Back in May AMD made Ryzen Mobile official indicating that the APUs previously known as "Raven Ridge" would be launching in the second half of 2017. As that launch window closes, more details are starting to trickle out including benchmarks scores. The latest appearance of Raven Ridge is in a Geekbench benchmark score results page where a "Ryzen 5 2500U" APU achieves a single core score of 3,561 and a multi-core score of 9,421. These are fairly impressive results on their own, but especially considering that Ryzen Mobile chips are reportedly using up to 50% less power versus last generation Bristol Ridge processors while handily beating them in performance offered.
AMD has previously claimed that its Ryzen Mobile (Raven Ridge) APUs will offer up to 50% more CPU performance and 40% more GPU performance compared to its 7th Generation APUs. The leaked Geekbench scores seem to back up those claims (for the most part) with the Ryzen 5 2500U scoring slightly above 36% better single core performance and 48% better multi-core performance compared to the AMD A12-9800 APU with the latter being due primarily to the addition of SMT to the processor design allowing for twice the number of CPU threads (eight total). The performance improvements are also due to the move from Excavator to a Zen-based design on a smaller more power efficient process node. What is most impressive about this mobile part is that it is that much faster than a 65W quad core (4 core / 4 thread) desktop Bristol Ridge APU clocked at 3.8 GHz base and 4.2 GHz boost while using approximately half the power!
The Geekbench benchmark is only one data point, but is still a positive sign. One thing it does not reveal is clockspeed as while it lists 2.0 GHz that number is likely only the base and not the maximum boost frequency. Further, details on the Vega-based GPU are still unknown although the Infinity Fabric should help tremendously in reducing the bottleneck and keeping the on die GPU fed with data while gaming especially when paired with fast dual channel memory or HBM (I just hope that Ryzen Mobile is not held back like previous generation mobile APUs were with laptop manufacturers pairing them with single channel memory setups). We also do not know officially the number of stream processors that will be included in any of the Vega GPUs used in Ryzen Mobile with past rumors going up to 1024 SPs (mobile parts will likely be capped at 512 or 768 at the extreme). AMD claims that Ryzen Mobile will offer up to 40% more GPU performance, which to me suggests that we will possibly see higher GPU core counts but for the most part performance improvements are going to come from architecture improvements.
I am really interested to see how Raven Ridge plays out and hope that it is one step closer to finally realizing that HSA future AMD has been promsing me for years!
Also read: AMD Teases Ryzen Mobile APUs with Zen CPU Cores and On-Die Vega Graphics
Single-core is 3500 while
Single-core is 3500 while multi-core is 9500 and all this on 4 core, 8 thread CPU. What’s wrong?
So 2.5x versus 4x.
On other side Apple iPhone X has 4200/10000 with lower clock and not to mention other things.
yeah they also designed the
yeah they also designed the core and the GPU themselves to work in tandem with the OS they designed themselves.
the A11 Bionic chip is fast, no doubt, but people forget the extreme amount of vertical integration Apple does, and Optimisation can make or break a product
You do realise that you can’t
You do realise that you can’t compare an ARM core to X86. Goodluck trying to run a 64bit X86 application smoothly on the A11. Also, let’s not forget that the two CPUs have different instruction extension amongst many other ARCHITECTURAL DIFFERENCES…
Actually the more reading and
Actually the more reading and talking I do with devs, the more I believe that these scores are more comparable and relatable than we initially thought.
Its kinda hard to believe a
Its kinda hard to believe a ~5 watt phone soc is faster than say a 35-50 watt laptop cpu.
The laptop market will be
The laptop market will be damn lucky to get any Raven Ridge laptop APU SKU that can run above 15 watts by any laptop OEMs that cater to the wider laptop market. I’m hoping that AMD can get at least one Gaming focused Raven Ridge SKU that is above 35 Watts so as not to constrain the APU’s Integrated graphics for thermal headroom. Those thin and light laptop form factor designs are worthless for any heavy laptop as a desktop usage needed by the road warrior types that need a powerful laptop.
Then there is another big Question the Technology Press is ignoring about Raven Ridge and that is what about the Vega HBCC/HBC IP on any Raven Ridge APU that may not have any HBM2 for Vega’s HBCC to make use of as HBC. So AMD needs to be asked some questions about what will take the place of HBM2 on Raven Ridge to act as the HBC if these first Raven Ridge SKUs only have DDR4 DRAM to use as VRAM.
I’m Hoping that there will be at least some eDRAM on the processor die to act as HBC, But there appears to be very little information currently. AMD did mention that it was working on some Workstation Grade APU’s on an Interposer with “Greenland”(Probably Vega) graphics so I’m waiting for 2018 and maybe some new roadmaps to come out from AMD about the status of any Interposer Based Workstation grade APUs on and Interposer project status.
System76 the Linux Laptop OEM folks are taking their Laptop designs fully In-House so maybe they and other Linux Laptop OEMs need to be asked about any Raven Ridge plans. And AMD needs to be asked Directly about and AMD/Linux Laptop OEM partnerships where AMD can work with its Linux OEM Laptop partners in getting AMD’s Raven Ridge into that market that may see plenty of new users the closer it gets to 2020.
I mean, it’s called an ‘APU’
I mean, it’s called an ‘APU’ for a reason. The iPhone X’s GPU doesn’t even come close this the Ryzen’s.
Which “devs?” Are these app
Which “devs?” Are these app developers who build glorified repackaged mobile websites? I interface with devs on a continual basis. They and I affirm that ARM, especially Apple’s A series, are nowhere near Intel’s ULV Core processors in terms of raw, sustained performance.
Here is a far more accurate benchmark:
https://pbs.twimg.com/media/DJstTHbUIAAc0Iu.jpg
Here is why:
https://www.pcworld.com/article/3006268/tablets/tested-why-the-ipad-pro-really-isnt-as-fast-a-laptop.html
Geekbench is the proverbial kiddy pool of benchmarks for those who merely dabble in tech. I am fond of your site, but please do not discuss serious tech matters if you do not really understand the technical underpinnings.
They need to repeat such
They need to repeat such benchmarks with the A11; it is significantly more powerful than the A10 and A9. In addition to more powerful cores, it allows for all 6 cores to be active simultaneously. The older chips only had 2 cores active at a time. With more multithreaded software, that can make a big difference. Many of the older mobile intel parts are only dual core also, but with high base clocks. Intel recently released some new “8th genearation” 15 watt parts which are 4 core / 8 thread parts, but they will have lower base clocks to make up for it. Dual core will no longer be competitive.
If software optimization makes a big difference, then that must factor into buyers decisions. Apple achieves very good performance that users will actually see due to their high level of integration. The A11 processor has a much narrower design window. It is specifically designed for the power budget in Apples phones and tablets and may take advantage of many other characteristics, like the very limited amount of memory. That max memory in use with Apple’s processors is only 2 to 4 GB. When I look at the specs of AMD vs. Nvidia, The raw compute power should have AMD in the lead in some cases, but you cannot separate the hardware from the software optimization. At least part of the problem is that AMD doesn’t get as much software optimization or they get optimizations for Nvidia hardware that effectively deoptimizes for AMD’s hardware. Should we recommend that people buy AMD based on raw hardware specs? Obviously not. Software optimization is part of what people are buying. Even if you could judge these processors without taking that software optimization into account, what would be the use of that?
Anyway, I am of the opinion that most of these benchmarks are irrelevant to most phone or tablet buyers. I haven’t heard anyone complaining that the A10 in the iPhone 7 is too slow. Most of the things people complain about is because of the low amount of memory, not anything to do with the processor itself. Apple keeps the memory very low probably to consume less power, so developers just have to deal with the low memory sizes. As long as you have hardware video decoding, there really isn’t much other general use applications that need more performance. The iPad Pro and some of the laptop form factors are about the only place where performance benchmarks matter, since they may actually be used for photo editing or other such software. In those cases, you shouldn’t trust benchmark packages that attempt to cover everything. You should look for benchmarks of the applications that you are actually going to use.
What are your thoughts on the
What are your thoughts on the 2500u having a 12000~ score on android vs 9000~ on windows.
https://browser.geekbench.com/v4/cpu/search?utf8=%E2%9C%93&q=2500u
That’s true, but final
That’s true, but final customer does not care what is at lowest level. Only score matters!
So we, maybe, have a big problem in benchmark which uses same units of measurement for apples and bananas. And that’s wrong and misleading.
Comparing apples and bananas gives: 2 cores in A11 vs 4 in AMD. A11 has 4 low power ones which might be compared to 4 additional threads in AMD.
I browsed Geekbenchmark site and x86 is separated from phones. But, again same units – numbers.
For example Celeron 4 core 1GHz is 1430 and iPhone8 Android!? 4 core 1.3GHz ARM MT6580 is 425 – search for A11.
iPhone 7 A10 2 cores 2.3GHz is 3371 but also has 1800!?
If the benchmark is
If the benchmark is representative of the performance users will actually see, then I don’t see the issue with using it to judge performance. You cannot separate the hardware from the software infrastructure.
The Apple device has 6(2 high
The Apple device has 6(2 high performance, 4 lower performance) cores that can all be run simultaneously and the Apple A series cores since the A7 have execution resources very near a Broadwell level of x86 CPU micro-arch. So Apple’s latest shoud be up there on the CPU side and more needs to be Known About Apple’s in-house GPU micro-arch. There is also a lot of specilied DSP IP that can maybe also be made available for apps to nake use of for graphics enhancement so until the folks at the Chip IP evaluation services get their electron mciroscopes at that A11 there can be little said beyond any benchmarks that are run on the A11.
This also brings up the qusetion of Jim Keller’s/Teams othet AMD ptoject besides its Zen project in the Technology Press ignoring the fate of the Keller/Team custom ARMV8A ISA running K12 custom ARM project for AMD, that AMD had put off until 2018 and is not mentioned on any new AMD roadmaps.
I dont think we will be
I dont think we will be seeing K12 and the modular motherboards that can accept arm or x86 amd chips. From the lack of news it may be dead :-/. I wouldnt totally lose hope but dont hold your breath waiting at this point heh.
That’s nothing to do with
That’s nothing to do with project K12(Custom CPU micro-arch), that was AMD’s Project Skybridge that had the dual use motherboards Arm/x86. K12 was Jim Keller’s other CPU project to develop a custom ARMv8A running ISA custom micro-arch(Like Apples A7 is a completely Apple custom ARMv8A ISA running CPU micro-arch). So K12 was supposed to be AMD’s in house custom CPU micro-arch that was engineered to run the ARMv8A ISA for a design to replace AMD the Seattle ARM server chips besed on the A57(Arm Holdings reference cores) that Seattle(server chip) uses.
Go and watch the YouTube Interviews with Jim Keller where he talks about both his Zen x86 design Teams and his K12(Custon ARM Micro-arch) teams sharing ideas for cache subsystems and other similar subsystems ideas between the K12(Custom) Arm teams and theose Zen x86 teams.
I was hoping that Keller’s K12 custom ARM would offer SMT capabilities and AMD did announce that K12 was pushed back to 2018 some time but K12 is no longer on AMD’s roadmaps so(?) who knows!
AMD has to provide support to its ARM sever customers for at least the next 5-10 years based on advertised support for its ARM server product useng customers! So Lisa/AMD will have to at some point in time in AMD’s SEC 10K filings make the K12 projects status Known in the 2018 time frame. AMD’s current Opteron A1100 server SKUs are Arm Holdings refrence design A57 based and AMD will be support those desigs for at least 5-10 years and maybe updating those Opteron A1100 with some K12(custom in house AMD cores that run the ARMv8A ISA) design under a different branding(Epyc/Other) as AMD has retired the opteron branding.
We will probably see K12, but
We will probably see K12, but AMD has said they ditched the X86/ARM cross-compatible platform and so each will have their own platforms. They put K12 on hold to finish Zen, so I would imagine they are working on it now, but yeah we haven’t heard a peep about it in a while, so you never know.
That x86/ARM cross-compatible
That x86/ARM cross-compatible motherboard socket platform was called SkyBridge and was developed during the same time frame as AMD’s first ARM-based Server processor(Seattle), the Opteron A1100 with the A1100(Uses Arm Holdigs A57 reference design based cores), finally released around Jan 2016.
So AMD has it Opteron A1100(Arm A57 cores) server customers to support and K12(AMD Custom) if it did have SMT support would represent the first Custom ARM cores to provide such support. And Zen’s SMT is even more efficient than Intel’s SMT[marketed under the HyperThreading(TM) Branding]. With Zen gaining a little more performance with its SMT relative to the performance that Intel gains for its SMT[HyperThreading(TM)].
Keller ran 2 teams at the same time Zen one for x86 and one for K12(AMD’s in house custom ARMv8A ISA running design). Those Arm 57 reference design cores(Designed by ARM holdings) used in Seattle are a bit narrower in their superscalar design at 3 instruction issued per clock compared to Apple’s A7 Cyclone cores at 6 instructions issued per clock. So I was Hoping that any Keller designed custom ARMv8A running design would be just as much a wide order superscalar design as Kellers Zen design is a wide order superscalar design .
It isn’t really in AMD’s
It isn’t really in AMD’s interest to release a powerful ARM processor. It would open the server/workstation market to competition from any one who can throw some ARM cores on a die. While AMDs processor would probably perform better in the short term, they would have competition from very cheap, many core ARM processors. With the process tech stagnation, I wouldn’t be confident of AMD (or Intel) maintaining a meaningful performance lead going forward. Apple’s design probably isn’t too far off as is. Keeping the market AMD64 based protects them from cheap ARM competition for a while. ARM may still be coming eventually though, so having a high performance design could be useful eventually. Cheap x86 took over from RISC cores eventually, with lack of 64-bit support being the only thing that held them back for a while. I could see a similar thing happening with ARM.
Then why did AMD pay the top
Then why did AMD pay the top dollar to get a Top tier architectural license from Arm Holdings for the ARMv8A ISA so AMD could engineer their own custom micro-arch that is engineered to run that licensed ARMv8A ISA. AMD has its Seattle/Opteron A1100 customers that AMD has to support for the next 5+ years so why would AMD have Jim Keller/Team design a custom ARMv8A ISA running CPU Miro-Arch in the first place.
So the x86 ISA, even AMD’s implimation of the x86 ISA, takes more transistors to implement and uses more power than any RISC ISA based design. There are plenty of Arm/RISC instructions that decode on a one to one basis assembly language instruction to micro-code instruction but not so for the x86 CISC ISA.
Like x86 or not there is an ARM based server market out there that AMD is already a part of with its Opteron A1100 SKUs(Based on the Arm Holdings refrence design A57 cores) so why should AMD not make good use of the Custom ARM K12/custom Micro-Arch that may even Include some SMT capabilities that no other ARM/custom core on the market offers and the Keller/Team designed K12 may be more like Zen on the inside in other areas as well it’s only just that K12 was engineered to run the ARMv8A ISA and not the x86 ISA.
Look at the IBM power8/Power9 cores and earlier thay are RISC based and the Power8/9 cores are computing monsters even compared to the Xeon cores! So what’s this x86 fixation on a rather hacked together x86 ISA whose only reason for such marketshare in the first place is that IBM chose it for the first IBM PC designs way back when IBM created its PC(Personal Computer) that started the whole PC market in the first place. The x86 ISA is pretty ugly if you ask the computer architects out there but because the whole software ecosystem in so dependent on x86 folks have been forced to make us of the x86 ISA that they where stuck with.
But now even Microsoft is trying to get away from a total dependency on x86 and there are plenty of uses in the server market for the ARM server chips that can be more densely packed into smaller cabnets and do any workloads that do not require too much heavy FP workloads. There is even those new ARM SVE instructions that can do AVX types of workloads on values from 128-2048 bits and Fujitsu is building Japan’s first exascale supercomputer using the ARMv8A ISA and those new ARM SVE(scalable Vector Extentions) instructions.
Cheap x86 has always cost more than ARM, and that statement alone[“Cheap x86 took over from RISC cores eventually”] tell me that you do not know what you are talking about! Because The ARM market has so soundly beat even Intel’s Contra Revenue billions of dollars wasted attempts at trying to buy Intel’s way into the market where ARM comes in at fractions of any x86 CPUs BOM/Costs. Really Cheap x86 does not exist compared to the ARM CPUs that are even in the controllers in most PC/Laptops and most all consumer devices TVs to Toasters.
I have seen a lot of behind
I have seen a lot of behind the scenes decisions. Many decisions (not surprisingly) are not focused on producing the best solution for the customer, they are focused on producing the best outcome for the company in question. Intel has avoided producing more than 4 core chips for years even though the process tech made it economical a long time ago. Why didn’t they? Part of it is that producing 8 core mainstream parts would mean software will target many cores. That would help AMD who has offered more cores for the same price for a long time. It was much better for Intel to just sit back and make money off little 2 and 4 core parts for as long as possible. They also could charge very high prices for more than 4 core parts in the workstation/server market by keeping it that way. A single core with L3 slice is only around 12 square mm on 20 nm. That is easily small enough to allow 8 core mainstream parts. It allowed them to exploit their single thread performance lead for many years also. It is only ending now with the release of Ryzen. While AMD did purchase an ARM license, there is a reason we haven’t heard anything about it in a long time.
Apple’s chip is ARM RISC
Apple’s chip is ARM RISC (Reduced Instruction Set ..) and is really good at the particular tasks it is made for. HOWEVER this processor, i.e. R5-2500U is x86-64 CISC (Complex Instruction Set..). In simple words, you are comparing a speed boat (ARM) with an aircraft carrier (x86).
Excited to see how the Vega
Excited to see how the Vega GPU turns out. Desktop SKUs are looking like they were pushed far beyond comfortable efficiency to try and close the gap with Nvidias performance, curious to know where the mobile SKUs will stack up performance/power wise (will they still be more power hungry than the alternatives?).
Also curious to see where this item lands in the product stack, if Ryzen 5 is once again stacking up against i5s, that’s not boding to well as these scores are absolutely crushed by the new KabyLake-R i5s. Even the old 2C/4T i5U SKUs don’t seem to struggle too much, with the higher clocked i5-7360U matching up in both single threaded and multi core GeekBench scores.
R3 1300x beats these scores
R3 1300x beats these scores as well and amd has already announced 8core laptop parts. so that means the apus are where they always have been at the bottom of the stack competing with atom chips
Those 8 core parts are
Those 8 core parts are desktop parts(Ryzen 7 1700 based) used by some gaming Laptop OEMs in their offerings. So they will be clocked lower than if the CPU where in a desktop form factor and make use of Discrete GPU SKUs. I’ll be looking at those 8 Ryzen core laptop SKUs as soon as AMD gets some discrere mobile Vega variants to market that have at least 4GB of HBM2 with Vega’s HBCC/HBC IP able to use the HBM2 as the HBC and the Video Memory size extended out onto the regular system DRAM, or even system SSD paged memory.
I’m also waiting to see if AMD/portable workstation partners are goig to offer and Epyc/Portable Workstation variants as there are plenty of Xeon based portable workstation offerings on the market. So will AMD maybe be designing a specilized Epyc/Portable CPU, and motherboard, variant for the Portable workstation market.
What I really want is for AMD to get those Workstation Grade APUs on an Interposer designs to market with the Vega(On Navi) die/s and the Zen(Or Zen 2) cores die and HBM2 die stacks all on an interposer to market as that will be the best for any portable workstations in middle to late 2018.