Preliminary Tests Of Intel Sandy Bridge & Ivy Bridge vs. Broadwell

[dungeon]

Could we have a comparison of HD4400/4600 vs HD5500? Just curious.

HD5500 should be faster then HD4400, while i don't think it can exceed HD4600... but it might touch it a little and in best case scenario be te same as lowest clocked HD4600... so it should roughly speeking in between those two.

While HD6000 should be twice more powered then HD5500 , and Iris 6100 of course which is likely same GPU as HD6000 just plus IBM's eDRAM.

[l8gravely]

More performance/watt please!

Phoronix: Preliminary Tests Of Intel Sandy Bridge & Ivy Bridge vs. Broadwell

Chances are if you have a Haswell ultrabook/laptop, you're probably not looking at upgrading to a new Broadwell design unless your Haswell laptop had hardware issues, you really need a longer battery life via more power efficient hardware, or you just fall in love with one of the new Broadwell devices. If you're running an Ivy Bridge or Sandy Bridge laptop on the other hand, it might be time for an upgrade to get faster Intel graphics and greater power efficiency. Here's some preliminary figures I have for showing off the new Lenovo ThinkPad X1 Carbon Core i7 5600U compared to some older SNB and IVB laptops.

http://www.phoronix.com/vr.php?view=21427

Gotta say, the per/watt results were just amazing. If I was in the market for a new laptop, I'd be very interested in getting the latest one with such good perf/watt profiles. The comments where people focus exclusively on performance aren't getting the whole picture. Performance hasn't really climbed much in the past fix years or so. My main desktop at home is still an AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ and with 8gb of RAM, it's plenty fast for day to day work.

But in a laptop... where battery is never big enough, Broadwell is just fantastic.

Is there any hope of getting the linux results also showing the iterations/watt numbers as well?

John

[Michael]

Of course if you have one, it would be pretty simple to run pts using the same tests to compare. That's generally what I do. Micheals hardware preferences are different than mine, so I usually just ignore his bench results and run his benches on my hardware so I know how my hardware performs

Also it's highly likely that openbenchmarking.com already has bench results listed for that hardware. That site is a pain to navigate, but there are tons of hardware configurations already benchmarked there.

Any mockups for improvements, etc, welcome. Unfortunately I don't consider UI design to be one of my strong areas.

[Michael]

Gotta say, the per/watt results were just amazing. If I was in the market for a new laptop, I'd be very interested in getting the latest one with such good perf/watt profiles. The comments where people focus exclusively on performance aren't getting the whole picture. Performance hasn't really climbed much in the past fix years or so. My main desktop at home is still an AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ and with 8gb of RAM, it's plenty fast for day to day work.

But in a laptop... where battery is never big enough, Broadwell is just fantastic.

Is there any hope of getting the linux results also showing the iterations/watt numbers as well?

John

Happy to run more tests with whatever new test profiles or PTS patches are provided by the community.

[duby229]

Any mockups for improvements, etc, welcome. Unfortunately I don't consider UI design to be one of my strong areas.

Oh same here. I have zero artistic flare. That side of my brain is completely dormant.

Right now the easiest way to find a specific hardware is to use the hardware index. But the layout and structure is weird. To search for a specific hardware the link is in the bar above the site banner in the leftmost corner. To search for a test profile it's the link in the middle of the blue bar below the banner. It's just awkward. Then a lot of the time (Like right now) the site search doesn't work, so you have no choice but to use the awkward navigation links.

The big font-little font thing is strange.

And then when looking at the list of GPU's for example there is "Average Price (USD)" which has no data filled in.

[dungeon]

On the other hand H.264@60 fps plays but consumes a lot of CPU, so it is not usable for watching when power is not connected to laptop .

H264@60fps

youtube-dl https://www.youtube.com/watch?v=6_W2w0oBwb8 -f 299 -o wideoh264.mp4
youtube-dl https://www.youtube.com/watch?v=6_W2w0oBwb8 -f 141 -o audioh264.m4a
avconv -i wideoh264.mp4 -i audioh264.m4a -c:v copy -c:a copy -bsf:a aac_adtstoasc h264.mp4

That does not use much CPU here maybe 30-40%, but is also decodable on Athlon 5350, see:



L4.2 is not advertised by vdpauinfo, but all those 60fps h264 Kabini decode with no problem .

But yeah, no idea how intel work for h264 60 fps videos

[duby229]

EDIT: While it is awkward to navigate, it's a fantastic resource. I really like it.

[Michael]

Oh same here. I have zero artistic flare. That side of my brain is completely dormant.

Right now the easiest way to find a specific hardware is to use the hardware index. But the layout and structure is weird. To search for a specific hardware the link is in the bar above the site banner in the leftmost corner. To search for a test profile it's the link in the middle of the blue bar below the banner. It's just awkward. Then a lot of the time (Like right now) the site search doesn't work, so you have no choice but to use the awkward navigation links.

The big font-little font thing is strange.

And then when looking at the list of GPU's for example there is "Average Price (USD)" which has no data filled in.

The search will improve once I'm able to move OpenBenchmarking.org off to a better server as right now it's crammed for resources, when that happens it should be easier in that regard to find comparisons. The price data is obtained via Amazon.com but either their API changed or it's having some other sort of communication issue in finding prices atm.

[curaga]

Re this article: you posted the OpenSSL graph twice, first with the per-watt results, then again on the last page.

[Michael]

Re this article: you posted the OpenSSL graph twice, first with the per-watt results, then again on the last page.

It was technically separate data. The first run was on battery power during power measurements and the last one on AC power.