AMD "Kaveri" APU Linux Review, Benchmarks Next Week

[pinguinpc]

Oh no, competence! We can't have any of that!

Maybe you wanted to say "competition"

Yes for this case maybe rival, i try correct but dont let me

[zoomblab]

A7850 looks an ideal for a steam machine. Am I right? This will have the performance of a ps4 without the need for dedicated card, which would mean lower price and less cooling requirements / smaller enclosing.

[pinguinpc]

A7850 looks an ideal for a steam machine. Am I right? This will have the performance of a ps4 without the need for dedicated card, which would mean lower price and less cooling requirements / smaller enclosing.

if you refer to AMD A 7850, GPU on this APU is lower than PS4 (PS4 have Radeon HD 7870 similar performance) but mainproblem for AMD, especially on GCN GPUs is Driver Support on linux (Low Compatibility and Performance compared to NVIDIA)

[guido12]

Do you do any LZMA2 compression and decompression benchmarks? I've been using Pixz ( https://github.com/vasi/pixz ) lately because it supports multi-threading for both compression and decompression. I apologize if these are already in the tests. I tend to gloss over the hardware benchmarks since I have no idea what most of them are and the articles often don't provide any description. A simple "this is a single-threaded benchmark that makes heavy use of integer arithmetic and relies heavily on memory bandwidth" blurb would be nice for each test.

Will there be any lower TDP Kaveri APUs? I'm planning to slap together a small mini-ITX system and want to get the most powerful CPU that could practically fit and not cost too much. About TDPs, have AMD's and Intel's measuring of TDP changed? The last thing I've read was an Intel whitepaper several years ago. Intel measures TDP by loading the system with what they consider a "typical" high load application while AMD maxes out the load using what I think was called power virus type loads.

[Michael]

Do you do any LZMA2 compression and decompression benchmarks? I've been using Pixz ( https://github.com/vasi/pixz ) lately because it supports multi-threading for both compression and decompression. I apologize if these are already in the tests. I tend to gloss over the hardware benchmarks since I have no idea what most of them are and the articles often don't provide any description. A simple "this is a single-threaded benchmark that makes heavy use of integer arithmetic and relies heavily on memory bandwidth" blurb would be nice for each test.

There are several compression test profiles supported (check out OpenBenchmarking.org for details) but not Pixz. I'll take a look at pixz and see if it's worthwhile to include in PTS benchmarks. For tests you don't understand, there's descriptions of each and every one at OpenBenchmarking.org -- from the Phoronix reviews/articles pages, usually just a matter of clicking on the title of the graph (for the SVG versions in modern browsers) to be brought to the page with plenty of details. Other feedback for improved usability is welcome.

[guido12]

There are several compression test profiles supported (check out OpenBenchmarking.org for details) but not Pixz. I'll take a look at pixz and see if it's worthwhile to include in PTS benchmarks. For tests you don't understand, there's descriptions of each and every one at OpenBenchmarking.org -- from the Phoronix reviews/articles pages, usually just a matter of clicking on the title of the graph (for the SVG versions in modern browsers) to be brought to the page with plenty of details. Other feedback for improved usability is welcome.

Thanks for the info! I guess I shouldn't use my phone to read benchmark articles. Even if my phone's browser supports SVG I can't tell if anything is a clickable link without a mouse pointer. I'll click on the graph titles now so I can see what aspect of the CPU/GPU is being tested and how.

[darkbasic]

A7850 looks an ideal for a steam machine. Am I right? This will have the performance of a ps4 without the need for dedicated card, which would mean lower price and less cooling requirements / smaller enclosing.

No, you aren't. Catalyst sucks.

[mmstick]

A7850 looks an ideal for a steam machine. Am I right? This will have the performance of a ps4 without the need for dedicated card, which would mean lower price and less cooling requirements / smaller enclosing.

You would still need a dedicated card for high end gaming. The GPU in Kaveri isn't as powerful as the GPU in the PS4. It's just 512 cores that are clocked rather low and suffer an issue of memory bottlenecks. You would need to buy the fastest RAM possible. A good high end GPU would be 6-10x faster than Kaveri's iGPU

[nightmarex]

Please include ArchLinux (e.g. via Antergos) in the testing loop, in order to test the out-of-box-experience via the latest stable software out in the wild. Especially OpenCL-tests via the proprietary driver (also installable via package management on Arch; also install catalyst-hook to auto-compile the kernel module) would be interesting.

Anyone who uses Arch and an AMD bull/pile/steamy (lol that's too funy).. and doesn't use ABS with march native & -O2 is asking for a let down. Unless your point was to see how mostly vanilla apps perform with steamy.

[nightmarex]

A7850 looks an ideal for a steam machine. Am I right? This will have the performance of a ps4 without the need for dedicated card, which would mean lower price and less cooling requirements / smaller enclosing.

Well... as much as everyone is shouting "No!"... if you follow semi accurate at all there is a debate going on. The DDR3 control on kaveri is very interesting. speculation is that is can do quad channel if used with a BGA motherboard... this is probably just wishful thinking but...