12 ms·
AMD Ryzen Full Lineup Prices, Specs and Clock Speeds Leaked
- jhoechtl 10y agoGiven the price and the feature comparison to Intel, Ryzen sounds awesome. Hopefully it will spark new innovation between AMD and Intel. And hopefully AMD will life up to where it was once when they gave name to the AMD64 architecture.
- chrisseaton 10y ago> feature comparison Are there interesting features that differentiate them from Intel beyond performance/cost? As in new instructions or things like that?
- dogma1138 10y agoNo, and as far as high end instructions then only 128bit AVX is natively supported on Zen, with 256bit AVX possible with half the throughput and no 512bit AVX support at all. That said because Intel disables AVX2 in half its SKU's at least that will likely not to be a major impact. As far as other features it's still TBD, virtualization works better on Intel currently, some other platform dependant issues would also probably still favor intel. Intel makes their storage, networking and USB in house, AMD doesn't, AM3 storage and USB performance were considerably slower than Intel's solution and it's yet to be seen what AM4 will provide. AM4 will come with USB3.0/3.1 from ASMEDIA and will only support a single NVME drive, so if you are looking for a professional workstation, need high speed storage, ECC, and can utilize quad channel low latency memory then Intel is still likely the way to go even with blue team tax. If you need multi GPU support and NVME at the same time, as well as other peripherals that use PCIE then you also might need to look at intel still since Zen will only come with 28 PCIE lanes. Hopefully they've sorted their PCIE performance issues at least, AM3 not only had bandwidth limitation due to PCIE 2.0 support but also had a much higher PCIE latency for unexplained reasons.
- mcbain 10y agoWhich Intel processors have AVX2 disabled?
- dogma1138 10y agoCore based Pentium/Celeron and some of the ULV CPUs have had AVX disabled since Ivy Bridge.
- mcbain 10y agoA little overstated to say that's half their SKUs, but I see your point.
- nickpeterson 10y agoThere is even more nuanced though, because Intel runs avx instructions at a different clockspeed than the main chip. So certain SKUs have avx support but have differences that don't reflect the CPU clock difference.
- dogma1138 10y agoThere is even more nuance to this as AVX low level instruction support isn't going to be universal as different CPUs will support different instruction subsets of AVX-512 and Extended AVX https://en.wikipedia.org/wiki/AVX-512 https://en.wikipedia.org/wiki/AVX-512 There is also a mess in FMA support for both Intel and AMD cpus... https://en.wikipedia.org/wiki/FMA_instruction_set https://en.wikipedia.org/wiki/FMA_instruction_set This entire deal pretty much turned AVX into cancer outside of very specific circles, especially considering the wierd support for AVX and SSE in OpenCL.
- pooloo1 10y agoI think what AMD is really trying to do, or what they should do, is place emphasis on the gaming community and drop any/all enterprise related features, and let Intel keep doing that part.
- nattmat 10y agoZen should feature instructions for full memory encryption using a ARM co-prosessor. I find that pretty interesting, considering the limitations of Intel SGX.
- zamalek 10y agoThey are still somewhat in the domain of performance. * ML is used for branch prediction,[1] as well as [implied] pre-fetch. * Automatic clock and power tuning.[2] This is not per-die, but granular across the chip. * eXtended Frequency Range,[3] unlocks the ranges that [2] is allowed to use. Essentially, the chip overclocks itself when you provide it with more cooling headroom. [1]: https://youtu.be/X9NNOqzTbKI?t=13m53s https://youtu.be/X9NNOqzTbKI?t=13m53s [2]: https://youtu.be/X9NNOqzTbKI?t=14m38s https://youtu.be/X9NNOqzTbKI?t=14m38s [3]: https://youtu.be/X9NNOqzTbKI?t=15m22s https://youtu.be/X9NNOqzTbKI?t=15m22s
- tyingq 10y agoAssuming similarly low pricing, AMD's Naples chip for the servers might spark even more competition. Since it can do up to 32 cores / 64 threads in a single socket, it would compete with dual socket Xeon systems that have much higher motherboard costs, plus the cost of 2 chips.
- 65827 10y agoI'd really like eight cores, didn't know that was going to be possible soon!
- rincebrain 10y agoPossible on consumer chips, or possible? Because Intel and AMD have both shipped 6+ core parts for some time now, just usually not targeting the consumer segment... IIRC when asked, Intel folks say they haven't make a consumer 6+ core part because there's little benefit/demand, but I can't find a citation for that. [1] is a nice discussion about why we haven't gotten more than 4 cores in consumer parts yet, though. [1] - https://news.ycombinator.com/item?id=12304046 https://news.ycombinator.com/item?id=12304046
- rbanffy 10y agoDepending on how much you want it, Intel has an E7 that has 24 cores and 48 threads. It can go up to 8 sockets with up to 12 TB of RAM. I don't even want to know how much such a monster would cost.
- chx 10y agoMuch less than you'd think if you are willing to go with 6TB only. The 128GB modules are awfully expensive but the other day I saw 64GB modules at $600 https://memory.net/product/s26361-f3843-e618-fujitsu-1x-64gb-ddr4-2133-lrdimm-pc4-17000p-l-quad-rank-x4-replacement/ https://memory.net/product/s26361-f3843-e618-fujitsu-1x-64gb... here. So that's ~39K for 6TB of RAM. The relevant Supermicro server (4048B-TR4FT) with four E7 CPUs is ~14K http://www.thinkmate.com/system/superserver-4048b-tr4ft http://www.thinkmate.com/system/superserver-4048b-tr4ft comes with 32 4GB modules which is a waste but oh well, it's a very small loss compared what it would cost to buy 6TB at a server seller like this. So you are looking ~53K for the server, CPU, 6TB RAM, you will need a little storage as well, so perhaps ~60K for a 6TB 4U server. In closing, let me remind you that a mere two decades ago, Microsoft launched the Terraserver which was a demonstration of a server with a terabyte of storage and it was very big news. Now we are here with a machine at the price of an ordinary F-150 truck or so which can hold the 5TB uncompressed data of the Terraserver in RAM.
- gizmo 10y agoDoes anybody know if the Ryzen CPUs support the equivalent of VT-x and VT-d?
- Avshalom 10y agohttps://en.m.wikipedia.org/wiki/X86_virtualization#AMD_virtualization_.28AMD-V.29 https://en.m.wikipedia.org/wiki/X86_virtualization#AMD_virtu... I don't know how well its supported by software though.
- trome 10y agoKVM & Xen have good support for both in my experience, seems to work well. No clue about support in janky proprietary solutions, but seeing as Amazon, DO, etc all use either KVM or Xen, with the extreme budget providers on OpenVZ (where you get a glorified chroot), your probably gonna be using one of those two.
- j3097736 10y agoWhy wouldn't it?
- dippyskoodlez 10y agoIOMMU and AMD-V have been around for ages, I would expect so.
- untoreh 10y agoYes, and they don't (generally) feature lock like Intel does, but we will have to see
- jbmorgado 10y agoI wonder how well they will support some more advanced virtualisation. I.E. Booting a virtual machine with PCI GPU passtrough in Windows Guest with Linux host for gaming, or booting a virtual machine with PCI GPU passtrough in Linux Guest with Linux host for some deep learning with CUDA.
- pavanky 10y agoWhy would you want to do Linux Guest with Linux host for CUDA?
- jbmorgado 10y agoBecause it's (well, at least 6 months ago was, now it could have improved) the only solution to reuse the GPU without rebooting X. With a practical example: 1 - You boot up your Linux base system. 2 - You fire up KVM with GPU PCI passthrough with a Windows guest to play some games. 3 - You shutdown windows. Now you can't use the GPU anymore, the module was assigned by KVM and you can't - for instance - run a CUDA simulation in the Linux host. You need to reboot X (you don't need to reboot the system). The workaround/solution is to fire up KVM with GPU PCI passthrough to another guest (This time a Linux guest) and in there you will have full access to the GPU to do CUDA computations (or whatever else you want).
- pavanky 10y agoAh I see, that makes more sense. If you had two NVIDIA GPUs while one of them was passed through to a guest then I think that problem may still exist. However if you only have one nvidia gpu (and have a different kind of gpu for display), then I think simply running "sudo rmmod vfio" "sudo modprobe nvidia" would work.
- pYQAJ6Zm 10y ago>Now you can't use the GPU anymore, the module was assigned by KVM and you can't - for instance - run a CUDA simulation in the Linux host. Maybe we had different setups, so this wouldn’t apply – I used to unbind the device with the following script, and then load the `nvidia` module; the device was then available on the host: for dev in "0000:01:00.0" "0000:01:00.1"; do if [ -e /sys/bus/pci/devices/${dev}/driver ]; then echo "${dev}" > /sys/bus/pci/devices/${dev}/driver/unbind fi done
- hmottestad 10y agoNone of these chips are for mobile. TDP on the chips in the Macbook Pro is 47W, while lowest TDP of the leaked Ryzen chips is 65W. What could be of great interest would be if Ryzen came with mobile chips with LPDDR4, that would really put pressure on intel. Then they could also do 50W TDP and still compete with Intel for the Macbook Pros.
- dogma1138 10y agoThere aren't any Zen chips for mobile, there aren't any Zen "APUs" which have an integrated graphics card either. The AM4 APUs that are launching with Summit Ridge are still construction cores and they come at 65-90W TDP also. Until low TDP Zen chips are out and until Zen APUs are out AMD is effectively forfeiting the OEM and mobile markets. Intel has also managed to scale it's current Core based CPU's to as low as 4.5W TDP which I have a very strong feeling that AMD will not be able to do until 2019-2020 at best, not with the TDP for these chips currently. That said if AMD does release a 40-45W 4C CPU they'll have a strong chance of being a contender for at least some workstation level laptops, lack of native USB3 and no TB might hurt them but if PCIE over USB-C beats TB then they might still have a chance. However since the external GPU enclosures became reliable I will never buy a laptop that doesn't support one for personal use unless it's a very very small form factor device and that's unlikely. Lets see if AMD will be able to come with a similar solution to how seamless TB2/3 with Iris graphics and an external GPU work these days.
- AlphaSite 10y agoAMD already has such a solution and Raven Ridge (Zen APU) is coming 2nd half 2017 (as far as anyone is aware).
- dogma1138 10y agoIt has, it's not out yet, currently it just relaunched its construction core Bristol Ridge APUs. Once Raven Ridge will be out we can discuss it, currently Summit Ridge is nearly unknown.
- 10y ago
- std_throwaway 10y agoDo they support ECC memory?
- 65827 10y agoHopefully not, biggest scam out there.
- throwawaydbfif 10y agoThe biggest scam is actually the other way around. Intel pulled ECC support from desktop processors a few years ago to force datacentres to buy their Xeons instead. It's nearly impossible to find a desktop CPU that supports ECC ram now even though 5 years ago it was commonplace. Trying to run a NAS with some sensitive data is now impossible unless you buy their server chips
- dogma1138 10y agoIf you going the workstation route then 2011 Xeon CPU's do support ECC and are cheaper than Desktop HEDT parts. Getting an Intel Xeon E3 for socket 1151 isn't particularly hard or expensive either.
- throwawaydbfif 10y agoYeah have fun buying the motherboard that supports it. The whole thing reeks of bullshit
- mcbain 10y ago> It's nearly impossible to find a desktop CPU that supports ECC ram now That's not correct. All 6th Gen Core i3 have ECC support, and the 7th Gen Core i3 that have 'E' in the product name support ECC: http://ark.intel.com/products/97130/Intel-Core-i3-7101E-Processor-3M-Cache-3_90-GHz http://ark.intel.com/products/97130/Intel-Core-i3-7101E-Proc...
- 10y ago
- fictioncircle 10y agoWell, this sounds good and I'm happy I am holding out for my new machine until it comes out. We'll have to see real world benchmarks by independent 3rd parties to validate the performance/$ but it continues to look quite impressive as long as you are staying in the sub $500 range.
- VodkaHaze 10y agoThe next step up from those CPUs really is going to be dual xeon boxes, which is a huge jump. Intel tried to do enough market segmentation that the curve from a Pentium all the way up to dual CPU workstations was fairly continuous in price, with zen it's not the case anymore.
- shmerl 10y agoAMD Ryzen 7 1700, 8 cores, 16 threads, 65W TDP, 3.0-3.7GHz, $319. Comparing to my current Haswell CPU, that's quite a great value for the money: (84W TDP, 4 cores, 8 threads, $300+ price soon after launch, 3.4-3.9GHz). I'd prefer somewhat higher frequency though, more in line with thier X series, but without major increase in TDP. But I guess +10W is tolerable for a good processing power increase.
- O5vYtytb 10y ago4 cores / 16 threads? Does clock speed matter if IPC for your workload is good at a lower frequency?
- shmerl 10y agoI guess it matters for single core performance bottlenecks. I.e. let's say you are compiling. Higher frequency CPU will manage it in less time. More cores naturally help as well.
- ganeshkrishnan 10y agoMost compilers should be able to handle multi-threaded compiling. I would rather go for more cores than higher frequencies
- shmerl 10y agoSure, but higher frequency would speed it up on each core, right? The benefits aren't mutually exclusive.
- onli 10y agoI'd be hesitant to judge those AMD processors on the specified clock, even more than usual. The usual argument is that clocks are not comparable, because you don't know that a 4GHz cpu A is not slower than a 1GHz cpu B in doing a specific task. That I assume is known around here, but is not what I mean. AMD is marketing the feature of those processors to overclock automatically above the specified turbo clock. Meaning 3.9GHz should be a lower bound, and it is completely possible those cpus will routinely clock much higher in practice. See http://www.kitguru.net/components/cpu/jon-martindale/amds-extended-frequency-range-could-auto-overclock-ryzen-cpus/ http://www.kitguru.net/components/cpu/jon-martindale/amds-ex... (and it is also as XFR in the original article here). Or maybe they won't overclock well at all. Well, we'll see after they were released. Edit: I just realized that with regards to the specific comparison I answered to my point is moot, because according to the table in the article, the Ryzen 7 1700 does not have that feature. The Ryzen 7 1700X does. That might make picking the right processors more difficult than usual. I'll let the comment stand regardless, that feature and distinction might not be well known yet.
- sengork 10y agoI certainly hope that the performance/price is on par with the competition Intel used to get from AMD back in the Duron and Athlon XP days. Ryzen seems to be fairly low power consumption as well which should help with lowering the cooling requirements compared to Intel's CPUs.
- ilaksh 10y agoWill it work with Keras?
- rkrzr 10y agoThis would be great news indeed if AMD finally had a competitive high-end chip again. It feels like Intel has just been coasting on their success for a long time, since there was no competitive pressure from the outside they also had little incentive to add more cores/lower prices/take some engineering risks. The CPU market could become a lot more interesting again after all this time.
- pmoriarty 10y agoHow long is it likely to take for Linux to support these new chips? I need to build a new system and am trying to gauge how long I should wait after Zen comes on the market.
- notaplumber 10y agoThere's nothing specifically required by a kernel to support new processors, except maybe power saving features, arguably unimportant for desktop. If you mean compiler optimizations, those are usually sent upstream way early.. A cursory Google search shows AMD first sending Zen (now Ryzen) patches to GCC as early as 2015. So, presumably it has been in a few recent compiler releases already. https://gcc.gnu.org/ml/gcc-patches/2015-09/msg02311.html https://gcc.gnu.org/ml/gcc-patches/2015-09/msg02311.html
- pmoriarty 10y ago"There's nothing specifically required by a kernel to support new processors..." Does that go for motherboards that can support these new chips too?
- notaplumber 10y agoNot specifically, if there's a new chipset or southbridge it may require kernel driver update (..perhaps as simple as new PCI IDs) or a new driver entirely, but similarly those changes are upstreamed rather quickly to increase the change of landing in a stable release. AMD and Intel engineers have hardware quite early to work on this, long before the consumer product ships.
- frik 10y agoAMD announced Ryzen supports Win7 (drivers). A few days later, they changed the marketing to Win10. If AMD wants to sell their new Ryzen make sure Win7 drivers are available (even if it's not marketed, as Microsoft is bitter about it - Win7 is still most used OS).
- mstaoru 10y agoOfficial retail prices in China and real retail prices in China are not always the same. "White"-imported i7 6900K the companies will buy is ¥8199, but most people will buy from the marketplaces, where the price is about ¥7000-7400, i.e. the same as in the US. Then again, for cheap servers, we can get E5-2683v3 here for ¥1900 ($270), 35M cache, 14 cores.