Sunday, January 6, 2019 - link
its basically almost no contest market segment there right now... AMD cant just poof a part for it...so they repurpose an existing part to fit that segment .... remember for chromebooks .... the part should be low TDP as well as "cheap" enough for it... or else it defeats the purpose. ... current crop of products are low on power ..but being on cutting edge tech ..they are quite costly to offer for such a segment.notb - Sunday, January 6, 2019 - link
Actually Zen is NOT good for low power systems. The cores are efficient, but Infinity Matrix eats so much energy that it's almost impossible to design something usable below 15W. That's why AMD is reviving older tech.Zen was a server product from the start - perfect for 150W+ systems. Consumer desktops are already pushing this architecture out of comfort zone. To be competitive in laptops, AMD would have to design purpose-built CPUs and the simple fact is: they can't afford that right now.
jasaero - Sunday, January 6, 2019 - link
Yeah.... But they could do a more simplified Zen apu that would work fine at this power level. Single CCX that has some lower powered GPU and simple lower power connection to it.Generally though they do not seem interested in mobile...could be a lot of interest in well though out lower power Zen derivatives. I think AMD and Intel are slowly allowing too much to get lost to ARM as far as the broader computing market. Sub 10w stuff should be on newer even if not latest process at least and use best core IP levels they can. Stuff like excavator against more upto date ARM chromebooks at similar price just makes AMD look that much worse.
hukax - Sunday, January 6, 2019 - link
> Athlon 300W for entry-level devicesThat is some powerful entry-leve device.
tipoo - Sunday, January 6, 2019 - link
" and features a 14-inch display with a 1366x768 resolution."Almost could have been interesting, but died to me right there.
Hopefully AMD parts will be in more interesting Chromebooks from here.
zepi - Sunday, January 6, 2019 - link
AMD should try to offer some kind of a solution to the memory BW issue of integrated graphics.quadibloc - Sunday, January 6, 2019 - link
My guess would be that the difference between a 35 watt part, and a 15 watt part, with identical specs, is that the 15 watt part is set for more aggressive power management. So it will still yield less sustained performance, although peak performance will be identical, because it won't be able to keep everything active all the time.RBD117 - Sunday, January 6, 2019 - link
Precisely this. Hopefully the 35W sustained performance gain will be good enough on mobile this time around, maybe like the 35W desktop AM4 parts. That would make for some really great gaming laptops.BigMamaInHouse - Sunday, January 6, 2019 - link
Cinebench R15 nT ("CPU"):Core i5-8250U vs. Ryzen™ 5 3500U: 524 vs. 651 (24%/1.24X faster for AMD)
Core i7-8565U vs. Ryzen™ 7 3700U: 619 vs. 688 (11%/1.11X faster for AMD)
3DMark® Time Spy ("GPU"):
Core i5-8250U vs. Ryzen™ 5 3500U: 399 vs. 907 (127%/2.27X faster for AMD)
Core i7-8565U vs. Ryzen™ 7 3700U: 444 vs. 967 (118%/2.18X faster for AMD)
notb - Sunday, January 6, 2019 - link
8565U is a 2018 model and 8250U was launched in 2017.Intel will soon release 9th gen models and that's what these Ryzens will have to compete against.
Also, why would anyone care whether AMD managed to squeeze more power from an ULV chip (even less so about GPU performance)? Previous gen already had plenty.
The big question here is: did they improve efficiency and power management. Laptops built around 2500U and 2700U have atrocious battery life.
BigMamaInHouse - Sunday, January 6, 2019 - link
esting by AMD performance labs as of 12/4/2018. “Battery life” defined as hours of continuous usage before the system automatically shuts down due to depleted battery. Video playback tested according to Microsoft WER methodology, while “general usage” is tested via MobileMark 14. Results presented in hours, in order of: 1st Gen AMD Ryzen™ 7 2700U Mobile Processor (100%) vs. 2nd Gen AMD Ryzen™ 7 3700U Mobile Processor. General Usage: Ryzen™ 7 2700U: 8.1 hours vs. Ryzen™ 7 3700U: 12.3 hours (51% longer) Video Playback: Ryzen™ 7 2700U: 6.9 hours vs. Ryzen™ 7 3700U: 10 hours (40% longer) Ryzen™ 7 2700U Test System: Lenovo IdeaPad 530s, Ryzen™ 7 2700U, 2x4GB DDR4-2400, Radeon™ Vega 10 Graphics (driver 23.20.768.0), 1920x1080 AUO 403D 13.9” panel, 512GB Toshiba KBG30ZMT512G SSD, 45Whr battery, 150 nits brightness, Windows® 10 x64 RS4. Ryzen™ 7 3700U Test System: AMD Reference Motherboard, AMD Ryzen™ 7 3700U, 2x4GB DDR4-2400, Radeon™ Vega 10 Graphics (driver 23.20.768.0), AUO B140HAN05.4 14” panel, 256GB WD Black WD256G1XOC SSD, 50Whr battery, 150 nits brightness, Windows® 10 x64 RS5. Results may vary with drivers and configuration. RVM-164velanapontinha - Monday, January 7, 2019 - link
I hope so, I really do. Battery life needs to be fixed asap, it is just unacceptable right now.HStewart - Sunday, January 6, 2019 - link
I think this test is a little unfair - since GPU on AMD is more of discrete GPU until like iGPUs - maybe the test should be against say 8705G in Dell XPS 15 2in1ajp_anton - Monday, January 7, 2019 - link
What? The 8705G is 65W while all of the CPUs in the comparison are 15W. So why would it be unfair?ET - Sunday, January 6, 2019 - link
I'm surprised that AMD hasn't disclosed any architectural changes. They usually mention some. Hopefully that'd come later.And it's nice to see Stoney Ridge still alive and kicking.
neblogai - Sunday, January 6, 2019 - link
There are probably none- only better efficiency/clocks/latency from 12nm with new microcode.TheinsanegamerN - Sunday, January 6, 2019 - link
Paring these APUs with a discrete GPU like the 560x is completely pointless. When will OEMs learn to simply let the APU shine on its own instead of crippling it with worthless extra silicon?Hopefully that AMD driver fix works, forcing OEMs to do driver updates was an incredibly dumb thing for AMD to do in the first place that killed a LOT of enthusiasm for the platform once the poor performance was shown off, and there was no reason for it other then AMD being lazy. Hopefully their updated drivers will fix a lot of the longstanding mobile issues.
RBD117 - Sunday, January 6, 2019 - link
Not sure if it was forced, but typically I was under the impression that as a silicon vendor, you may prefer OEMs to handle releases since every mobile platform is technically different and OEMs would "be the best equipped" to do their own custom internal testing to catch bugs that a general release could cause on their particular systems. Obviously we understand now that in practice, OEMs dont do anything at all here, even up to a year later (I'm looking at you HP), so thank God AMD is going to do a general drive release. I gather tech-savvy users will take the risk for better perf and power management and fix their own issues, and the layman won't even notice. Win-win.mikato - Tuesday, January 8, 2019 - link
Will AMD have to be careful with all the different laptop chassis cooling abilities out there when they are sending drivers directly to laptops and skipping the OEM?ST33LDI9ITAL - Sunday, January 6, 2019 - link
Why won't AMD get it... and just make what most mainstream users want and need. A 4/4 or 6/6 4ghz cpu with a vega 10 or beefier gpu, with gddr or hbm. Sigh. One day... one day.ST33LDI9ITAL - Sunday, January 6, 2019 - link
FFS, just give us a console cpu/apu package with beefy GPU and hbm2 and watch sales soar!hanselltc - Sunday, January 6, 2019 - link
I wonder why does AMD not use their latest arch on the latest node for mobile. Wouldn't you want more power efficiency on mobile?neblogai - Monday, January 7, 2019 - link
For AMD, APUs are low margin products. Zen CPU dies of the same size sell for ~€200 on average, and go for both desktop and servers. APUs, on the other hand, sell for ~€100 on desktop and mobile- need to be pushed into Intel dominated mobile market, with volume being not high. And mobile market has a lot of inertia, so it is not easy to enter. So, it makes sense for AMD to do it slower. They still have a lot to fix- like drivers, switchable graphics support, adding LPDDR support needed for small/high end devices, making video playback more efficient, etc. And, like I said, bigger/faster money is in Zen dies- which already had a 12nm refresh and now get Zen2 + 7nm. Lower margin product- APUs- will follow the path of 12nm refresh of same silicon, that offers a better product at minimal investment, and only then get whole new CPU and iGPU design at 7nm. Probably again for a new product (7nm) and then a refresh cycle (7nm+ or 5nm), which is a tick-tock strategy of all AMD products.wanderson - Sunday, January 6, 2019 - link
AMD have unfortunately followed Microsoft Operating System (OS) support blindly for at least two generations, even while Linux was captiring the Server, Data cwnter and Cloud computing Market left to Intel for plundering. Even ARM has started serious entry into the Server market.Now the company is looking at Chromebooks, at a time when seveal Chromebook manufacturers are eyeing ARM based systems.
By the way, more than 52% of all school systems in USA are deploying Chromebooks, and Amazon has announced that for the last three Thanksgiving through Christmas shopping seasons, they have sold more Chromebooks than "every brand and model of (Intel) Windows based laptops combined"
While I sincerely hope that AMD can be financially successful in competing with Intel, these always late and always choosing the wrong OS “exclusively” practices has severely hurt the company, thus pushing them into a perpetual catch-up mode.
Eris_Floralia - Sunday, January 6, 2019 - link
Ian, according to AMD the 3200U/300U are the same arch as PCO but on 14LPP.These should be the new Raven2(RV2) die.
zodiacfml - Sunday, January 6, 2019 - link
I'd rather sit this one out and wait for next year 7nm mobile PC chips.neblogai - Monday, January 7, 2019 - link
7nm mobile APUs might be great, especially if they were to have LPDDR4X or 5 support. That would allow AMD to enter high-end ultrathin market, and offer decent graphics performance. But these 12nm chips are also very good, and only need good laptop designs. And this Asus TUF actually looks well built (XFastest review).1_rick - Monday, January 7, 2019 - link
Typo in paragraph 3: "15W Athlon 300W", should be "Athlon 300U", according to the table.1_rick - Monday, January 7, 2019 - link
" Starting in Q1, any laptop with a Ryzen Mobile CPU that uses integrated graphics will be supported direct from AMD with a single driver package to cover all devices. "This is really good news. Laptops--I'm looking at you specifically, Dell--have traditionally been hamstrung by vendor insistence on not letting stock AMD drivers work with their products, leading to massively-outdated drivers being the only ones available.
TheJian - Tuesday, January 8, 2019 - link
Chase the RICH AMD, not the POOR. Ask NV/Intel for advice on your next product announcements & perhaps more importantly, HOW TO PRICE THEM to make NET INCOME. IE, don't start a price war from day one launches. Nobody else wants one, so not sure why you keep starting them with EVERY launch. If your next gpu performs like a GTX 1080 (rumor 7nm small aiming here), then for god's sake, CHARGE 1080 PRICING (not $249 rumor? LOL, no net income), especially if you are LOWER watts while doing it. YOU WIN. YOU CHARGE! You lose, you are bargain bin, and so is your NET INCOME. It is that simple. Aim HIGH, or DIE.Your cpu's seem to be heading in the right direction, I only hope you don't charge like your in 2nd place for the next 12-18 months pissing away your 7nm vs. 14nm Intel competition lead. Intel's tech is not screwed because of a failed launch or failed 10nm. IF they are being truthful (surely as they have the money to work on multiple things at once), 7nm will make 10nm dead ASAP and is ON schedule unlike 10nm 2yrs-3yrs late. If you don't make hay while the sun shines for the next ~18 months, how will you combat Intel R&D for the next round? Again, see Intel/NV launches. You aren't in gpu (until 7nm maybe) above $350 really now, so NV stacks pricing ON TOP of old gen (perf of 2060 is worth it, just saying, price going up with perf). Intel loses ability to keep up with 14nm and thus screws poor people and moves production to RICH SERVER people. Both WISE decisions, unlike you chasing console (poor PC people?...LOL), apu, and er, chromebooks with 28nm chips. 720 low quality? ROFLMAO. 720p should not be uttered in 2019, and at 7nm you'd be talking 1080p gaming (with your gpu) and ALL DAY battery life like Intel's coming stuff. Also note those chips are in $1600 laptops, while you're aiming at what, $250-300 chromebooks? This HP starts at $269...LOL. Who's margins and income do you think are BETTER here people? Hint, it isn't AMD.
Quit trying to make small cheap chips too. Small=2nd rate. Larger (NV reticle limit, Apple large socs etc)=NET INCOME because you tend to WIN at least for a while until the other guy comes. You go small you get 7nm gpu@GTX 1080 (maybe...LOL) performance. Large=RTX with new features to brag about, DLSS+RT and 60-92% better than 1060 for 2060 depending on tech you turn on per game.
https://www.youtube.com/watch?v=9mxjV3cuB-c
The 2060 die looks like it's about 2x AMD 7nm gpus coming (low 200's). Thus, POWER gamers buy in droves, vs. a YAWN over your poor people ~230mm^2 that can't keep up with old 1080ti. That is NOT a good launch vehicle for new tech, NV way is FAR better for the bottom line (ie, NET INCOME). Do you want to be BILL GATES, or the HOMELESS guy? MSFT gets most of it's money from ENTERPRISE, meaning RICH folks. You get your money from POOR people mostly, and your NET INCOME shows it Q after Q, year after year.
Also, market share means NOTHING if you don't have decent margins on them. I'd rather own 10-15% of the market and 50% of the NET INCOME (see apple iphones), than 85-90% of the market and the other 50% NET being split over the REST of the market, meaning nobody can afford to keep up with Apple R&D at these rates. The same story can be said about Intel/NV vs. AMD. At least you're chasing server/HEDT with cpu, but why not GPU KINGS too? You will continue to fail if you aim low. Poor people complain about $10-50 on a gpu, while RICH people just keep buying titans yearly without blinking. If you're making $200k+ a year a titan card isn't even a Christmas gift, it's F5 Olympics at launch no matter what day of the year that comes. See the difference? Don't bother with poor people until the rich can't afford to buy more ;) When a rich product is stuck on the shelf, THEN and ONLY then, do you make poor gpu/cpu stuff :) PERIOD. You are not in business to be our friend, you are in business to make NET INCOME. PERIOD.
Your going small on gpu, so NV just can keep 12nm going and drop the price at most, then hit reticle limit again Q1/20 with 7nm and raise prices back up. Your strategy is helping the competition, rather than KILLING their HIGH margin cash cows (IE, gpus above $300, consumer/pro/server here). You won't even nick NV profits with cards under $250 and their R&D train will just keep going with 4B+ a year NET.
Get smarter AMD, chase the RICH first and poor LAST. They are laughing at your NET INCOME and so is Intel. R&D at Nvidia is now 1.8B, while AMD 1.1B and they make a lot more stuff. Why is this so hard for AMD to figure out. Chasing poor, just keeps you POOR.
TheJian - Tuesday, January 8, 2019 - link
BTW that AMD number for R&D has finally eclipsed 2014 data and BARELY (1.074 back then, 1.1 now). Meanwhile NV on the up train for years chasing RICH. OH, and you can help the poor if you're RICH (donate all you want Intel/NV with 60% margins), but you can't do that if you're POOR living paycheck to paycheck (AMD with 30-35% margins)...LOL.Lina Smooth - Thursday, January 10, 2019 - link
I going to buy HP Chromebook 14Need it for my work (I do SEO, have own blog you can look here <a href="https://www.linksmanagement.com/a-link-juice/"... I hope it is really powerful.
Lina Smooth - Thursday, January 10, 2019 - link
I going to buy HP Chromebook 14Need it for my work (I do SEO, have own blog you can look here https://www.linksmanagement.com/a-link-juice). I hope it is really powerful.
zamroni - Tuesday, October 1, 2019 - link
sorry x86, I will wait for thinkpad x series with snapdragon 8cx.