3DNews.ru: Gaming/Application results including Skylake i3 and Kaveri/Godavari

[Abwx]

You seem to have confused HTPC with NAS.

There s a related test on said review :

http://www.guru3d.com/articles_pages/amd_athlon_5350_apu_and_am1_platform_review,16.html

Also those APU are AMD Crismon drivers compatible and will benefit a longer lifecycle than most products..

 

[ShintaiDK]

Its an obsolete product for HTPC. The only products AMD got in that section is Carrizo, Tonga and Fiji. And the last two isn't fit for HTPC usage.

 

[Zodiark1593]

Lol, the "specialist" has spoken..

http://www.guru3d.com/articles-pages/amd-athlon-5350-apu-and-am1-platform-review,15.html



Indeed...


http://www.computerbase.de/2015-10/...m-multitasking-test-winrar-plus-the-witcher-3

Two applications stressing the CPU with merely 3 or 4 threads (for the two softs, one use 2 and the other use two at most !!) and the hyped IPC of Intel CPUs is collapsing.

I guess that s why some suspect insist on single thread, as if a PC had only one or two threads to deal with.

Actualy any Intel CPU below the i7 has catastrophic IPC in a lot of occurences that are trivial for any user, in short they have very poor throughput when it is actualy needed and good throughput when it is not needed....

It would seem that without Hyperthreading, Intel cpus are pretty heavily penalized when switching between threads, and starves the hungry cores as a result.

 

[Abwx]

Its an obsolete product for HTPC. The only products AMD got in that section is Carrizo, Tonga and Fiji. And the last two isn't fit for HTPC usage.

You should then apply the term obsolete even more accurately to any Intel product prior to SKL, particularly for Celeron/Pentium and Baytrails since their video capabilities are below Kabini s..

 

[ShintaiDK]

You should then apply the term obsolete even more accurately to any Intel product prior to SKL, particularly for Celeron/Pentium and Baytrails since their video capabilities are below Kabini s..

Baytrail (1080P HEVC only), Braswell, Haswell, Broadwell and Skylake are all way ahead of Kabini. Of the listed, Kabini is the only one not supporting HEVC/VP9.

You should do some homework,

 

[Abwx]

It would seem that without Hyperthreading, Intel cpus are pretty heavily penalized when switching between threads, and starves the hungry cores as a result.

This has surely nothing to do with HT given that 2C/2T CPUs behave better than the 2C/4T ones in the Integer + ST FP test but not as well in the Integer + Integer test, besides the i7 does not suffer from this "illness" at all, or at least with a reasonable penalty..

The determining factor is likely the cache size as the more the cache the more the threads that can be executed concurrently within a same core whithout too much throughput penalty, no surprise that the servers chips are granted huge amount of cache.

 

[BigDaveX]

http://www.computerbase.de/2015-10/...m-multitasking-test-winrar-plus-the-witcher-3

Two applications stressing the CPU with merely 3 or 4 threads (for the two softs, one use 2 and the other use two at most !!) and the hyped IPC of Intel CPUs is collapsing.

I guess that s why some suspect insist on single thread, as if a PC had only one or two threads to deal with.

Actualy any Intel CPU below the i7 has catastrophic IPC in a lot of occurences that are trivial for any user, in short they have very poor throughput when it is actualy needed and good throughput when it is not needed....

Um, you do realize that link actually shows that AMD fares worse in the Cinebench & WinRAR test than Intel? Haswell-E fares the best, dropping to 81% of its normal Cinebench score, all the quad-core i7s drop to 65% of their normal scores, and the FX-8350 actually fares the worst of the 8-thread CPUs, getting just 61% of its normal score.

If there were an FX-6350 in that mix and it significantly outperformed the i5s then yeah, it might prove something. For those few users who perform intensive encoding/rendering apps while also extracting ginormous WinRAR archives, anyway.

 

[Abwx]

Baytrail, Haswell, Broadwell and Skylake are all way ahead of Kabini. Of the listed, Kabini is the only one not supporting HEVC.

You should do some homework,

You, you should do your homework, or seriously think about another hoy than spreading fud about whatever AMD.


At the bottom of the page, with the CPU usages :

http://www.hardware.fr/articles/921-5/gpu-performances-jeux-opencl-h-264.html

Notice that Baytrail has a higher CPU usage in streamed video, and the Celeron G1820 is no better than Kabini...

 

[ShintaiDK]

You, you should do your homework, or seriously think about another hoy than spreading fud about whatever AMD.


At the bottom of the page, with the CPU usages :

http://www.hardware.fr/articles/921-5/gpu-performances-jeux-opencl-h-264.html

Notice that Baytrail has a higher CPU usage in streamed video, and the Celeron G1820 is no better than Kabini...

And Kabini still doesn't support HEVC or VP9. Its obsolete as HTPC, unlike Haswell Celerons and so on. Then you can try change the goalpost as much as you like. But support isn't going to happen. AMD only got 3 designs able to do it. Carrizo, Tonga and Fiji.

 

[Abwx]

Um, you do realize that link actually shows that AMD fares worse in the Cinebench & WinRAR test than Intel? Haswell-E fares the best, dropping to 81% of its normal Cinebench score, all the quad-core i7s drop to 65% of their normal scores, and the FX-8350 actually fares the worst of the 8-thread CPUs, getting just 61% of its normal score.

If there were an FX-6350 in that mix and it significantly outperformed the i5s then yeah, it might prove something. For those few users who perform intensive encoding/rendering apps while also extracting ginormous WinRAR archives, anyway.

Still you missed that the FX is at 61% for CB but also at 92% for Winrar while the i7 are at 67% for the 4790K and only 71% for the 6700K.

So 4% advantage on FP is hardly compensated by 21-24% in Integer.

So much for FX being worse when it s actualy much better than even Intel s latest in matter of sustained throughput..

Edit : FX6 should behave better than FX8 given that less cores benefit from the same 8MB cache, the E5820 is a hint that cache size matter and it should be no different for the FX a some point.

And Kabini still doesn't support HEVC or VP9. Its obsolete as HTPC, unlike Haswell Celerons and so on. Then you can try change the goalpost as much as you like. But support isn't going to happen. AMD only got 3 designs able to do it. Carrizo, Tonga and Fiji.

What is obsolete is your fud and innaccuracies, no Intel CPU prior to SKL has H265 hardware decoding, assuming their decoder work accordingly as Hardware.fr stated that Intel drivers had never been so instable since Windows Vista, so perhaps that you ll end complaining that reviewers use older ones...

 

[BigDaveX]

Still you missed that the FX is at 61% for CB but also at 92% for Winrar while the i7 are at 67% for the 4790K and only 71% for the 6700K.

So 4% advantage on FP is hardly compensated by 21-24% in Integer.

So much for FX being worse when it s actualy much better than even Intel s latest in matter of sustained throughput..

Edit : FX6 should behave better than FX8 given that less cores benefit from the same 8MB cache, the E5820 is a hint that cache size matter and it should be no different for the FX a some point.

If you look through the full list, Haswell-E, the Sandy Bridge and Ivy Bridge i7s, the FX-8350, all the i5s and the two 2C/2T chips all lose around just 10% or so from their WinRAR times, whereas the Skylake and Haswell i7s (and the Haswell i3) completely tank.

So, either Haswell and Skylake (but not Haswell-E, for some odd reason) have some horrible glitch that causes performance to tank with Hyper-Threading enabled, which has somehow gone completely unnoticed until now... or we're just looking at some sort of scheduling error that's hurting performance on those chips. And considering that the i3 performs pretty much like you'd expect a higher-clocked Pentium to, my money's on the latter.

 

[TheELF]

Two applications stressing the CPU with merely 3 or 4 threads (for the two softs, one use 2 and the other use two at most !!) and the hyped IPC of Intel CPUs is collapsing.

I guess that s why some suspect insist on single thread, as if a PC had only one or two threads to deal with.

Actualy any Intel CPU below the i7 has catastrophic IPC in a lot of occurences that are trivial for any user, in short they have very poor throughput when it is actualy needed and good throughput when it is not needed....

Are you living in 1995?
Winrar launches a large amount of threads,oh and notice the difference in CPU % usage between real case scenario and benchmark...
https://www.youtube.com/watch?v=mcbICLEeIBg
Witcher 3 also launches a dozen threads and more.
https://www.youtube.com/watch?v=Lejul04cMmk Witcher 3

IPC is the amount of work a core can do,how you decide to split it up has nothing to do with the core,or with IPC collapsing,the celeron manages 60FPS in the witcher so of course if you decide to run something else FPS will go down in the amount of work that is transferred to the other task(winrar) in other words 20FPS of processing power go towards running winrar.

If you so decide you can command the second task to only use up the idle cycles,where is the collapsing IPC now?
GTA IV + video converting

 

[ShintaiDK]

What is obsolete is your fud and innaccuracies, no Intel CPU prior to SKL has H265 hardware decoding, assuming their decoder work accordingly as Hardware.fr stated that Intel drivers had never been so instable since Windows Vista, so perhaps that you ll end complaining that reviewers use older ones...

You obviously missed the news:
http://techreport.com/news/27677/new-intel-igp-drivers-add-h-265-vp9-hardware-decode-support

Haswell and forward all got HEVC and VP9 decode (Later driver). Including Main10 that AMD doesn't support in Carrizo/Tonga/Fiji.

Again, your Kabini like most of AMDs lineup is simply obsolete for video decode.

 

[TheELF]

where is the collapsing IPC now?

As a matter of fact look at those two as well,beloved by every AMD fanboy x264 and 7zip, running additional threads per core "raises IPC" on intel cores.

7z benchmark with 1 to 10 threads on g1820 with CPU load and threads showing
Multicore Myth G1820 x264 transoding with 1 to 24 threads

 

[Abwx]

Are you living in 1995?
Winrar launches a large amount of threads,oh and notice the difference in CPU % usage between real case scenario and benchmark...
https://www.youtube.com/watch?v=mcbICLEeIBg
Witcher 3 also launches a dozen threads and more.
https://www.youtube.com/watch?v=Lejul04cMmk Witcher 3

Softwares are core count aware, they will as much threads a the number of cores and not a single more, it would be extremely bad software design than to launch too much threads as it stall the pipeline uselessly.

IPC is the amount of work a core can do,how you decide to split it up has nothing to do with the core,or with IPC collapsing,the celeron manages 60FPS in the witcher so of course if you decide to run something else FPS will go down in the amount of work that is transferred to the other task(winrar) in other words 20FPS of processing power go towards running winrar.

If you so decide you can command the second task to only use up the idle cycles,where is the collapsing IPC now?
GTA IV + video converting

IPC will be the same but with very poor throughput as the pipeline is filled with too much non dependent datas.

If you so decide you can command the second task to only use up the idle cycles,where is the collapsing IPC now?
GTA IV + video converting

Are you serious.?.

Where are the numbers, what is the fps of the X264 encoding.?.

And you call that playing a game, at ultra low def on a tiny applet that has even less pixels..?..

 

[Abwx]

As a matter of fact look at those two as well,beloved by every AMD fanboy x264 and 7zip, running additional threads per core "raises IPC" on intel cores.

7z benchmark with 1 to 10 threads on g1820 with CPU load and threads showing
Multicore Myth G1820 x264 transoding with 1 to 24 threads

As long as there s not the slightest background apps running, it is well known that with an Intel CPU one doesnt need any antivirus, firewall, while the wifi connection need no driver and demanding CPU mangement of the flux..

Because it s about supporting four threads for 2 apps, not a lot of threads for a single app, what you re pathetically using as a mean to change the goal posts, so much for the fanboy moniker, apply it to yourself, with the bad faith one as well..

The Computerbase test is clear, load an Intel CPU with one apps and it will work as expected, launch a second app and the throughput will collapse, and dramaticaly with the Celeron and i3s wich can get as low as the equivalent of a 1C/1T and 1C/2T of such a CPU...

 

[Zodiark1593]

As long as there s not the slightest background apps running, it is well known that with an Intel CPU one doesnt need any antivirus, firewall, while the wifi connection need no driver and demanding CPU mangement of the flux..

Because it s about supporting four threads for 2 apps, not a lot of threads for a single app, what you re pathetically using as a mean to change the goal posts, so much for the fanboy moniker, apply it to yourself, with the bad faith one as well..

The Computerbase test is clear, load an Intel CPU with one apps and it will work as expected, launch a second app and the throughput will collapse, and dramaticaly with the Celeron and i3s wich can get as low as the equivalent of a 1C/1T and 1C/2T of such a CPU...

Seems quite extreme to me. A dual core celeron dropping to the performance of a single core? You'll have excuse me if I dont believe those claims without seeing such results firsthand.

Perhaps the Blender Cycles BMW benchmark might be a viable method to verify this. The software is free, will max out threads it's given, and the amount of threads are easily configured (for example, load 8+ threads on an i5). Should render times jump by %50+ percent, that would/prove you as correct.

 

[TheELF]

Ok at least look at the videos before posting nonsense, you can see all the icons,you can see the usage in process explorer.

 

[TheELF]

Are you serious.?.

Where are the numbers, what is the fps of the X264 encoding.?.

And you call that playing a game, at ultra low def on a tiny applet that has even less pixels..?..

Yes I call it playing because the graphics don't influence the CPU,the game still uses up 70% of the CPU.
The numbers don't matter you can control how fast each one will run that is what matters,no core not even amd's can run something with 100% speed and then run something else as well without loosing speed.

 

[Sweepr]

Um, you do realize that link actually shows that AMD fares worse in the Cinebench & WinRAR test than Intel? Haswell-E fares the best, dropping to 81% of its normal Cinebench score, all the quad-core i7s drop to 65% of their normal scores, and the FX-8350 actually fares the worst of the 8-thread CPUs, getting just 61% of its normal score.

Ouch, this one backfired fast.
Note that ComputerBase didn't even include any AMD APU in the mix. They would perform even worse than Vishera here.

So, either Haswell and Skylake (but not Haswell-E, for some odd reason) have some horrible glitch that causes performance to tank with Hyper-Threading enabled, which has somehow gone completely unnoticed until now... or we're just looking at some sort of scheduling error that's hurting performance on those chips. And considering that the i3 performs pretty much like you'd expect a higher-clocked Pentium to, my money's on the latter.

Sounds likely.

 

[Sweepr]

Just to show how much importance this website gives to the 'multitasking' tests often quoted by a certain shill, ComputerBase itself recommends Haswell Core i5 over FX if you're spending 150-200 euros on a CPU.

www.computerbase.de/thema/prozessor/rangliste

 

[Dresdenboy]

And Kabini still doesn't support HEVC or VP9. Its obsolete as HTPC, unlike Haswell Celerons and so on. Then you can try change the goalpost as much as you like. But support isn't going to happen. AMD only got 3 designs able to do it. Carrizo, Tonga and Fiji.

How old are you? Don't you remember the times of software decoding? An HTPC is not running on battery, so it might use the CPU/OpenCL decoders for these codecs instead or receive a differently encoded stream instead.

 

[ShintaiDK]

How old are you? Don't you remember the times of software decoding? An HTPC is not running on battery, so it might use the CPU/OpenCL decoders for these codecs instead or receive a differently encoded stream instead.

Old enough to know what software decoding takes of resources.

CPU+OpenCL gives you a limited support of 1080P HEVC 8bit decode at 30FPS.

Now show me software decode of 4K, including Main10 and VP9 all in 60 FPS

So maybe you should try see if you could get your AMD to actually deliver the needed hardware decode. Rather than searching for an excuse or terrible hotfix that wont work just like AtenRa.

 

[Sweepr]

Old enough to know what software decoding takes of resources.

Which is bad, especially for mobile.

CPU+OpenCL gives you a limited support of 1080P HEVC 8bit decode at 30FPS.
Now show me software decode of 4K, including Main10 and VP9 all in 60 FPS

Considering their focus on GPU, that's a shame. AMD's graphics performance advantage is being reduced to the point I'm sure some people are really looking forward to 2017 and praying for some HBM solution (that might or might not reach client segment anyway). Before that, there's Skylake GT4e and a new lineup of Kaby Lake CPUs with more powerful iGPUs and fixed function HEVC Main10/10bit / VP9 10bit hardware decoding.

 

[ShintaiDK]

The best they can hope for APU wise, will be some Zen APU in 2017 to fix these issues. That's just plain awful.

And who knows how the GPU lineup will look and how much that will be rebrand without support.