The original post was to evaluate the performance of desktops versus laptops for 4k 264 and 265 video editing



This was posted by JL_JL on a different thread and is the heart of the subject::

TonyL/others, the simple little test was suggested to Eugene as he was questioning the idea, how his CPU could be limiting the capability of his GTX960 GPU considering his CPU was only operating in the ~30% range during encoding. I simply suggested a 4K profile as that’s Eugene’s typical interest and it does present a significant challenge of elapsed wall time to encode.

Maybe this set of instructions will suffice:
1) Put 10 of the PD default media “Kite Surfing.wmv” into the timeline. If you don’t load PD default media modify your PD pref to load. Pref > Project.
2) In the PD “Produce” tab, set “Profile name” as H.265, MKV and default preset profile, HEVC 4K 4096 x 2160/30p (37Mbps)
3) In the PD “Produce” tab, set “Fast video rendering technology:” to enable “Hardware video encoder”
4) Start the Produce operation
5) Monitor GPU load of the GTX960 with free utility TechPowerUp GPU-Z, under sensor tab, the “Video Engine Load” ( https://www.techpowerup.com/downloads/SysInfo/GPU-Z/ )
6) Monitor CPU and RAM usage during the encode processes with default windows Task Manager
7) Items 5 and 6 are usually rather stable throughout the encoding process for this simple test provided if you don’t have significant “other” stuff running
When produce operation is complete, the “Time elapsed:” and “Produced” files size from the PD window.

Repeate above steps with item 2) of: H.264, MP4, MPEG-4 4K 4096 x 2160/30p (50Mbps)

Sample pic of results attached. From my experience what one will see is a reasonable separation of GTX960 capability, probably on the order of 50% difference or so for the exact same task on different computers.

Jeff
[Thumb - GTX960_performance.png]
Filename GTX960_performance.png [Disk] Download
Description
Filesize 23 Kbytes
Downloaded: 7 time(s)
-----------------------------------------------------------------

This too is from an other thread and contains my test results:

The encoding time test is a GREAT idea Jeff!! No guess work here and everybody can do the same exact test.

Using 10 ea kite surfing as suggested ,total length 1.42

Encoding 4K MKV 265 HEVC, 37 Mb, 30FPS. Encoding time 2.02 min, GPU video engine 80%, CPU 35% 6.3 GB memory used.Video 460MB


Play back MPC64, GPU 26%, CPU 6%.

-------------

Encoding to 4K 264 MP4, 49Mb, 30FPS encoding time 2.02 min, GPU video engine 43%, CPU 39%,4.6 GB memory, Video 626MB.


Play back MPC64, GPU 29%, CPU 6%

I was surprised to see the 960 GPU video engine running at 80% during HEVC encoding, not much room left there.
Encoding is almost in real time, video only no transitions etc.

My jaws are still on the floor when I look at these numbers in regard to HEVC. W/O the 960 it would have taken at least 1Hour 30 minutes to encode the same 1.42 min in PD14 HPQ mode.

Encoded video looks as good as the original.



Looking forward to see other forum members numbers. ESPECIALLY from LAPTOP owners

Eugene

PepsiMan posted very interesting results but spread over mant pages, could you please combine the results on one page?



TonyL

I noticed that your computer is the same as mine with the exception of the GPU. Could you please run Jeff's test and post the results? You have everything neded for that. Took me less than 1/2 hr to do.

Also the data file is not visible.

Eugene 73s, WA6JZN ex DL9GC
CYBERLINK PLEASE ADD UHD BLU RAY BURNING SOFTWARE
PD14,
Win10,64bit.CPU i7 6700,16GB ,C= 480 GB SSD ,GPU GTX1060 6GB 1 fan. Plus 3 int, 4 ext HDD's for video etc.LG WH16NS40 reads UHD.
4K 24" ViewSonic monitor.Camera Sony FDR-A

Eugene

I don't know why you can't access the worksheet. When I click the link above it opens. Anyone with the link can access the data sheet.

Please try the link again and if it doesn't work it might help if you gave me a little more info if you can

Re my own data, I've already said in the penultimate para that I'll be providing it but have been too busy creating the worksheet. Did you not read it?

TonyL
Self build Xeon W3690, 12GB RAM, 850EVO SSD, Asus 2GB GTX960, Win10H 64bit, PD14
iMac 27" Retina 5K i5, 24GB RAM, Radeon M295X 4GB, Bootcamp Win10H 64bit

Laptops in general do not have enough reserve in either number crunching or data transfer capability for HD and hgher video projects. A desktop made in the past 5 years with an i5 or higher CPU can outperform. Win 10, i7

done, redone and posted online!



Eugene.

sorry. failed to follow instructions; however, gave everyone halloween candies...

followed TonyL's online worksheet this time.



PepsiMan

p.s.

i have one more payment to make on the Samsung NX1 &

i might speeded up and order MSI GS70 Stealth-608 17.3" Broadwell i7-5700HQ/GTX 965M for poops and giggles!

no. i am not that a coocoo to spend $3K on a laptop for GTX980. 'no bridge too far'

Yashica Electro 8 LD-6 Super 8mm
Asrock TaiChi X470, AMD R7 2700X, W7P 64, MSI GTX1060 6GB, Corsair 16GB/RAM
Dell XPS L702X i7-2860QM, W7P / W10P 64, Intel HD3000/nVidia GT 550M 1GB, Micron 16GB/RAM
Samsung Galaxy Note3/NX1

PepsiMan

I cannot find any data. What is the location?

Eugene 73s, WA6JZN ex DL9GC
CYBERLINK PLEASE ADD UHD BLU RAY BURNING SOFTWARE
PD14,
Win10,64bit.CPU i7 6700,16GB ,C= 480 GB SSD ,GPU GTX1060 6GB 1 fan. Plus 3 int, 4 ext HDD's for video etc.LG WH16NS40 reads UHD.
4K 24" ViewSonic monitor.Camera Sony FDR-A

Quote: i might speeded up and order MSI GS70 Stealth-608 17.3" Broadwell i7-5700HQ/GTX 965M for poops and giggles!

no. i am not that a coocoo to spend $3K on a laptop for GTX980.

Check out the new MSI GT72 Dominator with latest Skylake processor! Several different models and GTX GPUs. Wish I had waited a few months! Intel i7 4770k, 16GB, GTX1060 3GB, Two 240GB SSD, 4TB HD, Sony HDR-TD20V 3D camcorder, Sony SLT-A65VK for still images, Windows 10 Pro, 64bit
Gary Russell -- TN USA

Quote: PepsiMan

I cannot find any data. What is the location?

Eugene

here it is. we're die hards! we're really retired and rest'em still working or love money!

https://docs.google.com/spreadsheets/d/1cvWno5C9QZ-73ncW0AtRnC5zeNA5i822YleknOqlLtU/edit?pli=1#gid=0

Quote:

... Anyone with the following link can access the sheet and add their data.

Click here to access the data sheet.

I've applied some protection but not sure how well it works so take care to not delete other people's data...

click the ' Click here to access the data sheet. ' sentence.

Quote:

Check out the new ' MSI GT72 Dominator ' with latest Skylake processor! Several different models and GTX GPUs. Wish I had waited a few months!

thank you. thank you for the link!

i'm drooooling all over... again.



GTX 965M is rated 2932 / GTX 970M is rated 4366 / GTX 950 is rated 5297 / GTX 960 is rated 5977 /

GTX 980M is rated 6074 / GTX 970(HA encode) is 8655 / GTX 980(HA encode) is 9765!!!

now that's smoking!



PepsiMan

p.s.

of course we're talking about HA H.265 HEVC encoding & decoding... 'no bridge too far'

Yashica Electro 8 LD-6 Super 8mm
Asrock TaiChi X470, AMD R7 2700X, W7P 64, MSI GTX1060 6GB, Corsair 16GB/RAM
Dell XPS L702X i7-2860QM, W7P / W10P 64, Intel HD3000/nVidia GT 550M 1GB, Micron 16GB/RAM
Samsung Galaxy Note3/NX1

Eugene,

Try clicking the link that says Click here to access the data sheet.

By the way, I don't have a GTX960 in my machine yet. It only arrived yesterday afternoon. I've got the data for my current card and will get it for the GTX after I've installed it, hopefully today.

TonyL Self build Xeon W3690, 12GB RAM, 850EVO SSD, Asus 2GB GTX960, Win10H 64bit, PD14
iMac 27" Retina 5K i5, 24GB RAM, Radeon M295X 4GB, Bootcamp Win10H 64bit

I posted my data, but I don't think encoding a 1280x720 video to 4K is a good test for "PC performance comparison".

It is at best a GPU compression test, and still not that good of one considering the source content isn't even 4K to begin with.

It doesn't really test the CPU much . There is no manipulation of the source content at all.

For example, no lighting adjustment, stabilizer, video denoise, color adjustment, white balance adjustment - to list just the common options under "fix/enhance". I often use at least one of those, sometimes two, occasionally three. I can attest that the processing for those usually trounces the time it takes to compress. Those are heavily CPU-limited, usually.

There is no title track also - almost all my projects have one.

There are many other factors in a project that seem to affect performance greatly - for example, one of my projects which displays half of two clips side by side - one 4K clip and anothe HD clip - is a major CPU killer. And I have seen some differences of 2 to 3x between my result and those of others, which are truly puzzling to me.

IMO, if we seriously want something that measures PC performance for PowerDirector, we should have a better test project than this.

Actually, probably a set of projects, to be processed in batch mode. MSI X99A Raider
Intel i7-5820k @ 4.4 GHz
32GB DDR4 RAM
Gigabyte nVidia GTX 960 4GB
480 GB Patriot Ignite SSD (boot)
2 x 480 GB Sandisk Ultra II SSD (striped)
6 x 1 TB Samsung 860 SSD (striped)

2 x LG 32UD59-B 32" 4K
Asus PB238 23" HD (portrait)

Quote: It is at best a GPU compression test, and still not that good of one considering the source content isn't even 4K to begin with.

It doesn't really test the CPU much . There is no manipulation of the source content at all.

For example, no lighting adjustment, stabilizer, video denoise, color adjustment, white balance adjustment - to list just the common options under "fix/enhance". I often use at least one of those, sometimes two, occasionally three. I can attest that the processing for those usually trounces the time it takes to compress. Those are heavily CPU-limited, usually.

The test was suggested for a very narrow question. The question was, will a GTX960 perform any faster that what I have? My CPU is only at 30%, so would a different CPU really reduce HA encoding times of the GTX960.

To answer this question only, I suggested the this test for the exact opposite reason you state. You want something that does not test all the other tasks a CPU does like lighting adjustment....etc, as then I'm not testing the capability of the GTX960, the test is evaluating their CPU and many PC's will never load the GTX960 to it's capability level. You don't want 4K source material as then the test depends on what does the decoding, CPU, GPU, and exactly how the user configured their pref. This suggested test won't care, a pure play on what a GTX960 is capable of HA encoding wise.

Jeff

I've input my readings to the data sheet and made a slight change to the topic header to hopefully avoid further misunderstanding.

I will carry out further tests and input the readings after I've changed my CPU and then after a change to SSD. These will be to show the effect on the 960 of making those hardware changes.

TonyL Self build Xeon W3690, 12GB RAM, 850EVO SSD, Asus 2GB GTX960, Win10H 64bit, PD14
iMac 27" Retina 5K i5, 24GB RAM, Radeon M295X 4GB, Bootcamp Win10H 64bit

it will. from 7200 rpm HD to SSD drive...



Pierre killed my GTX 750Ti numbers on the H.265 HEVC encoding time!

mine was 1 hours+ compared to Pierre's 12:29!!!

• correction:

i've reinstalled GTX 750Ti and the result is...

Pierre still beat me at 17:01 vs 12:29!



PepsiMan 'no bridge too far'

Yashica Electro 8 LD-6 Super 8mm
Asrock TaiChi X470, AMD R7 2700X, W7P 64, MSI GTX1060 6GB, Corsair 16GB/RAM
Dell XPS L702X i7-2860QM, W7P / W10P 64, Intel HD3000/nVidia GT 550M 1GB, Micron 16GB/RAM
Samsung Galaxy Note3/NX1

TonyL Thanks, found the data.

My driver is 353.62.

CPU speed 2.67 GHZ

I am surprised that PepsiMan has such a high CPU use number even with the 960.

Thanks for compiling that.

Eugene 73s, WA6JZN ex DL9GC
CYBERLINK PLEASE ADD UHD BLU RAY BURNING SOFTWARE
PD14,
Win10,64bit.CPU i7 6700,16GB ,C= 480 GB SSD ,GPU GTX1060 6GB 1 fan. Plus 3 int, 4 ext HDD's for video etc.LG WH16NS40 reads UHD.
4K 24" ViewSonic monitor.Camera Sony FDR-A

Quote:

• correction:

i've reinstalled GTX 750Ti and the result is...

Pierre still beat me at 17:01 vs 12:29!

Most likely has virtually nothing to do with the 750Ti. This particular encode on a GTX750Ti has to be done by the CPU, not supported on GPU. It is supported on second gen Maxwell, GTX9xx series, the focus these tests. Your CPU, 8370E is simple slower than Pierre's OC 8350. You can simply ratio the numbers from PassMark http://www.cpubenchmark.net/ adjusting for Pierre's OC condition and you get he should be about 1.32 faster. Your timed encode test, 1021/749=1.36.

Jeff

Quote:

Quote: i might speeded up and order MSI GS70 Stealth-608 17.3" Broadwell i7-5700HQ/GTX 965M for poops and giggles!

no. i am not that a coocoo to spend $3K on a laptop for GTX980.

Check out the new MSI GT72 Dominator with latest Skylake processor! Several different models and GTX GPUs. Wish I had waited a few months!

The GTX970 and GTX980 make less sense for PowerDirector than GTX960 due to the lack of H.265 decoder.

And the GTX970/980 are also power hungry and don't make much sense on a laptop.

If you are really going for a laptop, then you would probably want one with a GTX960M, or GTX950M if you care at all about battery life. MSI X99A Raider
Intel i7-5820k @ 4.4 GHz
32GB DDR4 RAM
Gigabyte nVidia GTX 960 4GB
480 GB Patriot Ignite SSD (boot)
2 x 480 GB Sandisk Ultra II SSD (striped)
6 x 1 TB Samsung 860 SSD (striped)

2 x LG 32UD59-B 32" 4K
Asus PB238 23" HD (portrait)

JL_JL:

thank you.



Julien Pierre:

you're right on the GTX 980/970 and having thirst for electricity and the cooling fan noises; however,

GTX 980M/970M supports both HA HEVC encoding and decoding where GTX 960M/950M

DOES NOT.

did some homework at the NotebookCheck.net Reviews.



just drooling over MSI with GTX 965M, when GGRussell who bought a brand new laptop

recommended the Skylake than Broadwell cpu.



if Santa or Mrs. Santa is generous then i'll be a good boy till end of the year!



PepsiMan 'no bridge too far'

Yashica Electro 8 LD-6 Super 8mm
Asrock TaiChi X470, AMD R7 2700X, W7P 64, MSI GTX1060 6GB, Corsair 16GB/RAM
Dell XPS L702X i7-2860QM, W7P / W10P 64, Intel HD3000/nVidia GT 550M 1GB, Micron 16GB/RAM
Samsung Galaxy Note3/NX1

Quote:

Quote:

• correction:

i've reinstalled GTX 750Ti and the result is...

Pierre still beat me at 17:01 vs 12:29!

Most likely has virtually nothing to do with the 750Ti. This particular encode on a GTX750Ti has to be done by the CPU, not supported on GPU.

Correct, my H.265 number was a software encode/software decode, since the 750Ti GPU does not support H.265 at all.

For H.265, there was no need for Pepsiman to switch the card - simply unchecking the "fast rendering" button would have sufficed.

Pepsiman's number for H.264 is virtually identical between his GTX960 and GTX750Ti however - 3:15 and 3:19, which suggests he is CPU limited.

gen Maxwell, GTX9xx series, the focus these tests. Your CPU, 8370E is simple slower than Pierre's OC 8350. You can simply ratio the numbers from PassMark http://www.cpubenchmark.net/ adjusting for Pierre's OC condition and you get he should be about 1.32 faster. Your timed encode test, 1021/749=1.36.
Jeff

That's true for the H.265 case.

But for the H.264 case, there is something wrong on Pepsiman's system, IMO.

My CPU usage is only 35% with my Gigabyte GTX750Ti, and rendering time is 1:42 - 102s.

His CPU usage is 98% with his MSI GTX750Ti, and rendering time 3:19 - 199s.

This is a huge difference - he is using far more CPU, and yet getting a rendering time that's 95% longer, almost double.

IMO, this is probably explained by a software issue. Possibly the combinations of the older nVidia drivers, and the older PD build Pepsiman is running have something to do with it. But since the difference is so large, I would say there is more likely something else on the system interferring - maybe some CODEC package - I don't know. I don't have any 3rd party CODEC installed on my system.

FYI, the specific 750Ti GPU I have is this :

http://www.gigabyte.us/products/product-page.aspx?pid=4948#ov

Anyway, I see that the title of this thread has been renamed to "GTX960 Performance Comparisons", so I guess I don't belong here. I am still considering getting a GTX960 on black friday. But to be honest, since I don't have an immediate need for H.265, a faster CPU is much more appealing. I might do both.

If I go Intel, I'm not sure if the Skylake 6700k or the Haswell-e 5820k makes more sense. The 5820k regularly sells at Fry's for under $300 . It was at $277 on a 1-day sale last saturday, which I missed. They were out of stock in the whole bay area by the time I showed up near store closing time. But the salesman told me it will go on sale again when they get more inventory, usually about every 2 weeks. The 6700k is $349 regular price. Of course, the X99 motherboards cost more than the Z170. In both cases, I would have to purchase DDR4 memory.

If I do this, I will upgrade my older Phenom II x6 1055T box which is OC at 3.6 GHz and currently a secondary box. This would become the primary box, and the FX-8350 would become secondary. This Phenom II box had a pair of 560Ti which I just sold on Amazon. Now it's running with a combination of the 9800GT and the built-in AMD 4290, in order to drive 3 displays.

Does anyone want to buy my 9800GT and also the 8600GT ? Both can still drive my two HP LP3065 2560x1600 monitors with the two dual-link DVI. They are fine 2D GPUs. Just not suitable for hardware encoding. MSI X99A Raider
Intel i7-5820k @ 4.4 GHz
32GB DDR4 RAM
Gigabyte nVidia GTX 960 4GB
480 GB Patriot Ignite SSD (boot)
2 x 480 GB Sandisk Ultra II SSD (striped)
6 x 1 TB Samsung 860 SSD (striped)

2 x LG 32UD59-B 32" 4K
Asus PB238 23" HD (portrait)

Quote:

GTX 980M/970M supports both HA HEVC encoding and decoding where GTX 960M/950M

DOES NOT.

did some homework at the NotebookCheck.net Reviews.

Sorry, my bad. nVidia marketing strikes again. Indeed, looks like the GTX950M and GTX960M are both first gen Maxwell. Sigh.

https://en.wikipedia.org/wiki/GeForce_900_series#GeForce_900M_.289xxM.29_series

According to this table, the GTX965M/970M/980M/980(Notebook) are all GM204 , which supports HA for H.265 hardware encoding, but not decoding.

Only the GM206 in the GTX950/960 (desktop versions) support H.265 hardware decoding. Seems there is no laptop version of GM206 at this time.

https://en.wikipedia.org/wiki/GeForce_900_series#GeForce_900_.289xx.29_series

just drooling over MSI with GTX 965M, when GGRussell who bought a brand new laptop

recommended the Skylake than Broadwell cpu.

I don't think there are any laptop equivalents of the 5820k (6-core/12 threads), are there ?

On the desktop I'm not sure the choice is so clear as on a laptop./ MSI X99A Raider
Intel i7-5820k @ 4.4 GHz
32GB DDR4 RAM
Gigabyte nVidia GTX 960 4GB
480 GB Patriot Ignite SSD (boot)
2 x 480 GB Sandisk Ultra II SSD (striped)
6 x 1 TB Samsung 860 SSD (striped)

2 x LG 32UD59-B 32" 4K
Asus PB238 23" HD (portrait)

Quote:

Quote: It is at best a GPU compression test, and still not that good of one considering the source content isn't even 4K to begin with.

It doesn't really test the CPU much . There is no manipulation of the source content at all.

For example, no lighting adjustment, stabilizer, video denoise, color adjustment, white balance adjustment - to list just the common options under "fix/enhance". I often use at least one of those, sometimes two, occasionally three. I can attest that the processing for those usually trounces the time it takes to compress. Those are heavily CPU-limited, usually.

The test was suggested for a very narrow question. The question was, will a GTX960 perform any faster that what I have? My CPU is only at 30%, so would a different CPU really reduce HA encoding times of the GTX960.

To answer this question only, I suggested the this test for the exact opposite reason you state. You want something that does not test all the other tasks a CPU does like lighting adjustment....etc, as then I'm not testing the capability of the GTX960, the test is evaluating their CPU and many PC's will never load the GTX960 to it's capability level. You don't want 4K source material as then the test depends on what does the decoding, CPU, GPU, and exactly how the user configured their pref. This suggested test won't care, a pure play on what a GTX960 is capable of HA encoding wise.

Jeff

OK, fair enough. But in that case, don't call it "PC performance comparison". Call it "GTX960 compression comparison".

I'm really much more interested in the differences between CPUs.

In the following thread

http://forum.cyberlink.com/forum/posts/list/45843.page

You posted some results with the project 4ktest-minuet_DT.pds where your encoding time was 98s, to my 208s.

Specifically, this post :

http://forum.cyberlink.com/forum/posts/list/25/45843.page#238049

I have asked you multiple times what CPU you had in this test, as I simply can't understand how your encoding time could be half mine, when my CPU is hardly a low-end one. I even sent you a PM about it. Did you miss it ? I would really appreciate a response as I'm considering a CPU upgrade.

Same question for the CPUs used in this post :

http://forum.cyberlink.com/forum/posts/list/45843.page#237065

As the words "mid level CPU" and "higher end CPU" are fairly meaningless to me. My FX-8350 OC is as high end as it gets in the AMD line. MSI X99A Raider
Intel i7-5820k @ 4.4 GHz
32GB DDR4 RAM
Gigabyte nVidia GTX 960 4GB
480 GB Patriot Ignite SSD (boot)
2 x 480 GB Sandisk Ultra II SSD (striped)
6 x 1 TB Samsung 860 SSD (striped)

2 x LG 32UD59-B 32" 4K
Asus PB238 23" HD (portrait)