Any idea why these Samsung results are so much better than the nearly identical QCOM part?
Even if both firms have used the exact same stock ARMH core, you can still get a wide variance in performance due to other factors, such as cashes and the memory controller etc . Then you add tweaks that both firms could do to the design, and ultimately they are on different processes. This would also be Samsung's second gen of this CPU cluster on 20nm.
That said, people on other forums have used the difference between the QCOM part and the previous Exynos to argue that the 810 does have heat and performance issues.
I dont know:)
Wally, as you love rumours, It's been suggested that this Exynos 7420 is actually a 14nm FinFet part. Wouldn't that be a hoot?
The new owner will have to renegotiate the cross license and terms with Intel for the x86 and access to Intel patents that the part requires.
Wouldn't that go the other way as well? As AMD 'owns' the 64bit extensions?
It's the season for benchmarks of next generation devices to leak into benchmark sites. Here we have the Exynos 7420 and a Cherryview tablet.
These aren't production devices, so apply a big dollop of scepticism to the results.
Within one year you saw 20 to 28% of the tablet market which was ARM fall to Intel. Intel forecasted 40 million tablets but grabbed a bit more. True they lost money on each sale, but with each passing month it was a bit less. Sooner than you think Intel won't have to pay the difference between it and an ARM based SoC/SiP.
What Intel got was credibility. now developers are starting to develop for Intel based devices running ARM (answer to your question "Why bother?" is because Android based Intel devices is where the market is heading).
Something that should keep you up at night was Intel more or less said 40 million tablets last year. It wasn't a goal, it was a limit. They could have picked up even more market share. When Intel can sell it's SoC's for the going rate in the market and turn a profit how much more ARM market share can they grab?
40 million tablets eh? Is that Intel's definition of a tablet, or someone else? That's an important distinction...
Can you name any tier 1 android tablets that Intel won with this programme? You see, the bulk of the Andriod shipments went to no name Chinese box shifters, with only a few devices flowing into western markets (where alot of software vendors are based). How many of these box shifters will remain loyal to Intel once the rebates and other 'help' goes away?
[No, a sick man is a man who has a fixation on another man on the internet and follows him everywhere and spends all day long trying to glean every bit of information he can about him. That's pathological. Get some mental health care. Your fixation on me is truly creepy. Stop using this message board as therapy and get some professional help.]
Perhaps, it's more simple than that? Perhaps he wants a date?
More a lack of financial commitment by anyone in the WinRT ecosystem to spend the money to make legacy Windows software run on ARM chips even by emulation the way Intel has with Android on x86.
Tut tut Marsavian, stepping into the Weaver land with that one.
WinRT was an attempt by MS to create a walled garden, an Applesk control that could drive profits via the store. The only way to run 'real' Windows applications in RT (other than MS office) was to jail break the device - which would break the warranty...MS, never supported 'legacy' applications, other than it's own, with RT. You should know that both OEM's and software firms don't go against MS directives (which is why you find very few dual boot (Andriod, Windows) devices supported out of the box).
Do not take this as a defence of RT. It was ill conceived, and should die ASAP.
[It doesn't change the "it's meaningless" conclusion. ]
You must be right. I was thinking that Amazon wouldn't drop 300 million on a start up, spend at least the same amount again on new designs, just to say "oops", how are we going to manufacture them? Especially, as they couldn't have just gone to Intel in the first place right?
[Yet more churlishness and meaningless generalizations from MonkBoy meaning nothing. More and more you are like the little dog that runs out from underneath the porch and tries to bite people on the ankles. Your posts are all form over substance these days. Little one liner gotcha stings that apparently bring you comfort as ARM struggles with its technology, manufacturing, fabrication and capacity.
Where's the meat, MonkBoy?]
I am not sure how you want me to reply? I attempted to answer the posters question (Apples move to 64bit), it's a simple as that. If you think my answer is wrong, just say so.
[Amazon has been previously tied into ARM designs with its tablet line. A continuation is no big surprise. And like you say it's fabless so that can change in the future. And you really want me to get excited about a mere 300 million investment? No way. Now if it were a $12 billion investment like a fab then that would bring me pause.
No, MonkBoy, this is relatively meaningless. In the meantime the ARM fabrication crisis continues unabated.]
Dear, Britney "oops i did it again" Weaver, this isn't phone, or tablet, it's for AWS (their cloud environment).
Wallis, you need to get on the phone to Amazon AWS HQ, as they are just about to drop 300 million on an fabless startup that uses ARM CPU's in it's designs. They clearly don't know about all these fab problems...
The best design doesn't always win. They don't.
So why will Intel gain share then?
"if the CPU benchmark scores are legitimate," If you're suspect of the results why should I trust them?
You can't trust the scores from any SoC/CPU that isn't in shipping devices, I would give the same proviso for ARM based SoC's.
Is this a business plan or simply a talking point for ARM? To me it sounds like a talking point. Here's what a business plan sounds like.
I pointed this out to show how 'quick' the market responded to Apple moving to V8. Do what you will with that information, it's nothing more than that:)
Will Qualcom, Nvidia, Microtek or any other supplier be so bold? No, they are in bed with A&B tier handset suppliers. Will they help pay for ads?
You've been talking about this for years, but it just hasn't worked...yet. Do you know that the reason why Xiaomi (big in China) can produce new designs so quickly? It just takes the reference designs from it's suppliers...
I thought about your question and here's the best way to answer it. Where ARM blew it happened over a year ago. The iPhone had current technology, but not always the latest technology. That changed in the 5s when it went to 28nm 64 bit... Intel 14nm will land against Apple's A8. For argument sake we'll say it's behind, but not far behind in benchmarks.
The quick move to V8 of the ISA by Apple certainly did throw roadmaps to the wall (esp QCOM), but 50% of phones this year will be able to run that ISA.
Lets make it clear cut, lets say Intel's 14nm chips are better performing than the A8. Intel will still struggle to get tier 1 design wins. Why? Intel's potential parters have no faith in the roadmap, it's going to take time to win that trust back (this point can't be emphasized enough). It's also worth remembering, Intel isn't competing with AMD like companies...
This is the C-tier handset manufactures eating the market share away from ARM by offering premium cell phone technology at the $200 price point. The fight will be with ARM at 20nm. How will ARM 28nm do against 14nm from Intel? Demand for 28nm and older nodes will die off as orders fall off. This concept is called 'Inferior technology." Intel did it with servers and will do it with phones. It's not top down, it's bottom up market penetration.
ARM is talking about $75 phones with LTE this year...
As for ARM 28nm against Intel 14nm, well, if the CPU benchmark scores are legitimate, then Cherrytrail (for tablets) is going to struggle to find clear water against these 'Inferior technology' 28nm parts. It clearly struggles against existing 20nm parts in the market today, let alone the parts coming later this year.
ARM isn't dead at 20nm. But they will start to face real competition from Intel starting now going into full pain mode within a year.
The fabs that can afford to push 20nm aren't the ones to worry, for now. The backlog will occur at 28nm and older.
[An Intel fan? Ha, then you must internally hemorrhage as you post your numerous bashes. I have been largely accurate with my predictions on ARM's fabrication. Please feel free to offer an in-depth analysis where you believe that I have been wrong instead of support-free rubbish comments. ]
When I have bashed Intel? I have not.
You have been an absolute *failure* in predicting ARM's fabrication failure from since the 40/45nm node. If you can't see that, you are truly blind. It was only a few months ago that you predicted that we wouldn't see 20nm and here were are *hundreds of millions* of 20nm SoC's later (from 2 foundries)...
[As I commented recently it's the end of the ARM dominance era, which is absolutely true. If you believe that ARM still has dominance in performance, energy efficiency, price and capacity please offer some type of support because there is a huge body of evidence that this is not the truth.]
Look, let me make this crystal clear. The ARM foundaries have *never* have the performance, power, price over Intel that you describe above. What the foundaries and the broader ecosystem have (over Intel) is the ability to bring new designs to the market quickly whilst continuing to improve the process and design for that node.
Intel has the best transistors. Intel has the best fabrication technology. I doubt anyone would argue this when speaking so broadly. However, this means little when you don't have the the right products in the market at the right time. Surely, even you can see that? Is that bashing?
Or are you confusing the design of SoC's V fabrication technology?
We can't even ask Nenni for his take because he went back to school to get a degree. And AE has fled all public forums. The ARM fanbois are down to the following three stooges: getanid61, MonkBoy and Twink.
I think you'll find that I am a bit of an Intel fan as well. I will continue to challenge the rubbish that you produce. So if that makes be a stooge in your eyes, then so be it:)
I must admit, I do smile when you type these "The End of the ARM Era" posts. You've been at this since the the 40/45 node. Begrudgingly, I can't help admire your ability to stick to something that has been proved to be wrong for so long.
Always the optimist eh?:)
who else is using (successfully) big.LITTLE
i am sure the bluered monk will come up with a long list
Actually, I only know of 1 other A57/A53 20nm design on 20nm and that's the Exynos 5433 in the Note 4. Sweepr over on the Anantech forms has done quite a bit of testing with the device. For what it's worth, it doesn't have the issues that are rumoured for the 810.
"I know, but it will be here soon enough:)"
[And soon enough Intel will have something better.]
Better than Red Hat? Are you suggesting that Intel is going to launch it's own Linux distro?
The ARM server threat is entirely negligible, any IO/memory benefits that you think Seattle/X-gene has over Avoton will be countered by 14nm Denverton which Intel can price low enough to extinguish any nascent ARM threat. As for Power 8 you really should look up its enormous bandwidth/cache/memory numbers, it really operates on another planet to your toy ARM chips.
I have used Power, so yes I know:) Again, you are focusing on raw CPU performance and that's not the problem that the ARMy 'toys' are trying to solve. Negligible is not zero, which is the point I was trying to make to the author of the other thread. If the threat was none existent, we wouldn't have seen the moves Intel has made (and continues to make).
Nope, you are wrong about IO/memory (14nm by itself means nothing in this context). Once you have saturated the io channel there is little you can do about it (the common solution is to add another server). Even the lowest of ARM cores can saturate it's network port, never mind Xeon.
p.s. Only Suse and Ubuntu ARMv8 releases are official, Red Hat which is 2/3 of the linux o/s market is still a WIP
I know, but it will be here soon enough:)
So let's recap, what is currently supported is LAMP...
New thread, couldn't reply to the old one.
I think you may be confused to what 'support' means. Basically all linux distro's 'support' the ARM ISA in one form or another, what's so special about Ubuntu LTE and the Red hat release is that you can get guaranteed support for x years (typical security fixes etc). This is a big deal to some firms.
Even if the distro just contained apache, mysql, perl/php/python we are talking of ten's of millions of lines of code in thousands of modules that just works out of the box. Rather silly example, 15% of websites run Wordpress (Apache, mysql and PHP). Of Course, the distro's don't just contain AMP...
That said, one of the projects I am involved with uses MongoDB and Node.js, both can be built for ARM but neither is officially supported by the tool vendor which would stop many firms using those products for production (on an ARM server).
Performance with the Power 8 - and I can't stress this enough - it's not just about raw CPU performance in servers. While it's certainly true that you have CPU bounded workloads it's a fact of server deployment that you are io/memory bound for many workloads. If you look at X-Gene and Seattle they are more integrated than Avoton or Rangeley and contain faster/more io and can use more memory. The question becomes: are Seattle/X-Gene better balanced (CPU V io) than Avoton/Low end Xeon for their target workloads and price points (including BOM and TCO)? Then you have to throw in to the mix less traditional server SoC such as ThunderX.
Intel segments it's chips, and this "balanced" argument is really about this segmentation. Intel can (and has) changed this to meet the ARMy head on.
So is it all singing, dancing and joy in the ARMy server land? No, far from it, but the idea that the threat to Intel is zero and doesn't exist is just fantasy.