So are these just Binnings of the existing laptop/desktop APUs? The top end SKU seems to be pretty close to the 2600H or 2800H. Has a bit high CPU clocks than the 2600H but the 2800H has 11 CUs. The lowest end one seems to be pretty close to the Ryzen 3 3200U. Seems strange to not have the 2 higher end models if they are just reusing stuff that was released several years ago.
Its common to reuse existing designs under different branding. In this case, it probably took a bit of time for AMD to finish the necessary validation to sell these as embedded processors for industrial usage. Better to test out chips in the consumer market and let stupid end users absorb the first few silicon revs before risking those same designs in longer lifespan hardware that will reside in places like assembly line control systems or thin clients.
The R1000 Banded Kestrel (2019) family has very similar specs to what was released as Athlon 3000 Dalì (2020). Both top out at 2 cpu cores and 3 gpu compute units. When compared to Ryzen 3000U Picasso (2019), the R2000 family has similar cpu core count (up to 4), but a little lower gpu compute unit count with only 8 compared to 11 (still close). What I found most interesting is that R2000 supports faster memory speeds at 3200 MHz, with Ryzen 3000U only supporting 2400 MHz. Another reason to move to the larger die is the 16 PCIe 3.0 lanes, up from 12.
When comparing SKUs: The top R2544 is essentially a Ryzen 3550H and R2514 is like Ryzen 3500U. The R2314 is similar to the Ryzen 3300U. The R2312 is an odd one, having half of most resources compared to the other R2000s. When compared to the old R1000 family, it has less L3 cache (and PCIe lanes I think), reducing any generational gains.
These are not exciting at all. I was pretty stoked with their first-gen product, the v1605b that was found on a SBC. The Udoo Bolt V8 is arguably still the most powerful SBC out there, despite using 2016-era technology.
I've been hoping AMD would give the crimson treatment for all product segments. As in use the latest node, with the latest CPU architecture and latest GPU architecture. Think: 2017 + 16nm + Zen1 + Vega... 2019 + 8nm + Zen2 + RDNA2... 2021 + 6nm + Zen3 + RDNA2.... 2023 + 4nm + Zen4 + RDNA3. Whilst the Thermal range from the 5W, 10W, 15W, 25W, 40W, and all the way up to the likes of the r9-3950x. They just need to make a compelling case even if produced in a small batch. You never know how popular they could get.
I agree that by now, we should normally expect at least Zen2. However, I expect TSMC N7 production is still under too much strain to hit the volume/price targets AMD has in mind. Meanwhile, they probably still have that wafer supply agreement in place with GF, giving them easy access to 12 nm.
Exciting? This is for a market that isn't willing to pay more for more performance. Underclocked 12nm Zen+ is OK for them, because cost per transistor is lower so they'll pay less for each IP block/feature that they include (E.g. for driving three 4K displays).
It's why the Raspberry Pi is stuck with a 28nm chip despite being fanless - 28nm is simpler to fab than FinFET nodes, so it's still got the lowest transistor cost.
If you want a powerful low-power machine, you go laptop or mini PC. There's not a market for powerful low-power SBC's. Consumers have no use for it, and anyone else would rather pay less for the hardware and let it pull more power (Resulting in more perf/$)
> If you want a powerful low-power machine, you go laptop or mini PC.
Due to supply chain issues, Pi is still stupidly expensive. For how much it costs, you'd do a lot better with one of the new RK3588 boards, finally starting to hit the market:
I disagree. If there is No Product, you cannot conclude there is No Demand. What AMD should have done is to setup a division which creates these high-end products. They can release these in a small batch (1,000 - 9,000 units), even give them away free, to staff, developers, influencers, and certain fans. Things like that do a lot of good-will and positive advertising.
Once they do something like this, they can see if there is demand from consumers and vendors. Then they can look into making mass produced runs of specific chipsets. Again, AMD has a huge advantage to scale things up anywhere from 5W to 200W, and they should take advantage of it. It won't really hurt their bottom-line to invest a fraction of their silicon-allocation into these adventures.
> What AMD should have done is to setup a division which creates these high-end products.
I think you're trying to solve a different problem than AMD. I think these embedded products tend to be lower-margin, higher-volume. When their N7 production is constrained, they have a disincentive to use their existing wafer supply to service this market.
> even give them away free, to staff, developers, influencers, and certain fans. > Things like that do a lot of good-will and positive advertising.
I disagree. If there is No Product, you only hurt yourself by creating more Demand.
As a consumer, if there's a product that I want to buy & I cannot get it, that only creates frustration. Case in point: Ryzen 3300X. Before the latest round of Zen 3 CPUs and Alder Lake i3's launched, this little CPU was the ultimate performance bargain. Unfortunately, it was only available for a short while, and then disappeared for years at a time! To those of us who wanted one, it was so frustrating to know that it existed but simply wasn't available.
> Then they can look into making mass produced runs of specific chipsets.
They already look into making mass produced runs of chiplets. That's their business, after all!
> It won't really hurt their bottom-line to invest a fraction of their silicon-allocation > into these adventures.
How do you know? Tell us: how big is this market, in terms of units sold annually. And if the market is so tiny that it wouldn't hurt to divert that many chips from their more profitable products, then you've got to ask whether it would be worth the overhead/NRE costs, as well as the distraction of attention and focus from more critical areas of their business.
Still disagree, because your point doesn't make logical sense. You say you KNOW there is ZERO demand BEFORE you even TEST the premise.
Also your analogy sucks, since you/other were never upset at the 3300x. You were merely upset at losing a bargain. As an example, if you personally were shopping for a 3300x and missed out, but managed to get a 3600 instead a week later at the same price (or maybe same value rating). Well you would stop complaining and forget about the 3300x. So there's an obvious solution. Now consider the opposite. Let's say you have a Zen1 rig and were looking to upgrade and saw the 3950x, but you missed out on buying it. Even if you bought a different chipset, like the 3900x you wouldn't feel as satisfied. That's what you're missing out on. Better to have had something than none at all. And if AMD did a limited run on the 3950x and found a promising market, they may strive for a second-batch of limited runs, or to make this a permanent part of the product portfolio, or do a return with the following generation of chipsets (ie 5950x).
There is an important nuance to consider between a "flagship product" versus a "good value product". Sometimes they can overlap, but they are intrinsically different markets.
It wouldn't hurt AMD's backpocket because "just trust me bro". Think about it, a unit allocation of 9000 quantity of a chip is nothing for the likes of Intel, AMD, Nvidia, Qualcomm, Samsung, MediaTek etc etc. But if you send all of those out, for free, to some key people such as developers, enthusiasts, and those with a platform you could get a lot of public-goodwill. And they can give you aid to make the product better, and give you free advertising to boot. Under the preface of "limited time product" or "prototype" etc. This is only for your niche products... your main products should still focus on traditional unveiling, distribution, and marketing. Sometimes you can have a No Market which evolves into an Enthusiast Market and perhaps later into a Mainstream Product. But there really is no way to know without doing the work. Netflix was supposed to have failed hundreds of times, but they found a niche, and their enthusiast market, then it evolved into the mainstream. As opposed to Blockbuster which had been the market leader for decades which vanished completely.
That is how the embedded V-series works. Zen 1 V1000 family came Dec 2018, Zen 2 V2000 in Nov 2020 and a Zen 3 V3000 is expected before the end of this year. I'd expect V3000 to feature Vega graphics.
This product isn’t even made for you it’s not meant to be exciting. Did you even read the article? It serves a distinct purpose and does that really well. You’re a classic example for a nerd that takes himself too seriously without having the knowledge to back himself up. Pretty cringe to be honest.
I love the idea of a low power durable SOC motherboard. preferably with dual nic. But the AMD options where priced like i7 options. I would be buying them instantly when they where like 100 euros for instance. Not to start about their Epyc brothers.
You don't have to be hung up too much on dual NIC, when you have USB3 ports.
While my main motivation was actually getting better bandwidth than GBit on my Gemini Lake Atom servers (SATA SSDs), these RealTek based 2.5Gbit USB3 NICs have become both cheap enough and use little enough power to even work with a PI4 or Jetson Nano (I just tested them there for the fun of it).
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
20 Comments
Back to Article
kpb321 - Wednesday, June 22, 2022 - link
So are these just Binnings of the existing laptop/desktop APUs? The top end SKU seems to be pretty close to the 2600H or 2800H. Has a bit high CPU clocks than the 2600H but the 2800H has 11 CUs. The lowest end one seems to be pretty close to the Ryzen 3 3200U. Seems strange to not have the 2 higher end models if they are just reusing stuff that was released several years ago.PeachNCream - Thursday, June 23, 2022 - link
Its common to reuse existing designs under different branding. In this case, it probably took a bit of time for AMD to finish the necessary validation to sell these as embedded processors for industrial usage. Better to test out chips in the consumer market and let stupid end users absorb the first few silicon revs before risking those same designs in longer lifespan hardware that will reside in places like assembly line control systems or thin clients.Rudde - Friday, June 24, 2022 - link
The R1000 Banded Kestrel (2019) family has very similar specs to what was released as Athlon 3000 Dalì (2020). Both top out at 2 cpu cores and 3 gpu compute units. When compared to Ryzen 3000U Picasso (2019), the R2000 family has similar cpu core count (up to 4), but a little lower gpu compute unit count with only 8 compared to 11 (still close). What I found most interesting is that R2000 supports faster memory speeds at 3200 MHz, with Ryzen 3000U only supporting 2400 MHz. Another reason to move to the larger die is the 16 PCIe 3.0 lanes, up from 12.When comparing SKUs: The top R2544 is essentially a Ryzen 3550H and R2514 is like Ryzen 3500U. The R2314 is similar to the Ryzen 3300U. The R2312 is an odd one, having half of most resources compared to the other R2000s. When compared to the old R1000 family, it has less L3 cache (and PCIe lanes I think), reducing any generational gains.
Kangal - Thursday, June 23, 2022 - link
These are not exciting at all. I was pretty stoked with their first-gen product, the v1605b that was found on a SBC. The Udoo Bolt V8 is arguably still the most powerful SBC out there, despite using 2016-era technology.I've been hoping AMD would give the crimson treatment for all product segments. As in use the latest node, with the latest CPU architecture and latest GPU architecture. Think: 2017 + 16nm + Zen1 + Vega... 2019 + 8nm + Zen2 + RDNA2... 2021 + 6nm + Zen3 + RDNA2.... 2023 + 4nm + Zen4 + RDNA3. Whilst the Thermal range from the 5W, 10W, 15W, 25W, 40W, and all the way up to the likes of the r9-3950x. They just need to make a compelling case even if produced in a small batch. You never know how popular they could get.
mode_13h - Thursday, June 23, 2022 - link
I agree that by now, we should normally expect at least Zen2. However, I expect TSMC N7 production is still under too much strain to hit the volume/price targets AMD has in mind. Meanwhile, they probably still have that wafer supply agreement in place with GF, giving them easy access to 12 nm.Wereweeb - Thursday, June 23, 2022 - link
Exciting? This is for a market that isn't willing to pay more for more performance. Underclocked 12nm Zen+ is OK for them, because cost per transistor is lower so they'll pay less for each IP block/feature that they include (E.g. for driving three 4K displays).It's why the Raspberry Pi is stuck with a 28nm chip despite being fanless - 28nm is simpler to fab than FinFET nodes, so it's still got the lowest transistor cost.
If you want a powerful low-power machine, you go laptop or mini PC. There's not a market for powerful low-power SBC's. Consumers have no use for it, and anyone else would rather pay less for the hardware and let it pull more power (Resulting in more perf/$)
mode_13h - Friday, June 24, 2022 - link
> If you want a powerful low-power machine, you go laptop or mini PC.Due to supply chain issues, Pi is still stupidly expensive. For how much it costs, you'd do a lot better with one of the new RK3588 boards, finally starting to hit the market:
https://www.cnx-software.com/2022/05/12/mekotronic...
For less money and still better specs than a Pi (except limited to 4GB), there's ever the ODROID-N2+
https://www.hardkernel.com/shop/odroid-n2-with-4gb...
mode_13h - Friday, June 24, 2022 - link
BTW, RK3588 is 4x A76 @ 2.4 GHz + 4x A55 @ 1.8 GHz. I think it could probably hold its own against a Tremont-powered laptop or mini PC.Here's one on a mini-ITX board, but it's marketed at industrial applications & therefore has a starting price of $459:
https://en.t-firefly.com/product/industry/itx3588j
Kangal - Friday, June 24, 2022 - link
I disagree.If there is No Product, you cannot conclude there is No Demand. What AMD should have done is to setup a division which creates these high-end products. They can release these in a small batch (1,000 - 9,000 units), even give them away free, to staff, developers, influencers, and certain fans. Things like that do a lot of good-will and positive advertising.
Once they do something like this, they can see if there is demand from consumers and vendors. Then they can look into making mass produced runs of specific chipsets. Again, AMD has a huge advantage to scale things up anywhere from 5W to 200W, and they should take advantage of it. It won't really hurt their bottom-line to invest a fraction of their silicon-allocation into these adventures.
mode_13h - Friday, June 24, 2022 - link
> What AMD should have done is to setup a division which creates these high-end products.I think you're trying to solve a different problem than AMD. I think these embedded products tend to be lower-margin, higher-volume. When their N7 production is constrained, they have a disincentive to use their existing wafer supply to service this market.
> even give them away free, to staff, developers, influencers, and certain fans.
> Things like that do a lot of good-will and positive advertising.
I disagree.
If there is No Product, you only hurt yourself by creating more Demand.
As a consumer, if there's a product that I want to buy & I cannot get it, that only creates frustration. Case in point: Ryzen 3300X. Before the latest round of Zen 3 CPUs and Alder Lake i3's launched, this little CPU was the ultimate performance bargain. Unfortunately, it was only available for a short while, and then disappeared for years at a time! To those of us who wanted one, it was so frustrating to know that it existed but simply wasn't available.
> Then they can look into making mass produced runs of specific chipsets.
They already look into making mass produced runs of chiplets. That's their business, after all!
> It won't really hurt their bottom-line to invest a fraction of their silicon-allocation
> into these adventures.
How do you know? Tell us: how big is this market, in terms of units sold annually. And if the market is so tiny that it wouldn't hurt to divert that many chips from their more profitable products, then you've got to ask whether it would be worth the overhead/NRE costs, as well as the distraction of attention and focus from more critical areas of their business.
Kangal - Thursday, July 21, 2022 - link
Still disagree, because your point doesn't make logical sense.You say you KNOW there is ZERO demand BEFORE you even TEST the premise.
Also your analogy sucks, since you/other were never upset at the 3300x. You were merely upset at losing a bargain. As an example, if you personally were shopping for a 3300x and missed out, but managed to get a 3600 instead a week later at the same price (or maybe same value rating). Well you would stop complaining and forget about the 3300x. So there's an obvious solution. Now consider the opposite. Let's say you have a Zen1 rig and were looking to upgrade and saw the 3950x, but you missed out on buying it. Even if you bought a different chipset, like the 3900x you wouldn't feel as satisfied.
That's what you're missing out on. Better to have had something than none at all. And if AMD did a limited run on the 3950x and found a promising market, they may strive for a second-batch of limited runs, or to make this a permanent part of the product portfolio, or do a return with the following generation of chipsets (ie 5950x).
There is an important nuance to consider between a "flagship product" versus a "good value product". Sometimes they can overlap, but they are intrinsically different markets.
It wouldn't hurt AMD's backpocket because "just trust me bro". Think about it, a unit allocation of 9000 quantity of a chip is nothing for the likes of Intel, AMD, Nvidia, Qualcomm, Samsung, MediaTek etc etc. But if you send all of those out, for free, to some key people such as developers, enthusiasts, and those with a platform you could get a lot of public-goodwill. And they can give you aid to make the product better, and give you free advertising to boot. Under the preface of "limited time product" or "prototype" etc. This is only for your niche products... your main products should still focus on traditional unveiling, distribution, and marketing. Sometimes you can have a No Market which evolves into an Enthusiast Market and perhaps later into a Mainstream Product. But there really is no way to know without doing the work. Netflix was supposed to have failed hundreds of times, but they found a niche, and their enthusiast market, then it evolved into the mainstream. As opposed to Blockbuster which had been the market leader for decades which vanished completely.
Rudde - Friday, June 24, 2022 - link
That is how the embedded V-series works. Zen 1 V1000 family came Dec 2018, Zen 2 V2000 in Nov 2020 and a Zen 3 V3000 is expected before the end of this year. I'd expect V3000 to feature Vega graphics.Khanan - Sunday, June 26, 2022 - link
This product isn’t even made for you it’s not meant to be exciting. Did you even read the article? It serves a distinct purpose and does that really well. You’re a classic example for a nerd that takes himself too seriously without having the knowledge to back himself up. Pretty cringe to be honest.krumme - Thursday, June 23, 2022 - link
How to live up to your wsamode_13h - Thursday, June 23, 2022 - link
Well, at least they're not backporting Zen2 to 12 nm. It'd be very telling, if they did something like that.Foeketijn - Friday, June 24, 2022 - link
I love the idea of a low power durable SOC motherboard. preferably with dual nic.But the AMD options where priced like i7 options.
I would be buying them instantly when they where like 100 euros for instance. Not to start about their Epyc brothers.
abufrejoval - Saturday, July 23, 2022 - link
You don't have to be hung up too much on dual NIC, when you have USB3 ports.While my main motivation was actually getting better bandwidth than GBit on my Gemini Lake Atom servers (SATA SSDs), these RealTek based 2.5Gbit USB3 NICs have become both cheap enough and use little enough power to even work with a PI4 or Jetson Nano (I just tested them there for the fun of it).
pogsnet1 - Thursday, July 21, 2022 - link
The low-end can be useful in feature rich internet routers