AMD Athlon 5350 APU On Linux

[curaga]

Yes it will an possibly substantially. In fact this is the only thing that bothers me about this early testing as you really need drivers that fully enable the chip. What we have right now is a biased set of results. Further I want to see testing with both DIMM slots filled with RAM. This is especially important for any GPU testing.

The AM1 socket is single-channel only?

[Spittie]

Why would anyone buy this junk? Crap performance and it needs a big heatsink and fan.

Because not everyone has >$200 to spend on a cpu, I guess?
The 5350 is probably fast enough for the people that only need a computer for facebook/twitter and similar stuff. IE, most of the world.
I agree on the fan, but since the chip TDP is 25W, it's not hard to find a fanless cooler (I wouldn't be so surprised if you could just remove the fan, and just use the stock cooler).

Meanwhile, this: http://hardkernel.com/main/products/...=G138745696275 costs $65 (or $59 if you buy just one: http://hardkernel.com/main/products/prdt_info.php ) and uses a 5VDC/2A power adapter... I guess AMD keeps taking the wrong path.

Yes, and it's much, much slower.

Yes it will an possibly substantially. In fact this is the only thing that bothers me about this early testing as you really need drivers that fully enable the chip. What we have right now is a biased set of results. Further I want to see testing with both DIMM slots filled with RAM. This is especially important for any GPU testing.

Kabini is single-channel, so I don't really see this helping here.

[blackiwid]

The AM1 socket is single-channel only?

I dont think it has something to do with the socket its about the cpu (if I am wrong tell me) the mem-controller is in the cpu/apu so it should be a problem of that.

To the cpu, I really want to give out money, and I would love to get a real update to my E-350 (Zacate), but if this cpu is really nearly not faster at all, it makes no sense.

Even if that is only because the benchmarks do not well show mutli-core performance or some of them, but if 1 core is only 10% faster its just not enough to make a big difference.

I think I have to wait for beema, and because beemo will come anyway this year it nearly makes no sense to buy it anyway even if it would be faster. But with that, I can just only see than that beema makes a big jump or amd is just giving up cpu market at all.

I can?t still belive that its a 1-1 Zacate with double cores after several years unbelivable.

[smitty3268]

That's the state of things with Ubuntu 14.04 out-of-the-box as mentioned multiple times in that article. As also said in the article, more tests are forthcoming.

Can you check to make sure the Jaguar cores are getting clocked to the proper speeds when running tests?

[Spittie]

I dont think it has something to do with the socket its about the cpu (if I am wrong tell me) the mem-controller is in the cpu/apu so it should be a problem of that.

It's a problem of both. Yes, the controller is in the cpu, but you also need the right pins and support in the motherboard.
Anyway, it's not a problem here, since the controller in the 5350 is single channel

I can?t still belive that its a 1-1 Zacate with double cores after several years unbelivable.

It does look pretty good in the other (few) reviews that I've seen: http://www.modders-inc.com/amd-athlo...-apu-review/3/ and http://www.anandtech.com/show/7933/t...lon-5350-am1/3 (only the x264 benchmark include the e-350)
I'm not sure why Michael's benchmark show so little difference.

[Luke]

HD Video benchmarks should be for local files

Problem with that is lack of free, well-mirrored content in all the different bitrates, etc. If there is, feel free to add it to PTS and it will then be tested.

I've never seen or use an internet connection capable of streaming 1080p video in realtime. Doubtless many others have not either. Perhaps this is a request not for Internet video playback benchmarks, but local video benchmarks?

A high bitrate HD video should be considered a file useful only for local playback after creating or downloading it. As such, I usually benchmark video by taking a selection of my own videos in various sizes and seeing whether a CPU can play 1080p without stuttering, if not try 720p, etc. They are played from the desktop in most cases. When the CPU is at its limits, closing all browser windows and running only mplayer/mpv really helps!

[Michael]

I've never seen or use an internet connection capable of streaming 1080p video in realtime. Doubtless many others have not either. Perhaps this is a request not for Internet video playback benchmarks, but local video benchmarks?

A high bitrate HD video should be considered a file useful only for local playback after creating or downloading it. As such, I usually benchmark video by taking a selection of my own videos in various sizes and seeing whether a CPU can play 1080p without stuttering, if not try 720p, etc. They are played from the desktop in most cases. When the CPU is at its limits, closing all browser windows and running only mplayer/mpv really helps!

It's for local playback but it's a matter of for me to test it needs to meet PTS requirements, which includes reproducibility for tests that can be done by anyone.. Thus needs to be publicly hosted somewhere.

[brent]

To the cpu, I really want to give out money, and I would love to get a real update to my E-350 (Zacate), but if this cpu is really nearly not faster at all, it makes no sense.

It is much faster compared to an E-350. There is definitely something very wrong with these benchmarks! See basically all the other Kabini benchmarks, which can be found on the web. The newer Jaguar CPU cores in Kabini have approximately 20% more IPC and run at 30% higher clock, and you get twice the cores. It is a huge improvement over Brazos, on all fronts.

[profoundWHALE]

Meanwhile, this: http://hardkernel.com/main/products/...=G138745696275 costs $65 (or $59 if you buy just one: http://hardkernel.com/main/products/prdt_info.php ) and uses a 5VDC/2A power adapter... I guess AMD keeps taking the wrong path.

Hey, genius, that's an ARM processor, which means I would need ARM compatible programs. So if I wanted to play a game...

It's for local playback but it's a matter of for me to test it needs to meet PTS requirements, which includes reproducibility for tests that can be done by anyone.. Thus needs to be publicly hosted somewhere.

I believe blender or Xiph could provide you with that, there's plenty of sample videos that they have hosted on Xiph's website. Some of them from the Blender Foundation.

[RussianNeuroMancer]

Meanwhile, this: http://hardkernel.com/main/products/...=G138745696275 costs $65 (or $59 if you buy just one: http://hardkernel.com/main/products/prdt_info.php ) and uses a 5VDC/2A power adapter... I guess AMD keeps taking the wrong path.

Maybe you'll find something here: http://wwwd.amd.com/catalog/salescat.nsf/shop?openform.