Slashdot Mirror
AMD's Fusion APU Pitted Against 21 Desktop CPUs
crookedvulture writes "When AMD unveiled the Bobcat CPU architecture behind its first Fusion APUs, the company claimed its Atom-killer would achieve 90% of the performance of mainstream desktop processors. But does it? This article compares the AMD E-350's performance to more than 20 desktop CPUs between $87 and $999 to find out, and the results aren't particularly encouraging. Although Fusion offers much better integrated graphics than Intel's latest Atom, neither stands much chance of keeping up with even low-end desktop CPUs. The E-350 does offer very low power consumption and impressive platform integration, making it a good choice for home-theater PCs and mobile systems. Desktop users are better off waiting for Llano, a Fusion iteration due out this spring."
Ok wait, so AMD's next-gen "atom-killer" successfully trumps Intel's next-gen Atom, but "the results aren't particularly encouraging" because it doesn't also beat full-fledged desktop processors? Seriously, talk about misleading.
In other news, iPods aren't the best at 3D graphics rendering, and cars are not the best choice for transatlantic shipping.
This is a test of CPU/GPU integration at the low end to start with - and a successful test at that.
Alphanos
I bought an Acer Aspire One 522 recently. It's a netbook with a 10.1" screen, 1280x720 resolution, and the new Fusion chip, so it has a Radeon 3250... I can actually run games on this device. I installed StarCraft II, dropped all the settings to minimal, and received playable framerates (after upgrading to 2GB ram). I blogged about it for those wanting more info. I need to make another post about Linux, because I have Ubuntu running near perfectly on it now.
I have no idea what business this new architecture has going against powerful desktop rigs, but for low-power applications, like a netbook, this offers a balance of computing power and energy consumption that's really nice, and beats what I've seen before.
Really most users today do not do much with there PCs but run a browser and email. It will run Office just fine and most software you would expect to find in most offices today. It should sell like hotcakes. Look how well the Atom does for so many tasks.
Yes if you are doing CAD, Gaming, editing video then this sucks.
For most other people it will be small, cheap, cool, and good enough.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
"Full-sized desktop processors are quite a bit faster than either the Atom D525 or the E-350.... Once you get past that realization, the next one follows almost immediately: the E-350 APU is the new champ in its weight class.... The E-350 Fusion APU is a wonder of integration, and AMD has set a new standard for basic computing platforms with the Brazos platform. For users whose needs are confined to simple office productivity, communications, and media consumption, the E-350 may well be sufficient. For those places where you might have considered an Atom- or Ion-based slim desktop system before, you'd now do better to consider a Brazos-based offering. In fact, we're left wondering how Intel could possibly continue its long-standing (and seemingly intentional) neglect of the graphics and video capabilities of this Atom platform for another generation. AMD has forced the issue. Without some help, the Atom will deserve to lose badly, both in "nettops" and netbooks, from here forward."
why wouldn't they just make desktop procs with it and leapfrog ahead again?
When's the last time you saw a Ferrari F-40 pulling a tractor trailer?
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Because you make a very different chip if you're aiming at 9 or 18W vs 60 or 80W.
Put all those desktop chips in the same power envelope as Bobcat, and they'd suck ass. Give Bobcat the power headroom of the desktop chips' environment, and it wouldn't know what to do with it.
The results as far as I can see are pretty good, given realistic expectations. Of course the article points out AMD claimed 90% of desktop, which just might be where some unrealistic expectations came from. Knowing AMD, that probably wasn't completely bullshit. It was probably a statement about IPC at equivalent frequencies, not delivered performance at their respective TDPs, possibly confused by a PR person, with a bullshit multiplier in there somewhere. ;)
The enemies of Democracy are
This quote is probably taken out of context as usual by Intel fanboy's. It was designed to compete with the Atom, and it beats it hands down. Not to mention AMD allows you to use OpenCL on their Fusion lines now and in the future. Im sure once people start accelerating their software with the on die GPU's it will speed things up quite a bit. Also, they could be releasing more newer fusion processors using Bobcat cores that will reach 90 percent. These processors were released specifically to compete with Atom, not reach the 90 percent goal the first iteration.
That brings me to an interesting point, / . is just "the ramblings of socially-inept, technology-literate news-mongers".
Isn't that what Llano should be. We'll see if it works when it comes.
I've been using the mediocre Intel IGP's for years on the last couple laptops. The GPU on these new AMD chips wipe the floor with the 2 year old Intel IGP on the laptop that I'm typing this on. Even basic home video editing doesn't really use the GPU, those goofy home videos are all CPU work.
Having the fastest computer doesn't mean much for most people. It's the form factor and utility that counts. Heck, we're one hop-skip-and-a-jump away from perfectly adequate ARM based machines that people will use instead of Intel or AMD... oh, wait... that was the iPad and it came out last year.
LOL, and to think that we used to measure computer speed by how fast it could recalculate a "large" Excel spreadsheet.
The PC market has rarely been defined by "good enough". PCs are sold not by what they can do, but by specifications, even to consumers who know next to nothing about those specifications. I'd guess 90% of PCs being used now could be replaced with something with half the computational power, and the user would not notice. The point I'm making is that technobabble sells, and if someone tells your layman that for a hundred bucks more they could double the processor speed, despite that having little to no effect on real world performance, they'll go for it.
Personally, I'm a gamer, so this is all a little irrelevant to me.
Grammar Nazis' --- I'm assuming that was deliberate.
A fair point that TDP isn't the whole picture, and I didn't say anything about die area ergo cost as a distinguishing feature of the market Bobcat and Atom are going for. I'd believe it does do better with a bigger core and lots of die space spent on 3.5MB of cache. It would have been interesting to see it in this comparison.
Though the conclusion would have been the same, as clearly a $250 part is not competing in the same space.
BTW, you seriously need to stop projecting.
The enemies of Democracy are
EG, Pentinum 2ghz compared to atom D525 or amd E250
Not sure about current Atoms, but the 330 dual-core benchmarked about the same as a 3GHz P4 when running multithreaded apps (and quite a bit slower when single-threaded).
Or, if you prefer big iron, if I remember correctly it gets about the same as a Cray Y-MP.
Transmeta did your idea. They went out of business.
The x86 and x86_64 carry around a lot of transistor baggage needed to be maintain 4 decades worth of backwards compatibility
Not really. More and more of the CPU is cache, so the size of the instruction decoders becomes less important all the time. Plus I believe I read that Intel can now turn off the instruction decoder when it's not required (e.g. running tight loops from the micro-ops cache) so that reduces the power usage too.
Why do giants like Intel and AMD continue throwing money into improving what will always suck?
Because they want to run Windows apps.
Power7 is quite efficient, but also costs thousands of dollars. There are some embedded-market PPC chips that do okay though - you can get performance comparable with a low-end/midrange Nehalem, but the downside is that they cost far more than they should ($300+) and have hardly any software.
ARM sucks. Go look up its SPEC2000 results and see.
I've been waiting for a good low-power CPU to come along that'll play 1080p and fit in an ITX rig, and it looks like this could be the one. It'll be nice to run Boxee on a computer that actually fits in my TV cabinet.
"AMD's performance target for Bobcat was 90% of the performance of K8 at the same clock speed"
Note the key clause *at the same clock speed*. To turn around and complain that a 1.6 GHz Bobcat isn't 90% of the performance of any of a set of desktop processors running anywhere from 2.7 to 3.6 GHz just shows that you didn't bother to understand the initial claim.
Very low power consumption, decent processor throughput and decent integrated GPU performance. For me its really the first and third that matter most so I am definitely in the market for designs based on this architecture.
Have you got your LWN subscription yet?
You pay a premium for tiny. I remember when I got my Nomad Jukebox 3, the iPod of that generation was selling for a couple hundred dollars more and had less disk space, but it was a fraction of the size. People are willing to pay a lot of money for something that's tiny, whether or not that's the wisest course of action.
Umm.... It is an X86. What the heck are you talking about software support? Put Windows 7 or Linux on it and use what every you want?
And it is a lot better than the cheapest Atom and makes a lot less heat and uses a lot less power than chips above the Atom!
WHAT ARE YOU TALKING ABOUT?
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
high performance ?
The Cloud - because you don't care if your apps and data are up in the air.
http://pcper.com/images/reviews/608/02.jpg
Apparently, the cores are about 1/2 of the die (the rest being cache and interfaces). I'm not sure how much of the cores would be trimmed down by getting rid of the x86 compatibility layer (both Intel and AMD have been using non-native x86 processing cores for a while, x86 machine code is actually converted to different micro-ops). Assuming you're just talking bout moving x86 compatibility out of the hardware and off to software, while keeping the same hardware unist (SSE-x...), my guess would be 1/3 of the cores tops, so 1/6th of the whole chip. Not much, probably not worth the effort.
The Cloud - because you don't care if your apps and data are up in the air.
"not much but a browser" is an outdated statement, since that now includes watching video and playing games - even news sites' main pages weigh in the megabytes. The Linux Flash player sucks up a lot of CPU for some reason.
I have an old 800 MHz computer that really isn't useful for browsing the web any more, although it used to be fine.
I've seen nothing on Slashdot about Nufront's 2GHZ dual core A9 chip. I know it's early days for that, but I would have liked to see that in any benchmarks which can be done on non-windows boxes. I suspect that at 2 watts (1/7th the power -- at least for the chip itself) the Nufront demo boxes are quite competitive with the Atom and the AMD Fusion systems. But I need to see the benchmarks to see my suspicions confirmed or stomped.
What? If "good enough" wasn't what the PC was sold on, "IBM Compatibles" would have never gotten off the ground. The PC has lived off of "good enough" most of it's life. It didn't become the best option personal computer wise until about 1995, and by 2005, they had reached the point that people stopped caring about the speed so much. So, they only sold on performance for about 10 years.
Not so much. x86 may have a challenger for the desktop, but it looks completely unapproachable in the server space. X86-64 took off in the server space. Sun was selling Opterons with Linux. The top500 is packed with x86-64s. Etc.
The original 8088 instruction set is crap, but it has improved since. X64 is perfectly fine. And make no mistake, many x64 chips are running 32 bit code, because an in-house app that cost many millions to develop simply isn't worth rewriting, and the overhead of emulation is undesirable, and a negative selling point.
But more important than 8088 instructions, is RISC versus CISC. Time has shown that CISC chips like x86 just have so much room for expansion that they can easily maintain backwards compatibilty for decades, while adding every new feature under the sun, no problem. RISC chips weren't so lucky, and their former benefits, became a liability. And losing backwards compatibilty wouldn't affect Windows users that much, but corporate servers... it would be a blood-bath.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
And this chip has flash and H.264 acceleration features. Just run a browser and email.
Yes a modern browser can do a lot but it still isn't a cad system.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
MIPS (the architecture) did this a long time ago. Nowadays everyone is focused on ARM, but ARM isn't the only instruction set that can be low power. The company behind MIPS is still pushing it, and there are even some Android phones out with MIPS after they ported Dalvik to MIPS instruction set. MIPS also has an advantage in that the 64-bit version is already defined.
I miss my Nomad Jukebox it was the best MP3 player I ever owned for the money.
The results as far as I can see are pretty good, given realistic expectations.
It seems to me the real problem is that they stuffed a discrete GPU into the thing and then tried to compare it with desktop systems with discrete GPUs. That isn't the usage model for these at all. These only have 4 PCI-e lanes. Look at page 16 where they start to compare it with desktop systems where they're all using integrated graphics -- it comes out far closer.
And then you throw in the fact that these are dual-core processors being compared to desktop processors up to six cores. The ones with more cores get much higher scores on the benchmarks that are well-threaded. Who'da thunk it.
The idea that these are 90% as fast as modern desktop processors at all tasks is clearly false. But 90% as fast as desktop processors with integrated graphics, for a large subset of common tasks? It's not that implausible.
I've still got mine, I just need to get new batteries.
This isn't an ARM chip.
(I personally think AMD should be picking up ARM designs, to help us shake off the burden of x86, but that's not what this is about.)
Computer memory is just fancy paper, CPUs just fancy pens with fancy erasers; the 'net is just a fancy backyard fence.
Hi MR foaming at the mouth fanboy! you DO know that the part you are naming usually sells for something like 400% MORE than the part being discussed, yes? Your comparison is like saying "Well my Ferrari will beat your Camaro!" which is obviously true but doesn't take into account that one is MUCH more expensive than the other.
And considering the main stated goal for this chip is in ULV netbooks and nettops where Atom is currently being used I don't think it is a fair comparison, do you? I mean last I checked those selling Atom netbooks didn't have a "Drop in a Core i5" option at the checkout.
Now looking at this logically from the chips stated purpose and not from any "AMD yay, Intel boo!" position and its intended opposition (the Intel Atom) it seems to fit well into the category it was designed for. It uses a little more power but at the same time the addition of a Radeon GPU means that for multimedia, a task that pretty much everyone enjoys, the much higher performance should more than offset the slightly higher battery use.
ACs don't waste your time replying, your posts are never seen by me.
The original 8088 instruction set is crap, but it has improved since.
I've ranted on the theme of 8086 evolution and adequacy several times in the past. You do know that the 8086 instruction set was designed in 1976-1978? I recall 1976. Year of the first cheap four-function calculator, the TI-30. It was also the year of the Summer Olympics in Montreal, where Canada as host country failed to win a single gold medal (courtesy I figured out later of East German steroids). The 1976 summer Olympics also featured the original decimal conversion bug, when Nadia was awarded a 1.0 gymnastics score due to a shortage of display digits. Intel engineers took notes and later conducted a replay attack.
35 years later, the planet's deployed x86 execution capacity is on the order of 10^18 instructions per second. Failures don't come any bigger than that.
Would the course of history differ much if someone teleported back to 1976 and beat the Intel engineers responsible with the clue stick of winsome orthogonality?
I would settle for one change only: an eight bit segment register offset instead of the four bit offset, giving the original PC a 16MB address space. 16MB has you covered until machines become powerful enough for a multitasking OS. It also has the PC encroaching on traditional IBM markets right out of the starting gate, and might have died a grim death of a thousand memos before IBM publicly announced the product.
And then, if it hadn't been so crippled, maybe fewer people would have rushed into the marketplace to fix it. Maybe Apple would have controlled the original PC marketplace instead, and locked the whole thing down with pentalobular vengeance.
This first fusion chip underwhelms me, but it's the tip of a program with bigger splashes to come. Since I'm older than dirt, I also recall that the original Pentium 60 was nearly booed out of the marketplace. Hotter and more expensive than a DX4, without any real benefit to show for it unless you had a fetish for fast (and wrong) floating point math. There were howls of outrage at the unfathomable 30W TDP (IIRC, the figure doesn't come up in a quick Google). You needed a heat sink the size of your hand and maybe *gasp* even a fan.
Superscalar, DOA. Fusion, DOA. Perhaps let's not jump to conclusions yet. The entire industry is built on the ashes of suckhood.
Want to bet? Take a lot at sales number. i3s outsell i7s for the simple reason they are good enough and cheaper.
I think you would surprised just how few people will take that upgrade. If you are looking at say a $299 notebook that works well vs a $399 notebook I think a lot fewer people will get the upgrade.
Now if you are talking about an 1199 vs 1299 notebook you are correct.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
I don't know if it's really a problem. I mean, it's interesting to see how well they perform against a broader range of chips, and how close they can get to desktop performance. A lot of people were hoping pretty close and it turns out, no not really. Not exactly a surprise, but still good information. And they admit it's unfair, and basically give it a positive review in their conclusions where they say that for what it wanted to achieve -- beating Atom, particularly in media playback -- it worked.
That it does well against integrated solutions -- pretty much a given in the cost structure it's targeted at -- is also a good sign.
The funny thing is that someone further down the thread sourced the original claim from AMD, and it was pretty close to what I'd guessed -- 90% the performance of a K8 at the same frequency. Which is a much more reasonable statement.
The enemies of Democracy are
Sure there are tons of people out there that insists on buying computers that waste so much power they could cook meals for a couple dozen people with the heat dissipated by their CPUs alone. They have a massive ego that requires the notion they have the fastest computer possible, even though their computer will run at less than 10% utilization almost all the time. Game vendors keep writing code evermore inefficient, same for Microsoft OSes. Some actually do need a fast computer, perhaps for encoding HD video, performing very complex calculations or something else.
Since I migrated from Windows to Linux about 10 years ago, freeing myself from 99% of bloatware, I found out that CPU speed is becoming more and more irrelevant with each new CPU generation. Even with a 4 year old computer, I do simultaneously share a 10Mbps broadband connection, burn a DVD, play flight gear flight simulator, run e-mail server, MySQL, on the same 4GB RAM system with a dual core CPU, no damaged DVD medias, even with the game running at maximum CPU priority. Try that on Windows, any Windows !
People, save on your power bill, free yourself from wasteful computing, migrate to Linux. It will run exceptionally well on any new computer, it will also run very well on anything designed for Windows Vista. The minimum computer that Windows 7 requires just to boot and load Office is a huge computer for Linux.
Migrate to Linux. Make your computer work for you instead of to evil software corporations. The computer is *yours* after all.