https://www.pcper.com/news/Processors/AMD-goes-after-15B-embedded-space-two-new-embedded-processors
Embedded RR APUs support faster memory (3200)?
Embedded RR APUs support faster memory (3200)?
Anarchist4000
Veteran
The slides point towards networking and storage, which makes sense with all the PCIe lanes for IO. The actual processing power could be useful in the case of encryption/hashing, VPN, packet inspection, or routing. Maybe some sort of video editing console, but a big APU would make more sense there.More impressed with embedded 32 core EPYC
No idea what needs that kind of power embedded but wow.
Still wish they had embedded chips with HBM2 for bandwidth as an option to make some nice little NUCs. Seems likely Intel will bring AMD's products to market before they do. Considering AMD's PCIe lane advantage, 10GbE would have been a nice option in 1800/TR parts of the consumer segment. TR boards have been packaging expansion cards to provide the feature and all boards apparently use Intel networking chips over the integrated solution? Not to mention some of the encryption features to double down on security for some segments. For what should have been a disruptive product cycle, their competitors seem to be outpacing them.
Well Dell's EPYC server offerings appeared on their website a few days ago. Price-wise they're not particularly competitive with an equivalent FLOPpage of Xeons. For some specific workloads they may be advantageous I guess. I have a couple of shared-memory parallel codes that will scale to 128 threads, but not enough to make EPYC compelling over Xeon at this time.
Which is a shame.
Which is a shame.
hoom
Veteran
Toms has some interesting details of Precision Boost2 over here http://www.tomshardware.co.uk/amd-ryzen-5-2400g-zen-vega-cpu-gpu,review-34205-3.html in Raven Ridge review
Original Zen implementation was 2 core boost or all core, new one degrades more smoothly across multiple threads not just cores & better handles small tasks on mostly idle cores.
Apparently Raven Ridge has 500Mhz max boost which is a good improvement over 200Mhz max.
Original Zen implementation was 2 core boost or all core, new one degrades more smoothly across multiple threads not just cores & better handles small tasks on mostly idle cores.
Apparently Raven Ridge has 500Mhz max boost which is a good improvement over 200Mhz max.
Maybe not compelling if you need FLOPS. If you need dense virtualization, especially for large capacity IO/RAM tasks, it looks quite good, IMO.
Some cloud providers would probably want to hedge their bets a bit after the Meltdown debacle. Most cloud customers weren't affected much, but some, like Crowdstrike, saw a 30% load increase on their Intel based cloud platforms.
Cheers
Holy crap they had AMD make custom processors for them? That's impressive. I never imagined they had the capital for something like that.
Bondrewd
Veteran
It's not custom.
It uses a volume Ryzen Embedded SKU.
D
Deleted member 13524
Guest
The V1000 series consist of a Raven Ridge directly soldered onto a PCB (embedded).
If anything, the only custom part AMD may have done to/with Smach is the console's PCB, which looks pretty neat IMO:
Apparently, this PCB unit will be replaceable, so people can upgrade to a more powerful APU in the future, without having to purchase a whole new console.
I for one think it would be in AMD's best interest to help the Smach team as much as possible to make the console a reality.
AMD and Valve should be knee-deep in this, as the Switch just proved that there's a vast market for a dedicated portable gaming machine.
ok that's a bit different than an actual APU designed for them, thanks for the clarification.
Valve should be really into it but I guess it conflicts with their SteamOS ideal. I assume they'd still rather push that but either way if they get Steam front and center on a SmachZ, it has to be good for them.
Valve should be really into it but I guess it conflicts with their SteamOS ideal. I assume they'd still rather push that but either way if they get Steam front and center on a SmachZ, it has to be good for them.
Yes indeed, but soldered-in APU cooling fan would need to change though when going to production. Such fans rarely live very long, and it's not reasonable expecting people to break out a soldering iron to replace it.
I'm assuming this is a prototype board, and that such little flaws can/will be corrected before mass production. (One of the VRM sections is also curiously missing inductors... *shrug*)
D
Deleted member 13524
Guest
The Smach-Z alpha hardware with a Raven Ridge is on display in MWC:
https://www.computerbase.de/2018-02/smach-z-steamboy-ryzen-v1605b/
https://liliputing.com/2018/02/smach-z-alpha-handheld-gaming-pc-makes-debut-not-playable-yet.html
Why would it conflict with SteamOS? It's a x86 platform that can take Windows and SteamOS or any other x86 Linux distribution. From a platform perspective it's not that different from Valve's own Steam Machines.
But even if Valve was actually pushing for SteamOS (doesn't seem like it is.. SteamOS looks more and more like a knee-jerk reaction to the Win10 Store that proved to be anything but a menace to Steam), they should definitely welcome a larger variety of devices for them to sell PC games on.
46W.h in this system should be good for 2h-2h30 if the system is being used at full-throttle. That's about the same as the first 3DS and Vita systems.
Using e.g. Radeon Chill and FRTC could increase performance considerably.
https://www.computerbase.de/2018-02/smach-z-steamboy-ryzen-v1605b/
https://liliputing.com/2018/02/smach-z-alpha-handheld-gaming-pc-makes-debut-not-playable-yet.html
Why would it conflict with SteamOS? It's a x86 platform that can take Windows and SteamOS or any other x86 Linux distribution. From a platform perspective it's not that different from Valve's own Steam Machines.
But even if Valve was actually pushing for SteamOS (doesn't seem like it is.. SteamOS looks more and more like a knee-jerk reaction to the Win10 Store that proved to be anything but a menace to Steam), they should definitely welcome a larger variety of devices for them to sell PC games on.
46W.h in this system should be good for 2h-2h30 if the system is being used at full-throttle. That's about the same as the first 3DS and Vita systems.
Using e.g. Radeon Chill and FRTC could increase performance considerably.
Soldered-in? Even without looking at these latest pictures of the PCBs, you could see the screws in the cooler.
Anarchist4000
Veteran
Not custom as much as tailored towards their application with firmware most likely. 10GbE, multiple display outputs, SATA/NVMe, etc that would be supported would be of limited use on a handheld.
I'm not sure that works well in practice. That would essentially mean replacing everything but the display, battery, and buttons. Interesting concept, but that board would be most of the cost. Faster APU with the same RAM, storage, etc may be of limited benefit.
Depends on where they were taking SteamOS. Designing all games for linux/SteamOS with the intention of encapsulating them in virtual machines for sandboxing would be plausible. Mobile platforms already do that, but GPU pass-through could allow it to scale to more demanding games. Aligns with Intel's design with integrated GPU plus Vega. That should allow games to more easily jump platforms, but would require a push from the status quo or the ability to encapsulate Windows.
Always external or possibly interchangeable batteries to extend that life if really gaming hard. I doubt there is a good power solution for attempting PC gaming on a handheld without really expensive hardware. That would require over-sized/cored silicon, HBM, etc to cut power as much as possible to achieve similar performance to a system using 100W or more.
The power leads, man... The power leads.
monstercameron
Newcomer
It would’ve been interesting to set per game clocks for the cpu, gpu and memory. I think this would lead to more stable performance vs cTDP throttling.
Sent from my iPhone using Tapatalk
Sent from my iPhone using Tapatalk
D
Deleted member 13524
Guest
Perhaps AMD will launch a version of the Ryzen Master that only allows downclocking and lowering the max turbo values for the Ryzen U and V series, leaving total TDP and max boost values as fixed.
Or the BIOS may just lock the CPU cores at its base values. For example, the original BIOS in my old Clevo W110ER locks the CPU cores at the base 2.3GHz clocks whenever the dGPU is active.
Since the Smach-Z is meant for gaming first and foremost, we probably wouldn't need a 3.6GHz maximum boost in the CPU, and with the iGPU at hand it probably wouldn't be held back that much if the CPU cores were kept in the 2.0-2.4GHz range.
Some more leaked numbers for Zen+:
https://videocardz.com/75185/first-benchmarks-of-ryzen-7-2000-cpu-have-been-leaked
Access latency for L2, L3 and memory is lower.
https://videocardz.com/75185/first-benchmarks-of-ryzen-7-2000-cpu-have-been-leaked
Access latency for L2, L3 and memory is lower.
Similar threads
