Question about gpu clock speeds affecting video encoding time.

[wade7575]

I want to get an ASROCK B580 when they come out as I really like my A580 and would like to use it mainly for video encoding using Handbrake first and foremost and maybe a bit for gaming but not a lot if any.

What I'm wondering is that the Intel B580 Founders Edition has a clock speed of 2670mhz and the ASROCK B580 Challenger has a 2740mhz and the ASROCK B580 Steel Legend has a 2800mhz clock speed.

What I don't know much about is does the clock speed affect video encoding speeds,I can purchase a Challenger or the Steel Legend for only 15 dollars more and I was just thinking of clock speeds helps even a little bit I would get that card.

For anyone who wants to reply to this thread please don't tell me to buy an Nvidia card as I don't want a card the sucks as much power as a 3060 or a 4070,I would like to stick with Intel as my A580 has been so far but like everyone else I'm after that little bit of extra speed.

 

[mackerel]

Not sure about this, but I think the video block is separate from the cores and I don't know if they scale with core clock. You could try testing it on your A580. Prepare a encode and run it as normal. Then underclock your GPU and repeat.

 

[EarthDog]

I'm not sure either, but was also under the impression it doesn't scale as the encoder (nvenc, quicksync, and vce) is separate from that clock domain... Test it. Looking forward to confirming the answer.

I won't tell you to get another card, but I will share power use among the three cards is negligble (170, 190, and 200w with b580 @ 190).

 

[rainless]

I want to get an ASROCK B580 when they come out as I really like my A580 and would like to use it mainly for video encoding using Handbrake first and foremost and maybe a bit for gaming but not a lot if any.

What I'm wondering is that the Intel B580 Founders Edition has a clock speed of 2670mhz and the ASROCK B580 Challenger has a 2740mhz and the ASROCK B580 Steel Legend has a 2800mhz clock speed.

What I don't know much about is does the clock speed affect video encoding speeds,I can purchase a Challenger or the Steel Legend for only 15 dollars more and I was just thinking of clock speeds helps even a little bit I would get that card.

For anyone who wants to reply to this thread please don't tell me to buy an Nvidia card as I don't want a card the sucks as much power as a 3060 or a 4070,I would like to stick with Intel as my A580 has been so far but like everyone else I'm after that little bit of extra speed.

Well a lot depends on your encoder. I can't speak for Handbrake, specifically, but almost every NLE these days (and most encoders) use GPU excelleration. Some (such as DaVinci Resolve) have a strong bias towards NVIDIA gpus...

That said... since this is an apples-to-apples comparison between Intel and Intel... that small a difference in clock speed shouldn't have any impact on encoding speed AT ALL. I mean you're talking 60mhz... which is literally nothing.

The largest impact you'll ever see in encoding is an increase or decrease in the number of cores.

That's all I ever care about when upgrading my GPUs: How many more cores does the new card have over my previous card.

The clock speed has always been irrelevant.

This was true when I went from 960GTX to RTX 2060 to RTX 4060... this will ALWAYS be true.

So if there's no difference in cores between the two cards... get either one.

 

[Kenrou]

There are very small gains with both core and memory clocks on the GPU, but it's usually not worth overclocking over it. CPU is another matter, even if you don't overclock you will get better performance by running all the cores at the same speed instead of depending on the default core boost since it's a very intensive task, and it will start clocking down eventually if your cooling isn't top-notch. Yes, even if you encode on the GPU, and no, doesn't depend on brand. On that note, have been hearing very good things about ARC's AV1 encoding potential.

 

[wade7575]

I tried adjusting what was called Core Power Limit and didn't see any real changes,I also adjusted the Voltage Offset and the Performance Boost and didn't see any changes either.

The stock setting for the Core Power Limit is 150 and that's the Maximum Watts the card can draw,the Voltage Offset and the Performance Boost were set to stock settings which are zero.

I raised the Voltage Offset to 15 and ran a test and then left that setting alone at 15 and then raised the Performance Boost from zero to 15 as well and both those settings were on 15 and this was all done with 175Watt setting on the Core Power Limit.

The temp's of the card never went above 42C

135 Watt Limit .910 to .978v Watt's the card used while testing 55 and 60 Watts 1000.23fps

150 Watt Limit .966 to .990v Watt's the card used while testing 55 and 64 Watts 1000.98fps

160 Watt Limit .950 to .980v Watt's the card used while testing 56 to 63 Watts 1004.83fps

175 Watt Limit .978 to .982v Watt's the card used while testing 56 to 61 Watts 993.97fps

1017.94fps

The test file I used was a tv show that was 960x540 being encoded to 720x404 and the times were between 1:07 to 1:10 minutes,I also used MP4 for the format and the video encoder was Intel H.264 QSV as the encoder setting as it is like a hybrid AV1,I find that the Intel H.264 QSV can compress stuff a bit more without hurting the video quality and will play just fine on any device unlike AV1.

If you look at the last fps of 1017.94 I ran the last test again after getting 993.97fps and then got higher numbers.
With the last test that result didn't surprise me at all I have seen this before if I run the same encode back to back and you'll get a different number,most times number will be 4 to 7fps difference,also with Handbrake it doesn't matter where the compression setting is either from what I have found the time it takes to encode a video is pretty much the same with 1 to 2 seconds difference in time any in change at all.

Post magically merged: Feb 18, 2025

I wonder how much faster the B580 will beat encoding I know they are supposed to have changed thing's up a bit there as well.

 

[Kenrou]

AV1 > h265 > h264 quality (especially at lower bitrates) ⇾ h264 > h265 > AV1 compression time

 

[Woomack]

I want to get an ASROCK B580 when they come out as I really like my A580 and would like to use it mainly for video encoding using Handbrake first and foremost and maybe a bit for gaming but not a lot if any.

ASRock B580 is available for nearly two months. It's just a matter of finding a store that has it in stock. At least in the EU, you can buy them without issues. However, prices are not always as low as expected.
Acer/Predator looks interesting, too, and in some stores, it's cheaper.

I don't know what's with the US availability, as all stores I just checked have them out of stock or shipping from China.

 

[rainless]

Post magically merged: Feb 18, 2025

I wonder how much faster the B580 will beat encoding I know they are supposed to have changed thing's up a bit there as well.

I have a bad feeling about that...

Arc B580 vs Arc A580 [1-Benchmark Showdown]

Comparing Arc A580 with Arc B580: technical specs, games and benchmarks.

technical.city

Faster speed but fewer cores...

Not even sure why they'd do that...

It's entirely possible that the encoding speed could be exactly the same.

Encoding video isn't the same as running Crysis or something...

 

[EarthDog]

135 Watt Limit .910 to .978v Watt's the card used while testing 55 and 60 Watts 1000.23fps

150 Watt Limit .966 to .990v Watt's the card used while testing 55 and 64 Watts 1000.98fps

160 Watt Limit .950 to .980v Watt's the card used while testing 56 to 63 Watts 1004.83fps

175 Watt Limit .978 to .982v Watt's the card used while testing 56 to 61 Watts 993.97fps

If you want to test if the core/memory clocks make a difference, you should adjust them. As you can see, lowering the power limit when the card doesn't come close to hitting that ceiling and doesn't show you anything. Drop the core speeds a couple hundred MHz and test.

most times number will be 4 to 7fps difference,

Out of 1000 FPS, that's less than 1% difference. Run variance.


Again, as far as the encoder goes, I'm not sure if it uses the cores or not (I think this varies by what type of encoding you're doing). If it does, than clock speed matters, if not, it's up to the encoder chip...but NV/AMD/Intel have their own encoding chip that's seperate from the cores.

 

[rainless]

Again, as far as the encoder goes, I'm not sure if it uses the cores or not (I think this varies by what type of encoding you're doing). If it does, than clock speed matters, if not, it's up to the encoder chip...but NV/AMD/Intel have their own encoding chip that's seperate from the cores.

Like I was saying... it's entirely dependent on the NLE you're using. I have no idea what Handbrake is doing... but the number of cores is used, in DaVinci Resolve for example, to run the entire program faster which frees up resources for... well... everything else. Which ultimately leads to the whole thing running faster.

I'm guessing Handbrake is a pure encoder and not an NLE so... WHO KNOWS!

But, at any rate, that 1% difference won't matter at all. Not even a little bit.

 

[Kenrou]

Handbrake ⇾ Difference is negligible, 1%-5% at the most depending on the brand/model if you OC the GPU+VRAM (not sure on halo models from the last 2 gens), why I said it's not worth the OC. The bigger difference would be the CPU, even if you're using GPU hardware encoding, which to be honest, is usually only good for streaming and short videos quality-wise. I've been messing around with upscaling with Neural AI and I typically use Handbrake for all my encoding even though my 3800x slogs through it (this is the one of the best cases where a *950x or a Threadripper would be most useful)...

 

[wade7575]

ASRock B580 is available for nearly two months. It's just a matter of finding a store that has it in stock. At least in the EU, you can buy them without issues. However, prices are not always as low as expected.
Acer/Predator looks interesting, too, and in some stores, it's cheaper.

I don't know what's with the US availability, as all stores I just checked have them out of stock or shipping from China.

With the ACER card I'm going to stay away from that one for sure,I know back about 12 years ago if you wanted a pre-built tower in Canada most of the big block store were selling ACER pre-built towers and after 3 to 5 year something would always be burnt out on the mobo,the saddest part about the ACER's back then is that a pre-built Dell or HP was about the price and the ACER's often had better spec and better chip's in them as well.

I worked on so many ACER's and most of them turned out to be tower's I would just salvage what I could from them and toss them into the E-Waste bin.

I could never trust ACER again unless they did a massive push on their product's and they turned out to be really good.

I think I will go with the ASROCK B580 Challenger again,it's really small runs very very cool and it only needs one PCIe cable and man is that card quiet,when I say quiet I can't even hear it run ever.

It's funny ACER is a Taiwanese company and I find their Laptops and Towers to be and if you get one that last's more then 5 years you have a rarity,but then look at Team Group another Taiwanese owned company and they make great ssd's as far as I'm concerned when it comes to price,just look at the MP44 ssd they seem to last and a friend of mine who does a lot of builds only use's their ssd's and ram and in the last6 years and has done about 175 towers and not one has come back with a bad ssd or ram.

Post magically merged: Feb 18, 2025

I have a bad feeling about that...

Arc B580 vs Arc A580 [1-Benchmark Showdown]

Comparing Arc A580 with Arc B580: technical specs, games and benchmarks.

technical.city

Faster speed but fewer cores...

Not even sure why they'd do that...

It's entirely possible that the encoding speed could be exactly the same.

Encoding video isn't the same as running Crysis or something...

With the fewer core's thing you mentioned I seen an interview I think on Gamers Nexus where an Intel Rep was on and he said that with the core's in the B Series cards they decided to focus more on making them run faster and much more efficiently.

When he said more efficiently he wasn't just talking about power draw but also how they processed task's as well.

 

[wade7575]

Like I was saying... it's entirely dependent on the NLE you're using. I have no idea what Handbrake is doing... but the number of cores is used, in DaVinci Resolve for example, to run the entire program faster which frees up resources for... well... everything else. Which ultimately leads to the whole thing running faster.

I'm guessing Handbrake is a pure encoder and not an NLE so... WHO KNOWS!

But, at any rate, that 1% difference won't matter at all. Not even a little bit.

When you say I'm guessing Handbrake is a pure encoder and not an NLE so... WHO KNOWS!

Have you never tried Handbrake before,it's a great free small program and they are from France like you.

It doesn't have as many feature's as Davinci Resolve but you can shorten the start and end time of a video with it,you can also give your video an audio gain if need and also change the resolution and compress the video more if wanted to change the mb size.

I have been using Handbrake for years and it is regarded as one of the best encoding software's out there,I know Intel said they made the A Series cards with Handbrake and to optimize to work very well with Handbrake.

 

[Woomack]

With the ACER card I'm going to stay away from that one for sure,I know back about 12 years ago if you wanted a pre-built tower in Canada most of the big block store were selling ACER pre-built towers and after 3 to 5 year something would always be burnt out on the mobo,the saddest part about the ACER's back then is that a pre-built Dell or HP was about the price and the ACER's often had better spec and better chip's in them as well.

I worked on so many ACER's and most of them turned out to be tower's I would just salvage what I could from them and toss them into the E-Waste bin.

I could never trust ACER again unless they did a massive push on their product's and they turned out to be really good.

I think I will go with the ASROCK B580 Challenger again,it's really small runs very very cool and it only needs one PCIe cable and man is that card quiet,when I say quiet I can't even hear it run ever.

It's funny ACER is a Taiwanese company and I find their Laptops and Towers to be and if you get one that last's more then 5 years you have a rarity,but then look at Team Group another Taiwanese owned company and they make great ssd's as far as I'm concerned when it comes to price,just look at the MP44 ssd they seem to last and a friend of mine who does a lot of builds only use's their ssd's and ram and in the last6 years and has done about 175 towers and not one has come back with a bad ssd or ram.

Acer is only a brand, a name, not a manufacturer. Nowadays, Predator products are made by other companies. They just put an Acer label on it. Actually, Acer and Predator RAM, storage, and graphics cards are pretty good and get good reviews (I reviewed most of their SSDs and RAM). The same OEMs manufacture SSDs and RAM for HP consumer and gaming products (I also reviewed the HP products mentioned) and Lenovo consumer storage products (those I never reviewed).
You would also be surprised by what I would say about some TG products, and I have reviewed most of their product line in the last few years (some are on the front page). The mentioned MP44 and MP44s were pretty good.

I still wouldn't touch Acer laptops, as even service centers in my country don't want to fix them (the same lower Lenovo, Toshiba, Fujitsu, and some other brands). Most of them have canceled/discontinued repair agreements with Acer and some other brands. Most countries use third-party support for laptop and monitor repairs. At work, we almost only sell Dell and HP. Sometimes, Lenovo, but only high-series models with pro support.

In short, don't always look at the brand and don't live in the past. Everything changes each year. Every brand has good and bad products. For that, it is good to have a good warranty agreement.
On the other hand, those ASRock cards get good reviews, and brands like Sparkle too ... and I have really bad memories about Sparkle products.

 

[wade7575]

Acer is only a brand, a name, not a manufacturer. Nowadays, Predator products are made by other companies. They just put an Acer label on it. Actually, Acer and Predator RAM, storage, and graphics cards are pretty good and get good reviews (I reviewed most of their SSDs and RAM). The same OEMs manufacture SSDs and RAM for HP consumer and gaming products (I also reviewed the HP products mentioned) and Lenovo consumer storage products (those I never reviewed).
You would also be surprised by what I would say about some TG products, and I have reviewed most of their product line in the last few years (some are on the front page). The mentioned MP44 and MP44s were pretty good.

I still wouldn't touch Acer laptops, as even service centers in my country don't want to fix them (the same lower Lenovo, Toshiba, Fujitsu, and some other brands). Most of them have canceled/discontinued repair agreements with Acer and some other brands. Most countries use third-party support for laptop and monitor repairs. At work, we almost only sell Dell and HP. Sometimes, Lenovo, but only high-series models with pro support.

In short, don't always look at the brand and don't live in the past. Everything changes each year. Every brand has good and bad products. For that, it is good to have a good warranty agreement.
On the other hand, those ASRock cards get good reviews, and brands like Sparkle too ... and I have really bad memories about Sparkle products.

I have gotten bad product's by other company's in the past and now by their product's but with ACER there were so many bad ones that it just left an extremely bad taste in my mouth.

With Sparkle I like the look of their cards but they don't seem to much if any support in the States or in Canada which will most likely mean the shipping overseas will cost more then what it's worth for warranty,I would trust ASROCK before Sparkle also just because they are a bigger company and also have a good rep for making good product's without a ton of problems and their stuff seems to last.

When you say (You would also be surprised by what I would say about some TG products) have you ran into some crap product's of theirs,I sure they have some crap product's but I try to watch and make sure before buying ssd's that they have been out for 1 or better yet 2 years and see what people are saying about them.

 

[Woomack]

When you say (You would also be surprised by what I would say about some TG products) have you ran into some crap product's of theirs,I sure they have some crap product's but I try to watch and make sure before buying ssd's that they have been out for 1 or better yet 2 years and see what people are saying about them.

Yes. Most are little mistakes, like the additional profile in the Xtreem CUDIMM kit (front page) - someone failed the timings list, and it doesn't work, but the declared profile works fine. However, one of the M.2 SSDs was replaced twice (as I remember, it was the G70 Pro), and it still didn't work correctly (literally 50% lower performance than expected). In the end, they sent something else. Some other websites published fake results and said that all is fine.

Recently, I received an X1 MAX flash drive. It overheats after +/- 3 minutes of load (82°C). The thermal throttling lowers the performance by ~50%, which is still ~400MB/s, so it's more than fine, but still, the product shouldn't be released acting this way. In addition, the enclosure was so hot that the glue, which was holding it, came loose after 1-2 weeks of use.

Waiting for 1+ years is not always a good idea regarding SSDs. It depends on the manufacturer, but many are changing NAND or other things under the same product name, so you may get something worse or far from the initial reviews.

What is annoying is when I find any design flaws and see that in other reviews, they are skipped, and the product gets top scores. Yesterday, I saw Biostar Z890 mobo get a top score. The mobo has problems with RAM support, overclocks worse than any other Z890 that I tested (2 ratios worse), has some problems with BIOS, and generally looks like designed 3-4 years ago. One spot on my review sample looked like it was soldered by hand. Additionally, Joe reviewed one for Tom's and the one he got burned down during initial tests. This is one example of honest reviewing nowadays.

For graphics cards, I guess the best option is to check TPU, as there are more details and many additional tests. Most other websites give top scores and award everything as marketing expects. They usually do that because they fear they won't get more samples.

 

[EarthDog]

So... how'd that testing go @wade7575 with lowering the clocks? I'm curious now!

 

[rainless]

When you say I'm guessing Handbrake is a pure encoder and not an NLE so... WHO KNOWS!

Have you never tried Handbrake before,it's a great free small program and they are from France like you.

It doesn't have as many feature's as Davinci Resolve but you can shorten the start and end time of a video with it,you can also give your video an audio gain if need and also change the resolution and compress the video more if wanted to change the mb size.

I have been using Handbrake for years and it is regarded as one of the best encoding software's out there,I know Intel said they made the A Series cards with Handbrake and to optimize to work very well with Handbrake.

I guess I'll look into it. I happen to be working on a Kickstarter video at this very moment...

Like RIGHT now... (otherwise I'd get into the importance of "cores" vs "clock speed" in NLEs...)

 

[wade7575]

If you want to test if the core/memory clocks make a difference, you should adjust them. As you can see, lowering the power limit when the card doesn't come close to hitting that ceiling and doesn't show you anything. Drop the core speeds a couple hundred MHz and test.


Out of 1000 FPS, that's less than 1% difference. Run variance.


Again, as far as the encoder goes, I'm not sure if it uses the cores or not (I think this varies by what type of encoding you're doing). If it does, than clock speed matters, if not, it's up to the encoder chip...but NV/AMD/Intel have their own encoding chip that's seperate from the cores.

I don't know if you can adjust the core/memory clocks on this card,I don't know if you can tell by the picture if you can,that a screen shot of the only spot on my Intel Control Panel where you can adjust things,I tried adjusting all 3 before when I did my first testing,the middle slider you can raise or lower the value but the top and bottom slider can only be increased.