Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
ARM A15 Samsung Exynos 5 can compete with Intel i3.
CalmOceans
Member UncommonPosts: 2,437
in Hardware
They tested a chromebook with an Ubuntu ARM distro on it.
This isn't even their 8x core, this is just a dual core, 8x is coming soon.
As you can see, it's closing in on an i3 from Intel but using very little power. It's pretty ridiculous how fast this is, this is a SoC using a fraction of the power of an i3 from Intel.
Intel better be very scared of ARM.
http://www.phoronix.com/scan.php?page=article&item=samsung_exynos5_dual&num=4
Comments
Can compete with? That's kind of like saying that an FX-4300 can compete with a Core i5-3570K. It loses by a comparable margin--and it's a margin that an awful lot of people would care about.
Also, that's a Clarkdale Core i3, which is nearly three years old. Cortex A15 cores aren't exactly going to be the new high end or even mid-range hardware in desktops or even laptops.
I'm not sure that Intel should be scared by how badly Atom gets crushed, either. Bobcat cores have been doing the same to Atom for nearly two years now, and Jaguar cores are coming soon.
The more interesting comparison is the gaping chasm in performance between Cortex A15 cores and the Cortex A9 cores in the Nvidia Tegra 3 shown there. Tegra 3 is a quad core, too, as compared to the dual core Cortex A15 being tested. That's why I've been saying to wait for Cortex A15 cores if you want ARM in a tablet (or AMD Temash if you want an x86 Windows 8 tablet) for quite some time now.
And they didn't even test graphics, either. The new generation of graphics for ARM processors (ARM Mali 6, PowerVR Rogue, Nvidia Tegra 4) will be the difference between "supports modern graphics" and "only supports really old, gimpy APIs".
Well you're missing the point somewhat, people aren't excited because of the pure benchmark.
It uses a fraction of the power and ARM uses a very shallow pipeline unlike the i3, they can push this SoC way farther than the i3 has life in it. 8x coming up.
And no they didn't use graphics, they're running an ARM version of Ubuntu, not much to test.
Oh well.
The ARM Cortex A15 has a 15-stage integer pipeline. Hence the name. Some instructions take more than 15 stages, even. I'm not sure what Clarkdale Core i3 is, but I think it's right around there. I remember that when Bulldozer introduced a 19 stage pipeline, people thought that was unusually long.
As for power consumption, yes, the Cortex A15 is pretty light on power as compared to a Clarkdale Core i3. But it burns a lot more power at load than a Cortex A9 core. That extra performance doesn't come free. Will it be more efficient than Jaguar cores (and remember, AMD is preparing a complete SoC that includes four Jaguar cores, not to mention graphics, in a 5.9 W TDP)? How about ULV Haswell cores, which have the advantage of being on a 22 nm process node? In both cases, the answer is maybe--but maybe not. The verdict on that is still out. ARM's big.LITTLE initiative of pairing Cortex A7 cores with A15 cores exists for a reason: the Cortex A15 loses some efficiency to get its performance.
Amazing how similar the Samsung Chromebook looks to a Macbook Air...
Anyway, stale patent jokes aside.
The latest Core i3 would either be the 3217U (1.8G 17W ULV) or 3110M (2.4G 35W mobile). A netbook would likely use the ULV part, but a more apples to apples comparison would use the mobile with it's similar TDP. The 330m is a 35W bin at 2.1G
One way I can really relate the 3 i3's is via Passmark; however, they only really test x86 CPUs:
330m (2.1G): 1818
3110M (2.4G): 2847
3217U (1.8G): 2072
So from that I can infer that the ULV bin is maybe just a bit faster than the old 330m, but that the regular mobile part is a good deal faster, nearly twice.
I can sorta relate the ARM to Intel via Geekbench scores:
Nexus 10 Exynos 5 (1.7G): 2370 (only package I could find with the same CPU/clock)
330m (2.1G): ~3600 (depending on the computer package)
3110M (2.4G): ~5300
3217U (1.8G): ~3700
So take all of that in, and the kicker is that the Exynos has a TDP of 4W. That is impressive - but I still don't think you can really say that it competes with an i3 in terms of performance alone. If we were to calculate Performance Per Watt, ARM would trounce it (but it nearly always has, it's just that it's never been competitive on performance alone). Now, that may very well be above the magical threshold of "Enough" performance for most people.
It will be interesting to see what comes out next from ARM, and how Haswell plays into all of this.
A sloth can compete with a cheetah in a 100 meter dash.
The sloth will use very little energy in competing.
It also will be far, far slower.
It can compete.
But it'll lose.
Lol...Sloth
Offtopic but lol
''/\/\'' Posted using Iphone bunni
( o.o)
(")(")
**This bunny was cloned from bunnies belonging to Gobla and is part of the Quizzical Fanclub and the The Marvelously Meowhead Fan Club**
Nonsense. It's a long way from "Broadwell won't have a socketed version" (which seems to be the source of the rumor) to "Intel isn't going to make desktop CPUs anymore."
Desktop CPU dies are typically also sold as server and/or laptop chips. AMD's Thuban (Phenom II X6) might well end up as the last chip targeted primarily at desktops without also being used in servers and/or laptops. Once you've got the chip ready for laptops and/or servers, it isn't much added cost to release a desktop version. The primary reason not to is if it's decidedly dumb to use it in a desktop, such as Westmere-EX (too expensive) or Tegra 3 (too low of performance).
If Broadwell's desktop performance wasn't going to be any better than Haswell, then there's no point in releasing a socketed desktop version of it. Intel could just keep making Haswell for desktops, while putting Broadwell in the markets where the reduced power consumption matters.
The launch of the first 14 nm chip won't mean that Intel has the fab capacity appropriate to go 14 nm only and shut down all older process nodes. Remember that Intel is going to push Airmont Atom onto 14 nm very early on in the process node's life cycle. Broadwell will probably be a huge deal for laptops because of the reduced power consumption. Servers are increasingly concerned with power consumption rather than absolute performance, though this varies wildly by the use of the server. You can't move everything to the new process node all at once, and if desktops get the least advantage from the die shrink, then desktop chips might be the lowest priority for using the new process node. Think of how AMD prioritized Trinity for laptops over desktops; that hardly meant AMD would abandon the desktop market.
Not that Quiz needs its but I offer this as my personal opinion behind why any company that has the ability to compete in the PC market would be an idiot not to.
Valve is a company that makes a LOT of money off consoles but it's pretty obvious the purpose of their latest project is specitically targeted at the home based personal computer AKA PC, and /r othersise known as desktop computers (although I would argue this is probably targeting more the HTPC with more GPU influence).
Software companies build their empires based on the hardware of the future. So to think that Intel, the current leader in desktop CPUs, are going to stop production anytime soon all the while companies like Valve continue to develop for them is flat out ludicrous! Next tinfoil hat theory please.