MediaTek announces the first ARMv9 chip and boy if it's powerful!
Paolo Fabio Zaino (28) 1882 posts |
Well, looks like first ARMv9 based CPU is going out with a bang: https://twitter.com/MediaTek/status/1461470026146320384/photo/1 And wondering what Apple is going to announce on autumn 2022… I guess it’ll be an M2 (base chip) for MacBook Air and/or 13", but who knows… |
David J. Ruck (33) 1635 posts |
MediaTek is certainly upping it’s game from their current mid to low end chippery, but I don’t think their ARMv9 is going to be faster than the current best ARMv8.X chips. It will be interesting to see when/if Apple move from v8 to v9, as having their own ecosystem and pretending its Apple Silicon and not really ARM means they could go their own way. I suspect how the maturity of ARMv9 support in llvm will be a clue. |
Charlotte Benton (8631) 168 posts |
In my opinion, this is a lot more exciting than Apple’s M1, as MediaTek/TSMC are primarily interested in selling chips to manufacturers rather than closed system consumer electronics. Whether or not this could be the beginning of the end for x86 remains to be seen, but it’s almost certain that a wide range of ARM based PCs are coming. |
David J. Ruck (33) 1635 posts |
Qualcomm is dusting off their architecture licence and is working on new high end cores, after using off the shelf ARM designs for the last few years. That promises to give the rest of us non fruit users M1 class chips. |
Stuart Swales (8827) 1357 posts |
They’d be daft not to, given how our fruity friends have shown that the way is very very sensible indeed. I’d take 0.1% of that market! |
Paolo Fabio Zaino (28) 1882 posts |
Indeed.
So, on paper, they will be fast, the X2 core @ 3.05Ghz certainly beats the majority of the ARMv8.x chips on the market, not to mention that should be way faster than a Pi4/400. From a product stand point, I can see these chips cover the emerging market of “phone and PC” devices running Android OS and coming with a “fake” laptop to plug-in the phone and transform it into a PC_like device. For more than this, I agree with Druck, it has to be seen when ready.
Indeed. Although, a little corner of my head can stop thinking that it’s a bit of a shame that all this is happening when Aarch32 is reaching EOL, imagine ROS running on one of these CPU :) |
Paolo Fabio Zaino (28) 1882 posts |
After Intel has shown their 12th gen architecture, and the new Foveros 3D packaging, I am quite sure that, between Intel and AMD, x86 is receiving an injection of life. Also, after seen how Intel’s talks with SiFive ended, it seems like Intel has gained “new hopes” around the x86. From a technical point of view it’s hard to insist on x86, it’s way more expensive to improve than ARM or RISC V and it’s no longer a “must-have” like it was back in the 90s. However, it’s still widely used, and surely Intel doesn’t want ARM to become the de-facto standard. We’ll see :) |
Charlotte Benton (8631) 168 posts |
Unless, of course, Intel and AMD decide to make their own high performance ARM chips. Aren’t modern x86 chips basically RISC chips with an x86 emulator in the microcode anyway? |
Andrew Chamberlain (165) 74 posts |
Judging by how long it’s been taking for smartphone technology to filter into the laptop and SBC markets it could be three to four years before we see a Cortex X2 chip in a Chromebook etc. There should be some quite fast ARMv8 devices in the meantime. Rock Pi 5 is due to arrive at some point in the next six months with four Cortex A76 cores and four Cortex A55s in a handy SBC format. That could make a good RO target. |
David J. Ruck (33) 1635 posts |
We saw Intel’s ARM effort with the X-Scale – the worst ARM core ever in terms of instructions per cycle, littered with interlocks and an errata as long as war and peace. If they never touched their architecture licence again it would be too soon. |
Paolo Fabio Zaino (28) 1882 posts |
well, as a really general definition, yes, but I would not push this type of description of what microsocde is. The reason I always stay away from that definition of microcode is that it systematically can confuse people. Microcode is part of the microarchitecture of a CPU (some books refer to microarchitecture as “organisation”), hence the microcode “instructions” are more similar to control signals than to an ISA, for instance each microcode instruction is always followed by a sequence word until the last micro-op which also correspond to the completion of the execution of the corresponded macro-instruction, which is clearly not something that an ISA would presents. However yes, by definition, a microcode control signal must contain everything it needs to execute within a single clock cycle which is something taken from (or should we say “by”, given that microprogramming actually comes before) the RISC philosophy (please note the term “philosophy”). So, yes microcode does present some common elements with the RISC philosophy, but it also differs in nature etc. My apologies for being pedantic, it just that I often have chats where people tend to use the definition you’ve used for microcode and those chats tend to evolve in things like “the x86 is an ARM inside” which is absolutely not a correct definition and not even true. |
Paolo Fabio Zaino (28) 1882 posts |
Indeed, if they have left the AArch32 in the cortex A76. Fingers crossed :) |
Paolo Fabio Zaino (28) 1882 posts |
Oh yes, and if we sum that with their 960, which they then replaced with a StrongARM because it was much better than the 960, we clearly see that Intel has a really bad score at designing RISC CPUs. Which is probably another reason why they were trying to buy SiFive? Intel is probably more interested in controlling a different architecture than popularising ARM even more, although if they did say they will produce ARM chips too at the beginning of this year, but they meant only in therms of manufacturing for now. |
André Timmermans (100) 655 posts |
If your speaking about the new 8cx Gen 3, a French article mentioned that the first results are in Geekbanch: decent but far below the M1. Edit: seeing that the single core results are 3 times that of the ARMV8s Mediatek ones, I think that there is a fair chances that the Qualcomm chip will at least outperform the ARMv9 MediaTek one. |
Peter Howkins (211) 236 posts |
They haven’t, at least not supervisor 32bit. So no RISC OS here. |
Andrew Chamberlain (165) 74 posts |
Just the Linux port then. All this interesting new hardware is going to pass RO by unless some work goes into completing the Linux port. I don’t underestimate the importance of rewriting the OS in C for the long term future of RISC OS, but it seems to me that an opportunity will be missed by not getting it up and running on the faster ARMv8 systems that are arriving now. It’d be a shame not to benefit from the excitement that exists around ARM at the moment. |
David J. Ruck (33) 1635 posts |
Here’s details of Qualcomm’s answer to the Mediatek chip https://www.theregister.com/2021/11/30/qualcomm_snapdragon_8_gen_1/ and in the cloud Graviton 3 is coming https://aws.amazon.com/about-aws/whats-new/2021/11/amazon-ec2-c7g-instances-aws-graviton3-processors/ Things are exciting in 64 bit ARM land. |
Paolo Fabio Zaino (28) 1882 posts |
Oh yes! I saw the details from their Qualcomm’s Summit, very interesting new CPU the 8 Gen 1. Also, recently, I re-shared a video on twitter of the PineBook Pro capable of running quite a lot of modern desktop games, so these new generations will definitely bring a lot to the ARM64 gaming, indeed very exciting :)))) |