Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
AMD announces many products at CES
There are three major classes of products that AMD announced. None are available yet, but they're all coming soon.
The most important is the Ryzen Mobile 4000 series. AMD has already launched Zen 2 based CPU parts for mainstream desktops, servers, and HEDT, and soundly trounced Intel in most of the ways that matter in all of those markets. Conspicuously missing was laptops, where AMD should be able to offer a terrific product because their CPU cores are so much more efficient than Intel's. So we knew that the Ryzen Mobile 4000 series was coming, and even had a pretty good guess of what it would be called.
Today, AMD announced that it will offer 8 CPU cores (16 threads), a GCN-based GPU with 8 compute units, and a CPU max turbo of 4.2 GHz. Also, that's a 15 W part. AMD is also offering a 45 W version of the same thing, which laptop vendors are apparently only interested in using for gaming laptops. Naturally, there will be cut-down 6-core and 4-core versions of the same thing, and with analogously fewer GPU compute units.
It's interesting that AMD is finally offering an 8-core APU. They had topped out at 4 cores ever since Llano way back in 2011. AMD has long been known for offering more cores than Intel, even in the days when AMD's CPU cores weren't very good. So it's interesting that their APUs stayed at four cores for so long even as Intel moved to six and then eight. Well, AMD has sold plenty of 8-core APUs for consoles, but low power Jaguar cores are a different beast entirely.
At multi-threaded CPU or integrated GPU performance, there's little doubt that AMD will handily crush anything Intel has to offer in the same power envelope. We've already seen the same CPUs compared in desktops, and Intel's integrated GPUs aren't very good.
But there are two areas where Intel may be able to hold onto a lead. First is single-threaded CPU performance. There's quite a gap between AMD's max turbo of 4.2 GHz in a Ryzen 7 4800H and Intel's 5.0 GHz in a Core i9-9980HK--and that's much larger than the desktop part gap between AMD's 4.7 GHz and Intel's 5.0 GHz. A 15 W power envelope isn't going to clock terribly close to 5 GHz for very long, so AMD might well commonly catch Intel in single-threaded CPU performance there. But when a single CPU core has most of that 45 W all to itself, that allows a pretty high clock speed. You can't push several cores close to 5 GHz and stay inside of 45 W, so the advantage will eventually shift back to AMD as you start using more CPU cores.
The second is low idle power consumption, and hence long battery life. AMD has never been competitive with Intel there since Conroe dropped way back in 2006. There have been a variety of reasons for this. One is the lack of LPDDR* support, which the Ryzen Mobile 4000 series finally adds (LPDDR4X). But that hasn't been the only reason, and we'll see if AMD managed to catch or pass Intel there. Of course, if an Intel-based laptop can get you 10 hours of battery life when idle at desktop, and an analogous AMD-based part can "only" get you 9 hours, do you really care about that difference? If the battery will last longer than you need it, that's good enough.
AMD customarily doesn't publish prices on laptop parts. The laptops themselves will launch when they're ready. Laptops have to do enough custom stuff that to the extent that laptop CPUs even have a launch date, it isn't accompanied by a whole lot of laptop launches. But AMD does expect the first laptops to arrive in the first quarter of this year.
The second major announcement was the Radeon RX 5600 XT. This takes the same Vega 10 GPU chip as in the Radeon RX 5700 XT and cuts it down further. It doesn't just disable compute units and clock lower, but also disables a memory controller. This will be AMD's first memory size that isn't a power of 2 since the Radeon HD 7970 had 3 GB way back in 2012. For that matter, it will only be the second time that AMD has gone with a non-power of 2 memory capacity since buying ATI way back in 2006.
As the Radeon RX 5600 XT is just a new bin of an old part, there isn't really that much to know about it. It's going to be faster than an RX 5500 XT and slower than an RX 5700, and with a price tag of $280. Exactly how interesting of a part that is depends on exactly how fast it is. There won't be reference boards, but only custom boards from AMD's usual board partners. The new cards go on sale on January 21.
The third major announcement is the final third generation Threadripper part, the Threadripper 3990X. AMD had previously said that a 64-core part was coming this year. Now we have a date: February 7. We also have a price, where it will be one of the rare parts whose price tag is actually in the name: $3990. Normally, I'd call that $4000, but since it's in the part name, I'm willing to call it $3990.
While the Threadripper 3960X and 3970X could justifiably be called legitimate gaming CPUs that have a bunch of extra stuff added (more cores, more memory, more PCI Express connectivity, etc.), that's not the case for the 3990X. It will have the same 280 W TDP as the parts with fewer cores, but necessarily have to eat up more of that TDP on powering extra cores and bandwidth to connect them all. That means it will clock lower, with a max turbo of only 4.3 GHz.
The Threadripper 3990X will be a rather dumb part for anyone who can't actually push more than 32 cores. But for those workloads that will scale well to 64 cores and 128 threads, it's going to be an awesome part.
Having competitive CPU cores while being a process node ahead gives you a lot of advantages. AMD has been able to exploit that with their third generation Ryzen and Threadripper parts, as well as their Rome EPYC server parts. Now they're going to be able to do the same with laptops.
For now, AMD is offering superior products at competitive prices, as they're trying to gain market share. They pretty much own the HEDT market, and have dominated the enthusiast portion of the desktop market for the last several months, though it takes much longer for server market share to adjust. But if they have years of high market share while Intel doesn't offer competitive products, AMD won't feel the need to offer such competitive prices forever. I'd like to see AMD make a lot of money this year, as I don't want them to have another near-death experience that risks leaving Intel and Nvidia as near-monopolies in their respective markets. But we really need Intel to get their act together and offer some better products within the next few years or else AMD is likely to become the dominant player in x86 and exploit that in many of the ways that Intel has. When multiple vendors offer competitive products, their customers are the real winners.
That analysis notably doesn't apply to Nvidia. While AMD is ahead of Nvidia on the process node of launched products, the only efficiency metric in which Nvidia is notably behind is die size--which not coincidentally, is the easiest thing to fix with a die shrink. Furthermore, Nvidia has access to (and commonly uses) the same process nodes as AMD, so there's no risk of Nvidia being stuck with a process node disadvantage over the long term. Intel does run that risk unless its foundry can get its act together soon.
The most important is the Ryzen Mobile 4000 series. AMD has already launched Zen 2 based CPU parts for mainstream desktops, servers, and HEDT, and soundly trounced Intel in most of the ways that matter in all of those markets. Conspicuously missing was laptops, where AMD should be able to offer a terrific product because their CPU cores are so much more efficient than Intel's. So we knew that the Ryzen Mobile 4000 series was coming, and even had a pretty good guess of what it would be called.
Today, AMD announced that it will offer 8 CPU cores (16 threads), a GCN-based GPU with 8 compute units, and a CPU max turbo of 4.2 GHz. Also, that's a 15 W part. AMD is also offering a 45 W version of the same thing, which laptop vendors are apparently only interested in using for gaming laptops. Naturally, there will be cut-down 6-core and 4-core versions of the same thing, and with analogously fewer GPU compute units.
It's interesting that AMD is finally offering an 8-core APU. They had topped out at 4 cores ever since Llano way back in 2011. AMD has long been known for offering more cores than Intel, even in the days when AMD's CPU cores weren't very good. So it's interesting that their APUs stayed at four cores for so long even as Intel moved to six and then eight. Well, AMD has sold plenty of 8-core APUs for consoles, but low power Jaguar cores are a different beast entirely.
At multi-threaded CPU or integrated GPU performance, there's little doubt that AMD will handily crush anything Intel has to offer in the same power envelope. We've already seen the same CPUs compared in desktops, and Intel's integrated GPUs aren't very good.
But there are two areas where Intel may be able to hold onto a lead. First is single-threaded CPU performance. There's quite a gap between AMD's max turbo of 4.2 GHz in a Ryzen 7 4800H and Intel's 5.0 GHz in a Core i9-9980HK--and that's much larger than the desktop part gap between AMD's 4.7 GHz and Intel's 5.0 GHz. A 15 W power envelope isn't going to clock terribly close to 5 GHz for very long, so AMD might well commonly catch Intel in single-threaded CPU performance there. But when a single CPU core has most of that 45 W all to itself, that allows a pretty high clock speed. You can't push several cores close to 5 GHz and stay inside of 45 W, so the advantage will eventually shift back to AMD as you start using more CPU cores.
The second is low idle power consumption, and hence long battery life. AMD has never been competitive with Intel there since Conroe dropped way back in 2006. There have been a variety of reasons for this. One is the lack of LPDDR* support, which the Ryzen Mobile 4000 series finally adds (LPDDR4X). But that hasn't been the only reason, and we'll see if AMD managed to catch or pass Intel there. Of course, if an Intel-based laptop can get you 10 hours of battery life when idle at desktop, and an analogous AMD-based part can "only" get you 9 hours, do you really care about that difference? If the battery will last longer than you need it, that's good enough.
AMD customarily doesn't publish prices on laptop parts. The laptops themselves will launch when they're ready. Laptops have to do enough custom stuff that to the extent that laptop CPUs even have a launch date, it isn't accompanied by a whole lot of laptop launches. But AMD does expect the first laptops to arrive in the first quarter of this year.
The second major announcement was the Radeon RX 5600 XT. This takes the same Vega 10 GPU chip as in the Radeon RX 5700 XT and cuts it down further. It doesn't just disable compute units and clock lower, but also disables a memory controller. This will be AMD's first memory size that isn't a power of 2 since the Radeon HD 7970 had 3 GB way back in 2012. For that matter, it will only be the second time that AMD has gone with a non-power of 2 memory capacity since buying ATI way back in 2006.
As the Radeon RX 5600 XT is just a new bin of an old part, there isn't really that much to know about it. It's going to be faster than an RX 5500 XT and slower than an RX 5700, and with a price tag of $280. Exactly how interesting of a part that is depends on exactly how fast it is. There won't be reference boards, but only custom boards from AMD's usual board partners. The new cards go on sale on January 21.
The third major announcement is the final third generation Threadripper part, the Threadripper 3990X. AMD had previously said that a 64-core part was coming this year. Now we have a date: February 7. We also have a price, where it will be one of the rare parts whose price tag is actually in the name: $3990. Normally, I'd call that $4000, but since it's in the part name, I'm willing to call it $3990.
While the Threadripper 3960X and 3970X could justifiably be called legitimate gaming CPUs that have a bunch of extra stuff added (more cores, more memory, more PCI Express connectivity, etc.), that's not the case for the 3990X. It will have the same 280 W TDP as the parts with fewer cores, but necessarily have to eat up more of that TDP on powering extra cores and bandwidth to connect them all. That means it will clock lower, with a max turbo of only 4.3 GHz.
The Threadripper 3990X will be a rather dumb part for anyone who can't actually push more than 32 cores. But for those workloads that will scale well to 64 cores and 128 threads, it's going to be an awesome part.
Having competitive CPU cores while being a process node ahead gives you a lot of advantages. AMD has been able to exploit that with their third generation Ryzen and Threadripper parts, as well as their Rome EPYC server parts. Now they're going to be able to do the same with laptops.
For now, AMD is offering superior products at competitive prices, as they're trying to gain market share. They pretty much own the HEDT market, and have dominated the enthusiast portion of the desktop market for the last several months, though it takes much longer for server market share to adjust. But if they have years of high market share while Intel doesn't offer competitive products, AMD won't feel the need to offer such competitive prices forever. I'd like to see AMD make a lot of money this year, as I don't want them to have another near-death experience that risks leaving Intel and Nvidia as near-monopolies in their respective markets. But we really need Intel to get their act together and offer some better products within the next few years or else AMD is likely to become the dominant player in x86 and exploit that in many of the ways that Intel has. When multiple vendors offer competitive products, their customers are the real winners.
That analysis notably doesn't apply to Nvidia. While AMD is ahead of Nvidia on the process node of launched products, the only efficiency metric in which Nvidia is notably behind is die size--which not coincidentally, is the easiest thing to fix with a die shrink. Furthermore, Nvidia has access to (and commonly uses) the same process nodes as AMD, so there's no risk of Nvidia being stuck with a process node disadvantage over the long term. Intel does run that risk unless its foundry can get its act together soon.
Comments
factor... but I guess it’s being done and hasn’t stopped people from buying them
https://www.pcgamer.com/intel-says-its-next-gen-laptop-cpus-double-the-graphics-performance-of-10th-gen-processors/
And a surprisingly well-timed rumor says NVidia's Ampere GPUs will offer 50% increased performance and do it with 50% less power
https://www.tomshardware.com/news/nvidia-ampere-purportedly-50-faster-than-turing-at-half-the-power-consumption
Intel doubling their GPU performance by some reasonable metric is much more plausible. The question isn't whether Intel can do so. It's whether they're willing to burn the die space and power to do so.
"You should expect that our discrete graphics as we go through 2020 will also have ray tracing."
"you should expect that we will have a high-end Navi"
"you will see Zen 3 in 2020!"
The whole interview is here:
https://www.anandtech.com/show/15344/amd-at-ces-2020-qa-with-dr-lisa-su
Guessing they aren't going to make the same mistake here.
According to Kan Liu, Director of Product Management for Google's Chrome OS
"more powerful Chromebooks, especially AMD Chromebooks, are coming"
This comes together with news that Google is working with Valve to get Steam to Chromebooks.
Source: https://www.androidpolice.com/2020/01/17/exclusive-google-is-working-to-bring-steam-to-chrome-os/
No timeframe was given, but this looks promising for AMD because at the moment Chromebooks with AMD processors are really rare.
There's also the issue that in a sense, AMD has a lot more lines of products relevant to gaming desktops or laptops than Nvidia or Intel. Nvidia does GPUs and that's it. Intel does CPUs and that's it, though they'll get into discrete GPUs soon. AMD does both of those, and also separates their pure CPU products from their APUs that have both a CPU and GPU. That essentially gives AMD three lines of products relevant to gaming desktops or laptops, while Intel and Nvidia each have only one.
Stock market price increases are almost entirely based on speculative performance. That's how companies like Amazon can survive, and even thrive, for more than a decade of no profit.
AMD's price is riding high not because they have developed this technology - that was just a massive cost in R&D. It doesn't mean anything to Shareholders unless there is a market for it: they anticipate that the new technology compete well against competition in the near term, and there is a marketplace for it with potential for growth.
Managing shareholder opinion is just as important as the R&D itself - without that confidence, the shareholders could just as easily see R&D costs as a waste and punish your stock price for it. JSH at nVidia is masterful at this, and most of his company announcements have more to do with managing stock price than anything (in my opinion); and Lisa is coming a long way into that role and these statements show that.
In general: stocks hold value when they meet their expectations. They tank when they miss it or the market moves in a different direction than the company. They grow when there is a potential for future growth.
The first sentence there - my post explains exactly why. If you have been paying attention to stocks, as you have, and to what companies announce and when they choose to make those particular announcements then it's not difficult at all to see why nVidia is worth $200/share. Since they don't have to react to stiff competition, almost everything nVidia does is carefully calculated with respect to their share price. If you listen to nVidia, Raytracing is the next big thing and it's only brought to you by nVidia, by their extremely high margin and wildly successful Turing architecture. Exactly what shareholders want to hear.
nVidia has been trying for years to broaden market reach beyond just PC gaming - mobile, datacenter, HPC, AI, Vehicles, etc. They aren't entirely successful at all of it, but if you read their press releases, they are fabulous and the sky is the limit, and their shareholders and Wall Street believe them for the most part. The perception, what those people believe, not the reality of anything - that's what sets the stock price. You can lose money and marketshare year over year, but if your CEO can sell a vision, you can ride that for as long as you can keep the wool over their eyes.
The last two sentences, which you attempt to refute me on, I didn't address at all, and I still don't claim to address them. It's an opinion, your entitled to it, and I don't really have an opinion one way or the other on that particular topic.
Imagine an alternate universe in which the work that had been done on Zen cores up to that time was quite bad. Even if the CEO were able to know that, the decision would have been to start over and do it right or to press on with a terrible product. The latter would surely have led to AMD's bankruptcy. The former would have led to first generation Ryzen launching sometime around today, assuming AMD didn't go bankrupt first, which they probably would have. So, how much credit does Lisa Su get for the good work on Ryzen that happened years before she became CEO?
But that was at least work done within AMD. Maybe just having her around magically makes everyone at AMD really good at their jobs or something. That's not how it works, of course, but let's just suppose. There are plenty of things outside of AMD that have had a huge impact on AMD's stock price.
Imagine an alternate universe in which Intel's foundries didn't trip and fall on their face. Imagine that rather than being delayed and delayed and delayed, Cannon Lake launched and was actually good right on schedule, around the end of 2016. Ice Lake followed the next year and was able to clock decently high. Intel then moved to 7 nm about a year ago, leaving them with about a 20% advantage in IPC, and clearly faster CPUs that use less power than third gen Ryzen or EPYC Rome. And then Intel's second generation 7 nm CPU was due to launch any day now, expanding their advantage.
In that alternate universe, nothing about AMD changes, but only Intel executes well, as they often have historically. Think AMD would still be worth $50/share? AMD's stock price is currently up in anticipation of future profits. AMD already dominates the enthusiast desktop and HEDT markets by any efficiency metric you like. It looks like they'll soon have the far more lucrative x86 server market, though that takes longer to shake out. And in laptops, they'll be a premium option within a few months, not just the budget option that they've been for more than a decade now. That leads to reasonable expectations of large profits coming soon, and that leads to higher stock prices. But none of that would be true if Intel's execution had merely been comparable to what they usually did historically.
Now imagine yet another alternate universe in which both AMD and Intel did what they had done in this one, but Intel's foundry problems were hardly confined to Intel. Rather, TSMC and Samsung also had serious problems with shrinking past 14 nm. Imagine that TSMC 7 nm kind of worked, but yields were terrible, it was horribly leaky, and CPUs couldn't clock very high. Imagine that third gen Ryzen used more power than it does now, but couldn't clock above 3 GHz on simple air or water cooling. Navi-based GPUs wouldn't be much good, either. AMD did everything right on their design work, but TSMC just couldn't build it.
What do you think AMD's stock price would be in that alternate world? We'd still largely be looking at a competition between Sky Lake Refresh Refresh Refresh Refresh and Zen Refresh Refresh, as AMD probably would have done another refresh on 14 nm. AMD wouldn't be in dire trouble if that were the situation, but neither would they be poised for huge profits incoming based on the products they have today.
I don't mean to malign Lisa Su's performance as CEO of AMD. It's hard to distinguish what the CEO is responsible for versus various other people at all levels of the organization--including engineers at the bottom actually designing the chips--but by all accounts, she's done a fine job. But much of AMD's recent success was not her doing, unless you want to credit her for TSMC being well ahead of Intel in process node performance for literally the first time ever. If stock traders had known the day before Lisa Su was hired to be CEO that the foundry situation was going to shake out this way and that Intel would insist on producing their own CPUs anyway, AMD's stock never would have fallen to the $2/share range.
It's one thing to provide component parts. It's another thing to support an entire system. With AMDs renewed capital they should be thinking about how they improve relationships with their OEM partners above most other things. They won't hold the lead over Intel for much longer.