Anyone planning any new hardware purchase...

[Joe DeFuria]

Is it a better SSE2 implementation or is it just that the clockspeed is higher? I was under the impression that the higher clockspeed made the difference in video encoding and not the actual design of the SSE2 bits.

You can't really separate the two like that. I mean, "clock for clock" the Athlon certainly does much more. That's just the overall design implementation.

The bottom line is, that for a given speed "rating" (and similar price points), the P4 tends to offer better encoding performance. It may even be due to external memory bandwidth for all I know.

 

[Johnny Rotten]

Oh, Id also like to go with that dell 20.1" lcd. God that looks beautiful.

I am very pleased with it. No dead pixels either!

How happy are you with the hue/saturation performance? Everything look good or is there a slight blue push? Are you able to notice the 'grain' of the panel (ie some people are saying the picture sufferes from a graininess due to a relatively large pixel grid - you can see the lines between the individual pixels rather easily)?

 

[olivier]

i prefer amd for gaming , but p4 for general usage .. i really like hyperthreading when im compilling, encoding or doing anything who take 100% of the cpu and my system seems smooth.

 

[Tim Murray]

You can pretty much say that the P4 is always on top in video encoding because of its better SSE2 implementation compared to the A64.

Is it a better SSE2 implementation or is it just that the clockspeed is higher? I was under the impression that the higher clockspeed made the difference in video encoding and not the actual design of the SSE2 bits.

I'm trying to remember what was on the Inqy and why SSE2 performance seemed so low on A64s... I think the theory was the A64's SSE2 registers were 64 bits instead of 128, and so it offered no speed improvement (and sometimes slowed things down) over SSE alone? Something like that.

Either way, Intel > AMD for video encoding, any way you slice it.

 

[Vortigern_red]

Either way, Intel > AMD for video encoding, any way you slice it.

The bottom line is, that for a given speed "rating" (and similar price points), the P4 tends to offer better encoding performance. It may even be due to external memory bandwidth for all I know.

The above quotes are taken as given but how many video benchmarks do you actually see?

I thought about this for ages and considered buying a P4 in addition to my athlon (2500+ clocks upto 2.3 400fsb) for video testing and was staggered to see the cheapest P4 I could buy was about £130 (I paid about £70 for my 2500 and £60 for my 2200 before it). I decided in the end I could not afford my experiment (wife!) but I was considering trying all the popular video programs (I own some but for others I would have to use demos) with different codecs and seeing what the results were on both platforms, perhaps posting them online.

I do quite a bit of (home not pro) video work with MSP and it seems to me that all the assumptions about video encoding speeds (mine included at one time) come from the DivX encoding with flask or maybe Xmpeg benchmarks shown at popular websites.

I've just looked at a few websites athlon 64 3400 reviews and found a few with video benchmarks.

http://www.tomshardware.com/cpu/20040106/athlon64_3400-20.html

This is really interesting and is what made me think about the "P4 is better for video" in the 1st place (on an older review). This is a very popular program with home users and whilst the P4 is faster at the top end (by a small amount) I feel that if you compare the lower/middle end the athlon 3000+ comes out the same as the P4 2600 HT/200 (800)fsb and faster than the P4 2800 533 non HT (the difference with HT seems to be about 5% with this test)

The interesting part of this is why I have bolded part of joe's quote above the Athlon 3000+ in the UK currently costs about the same as the P4 2400 HT 200 (800) fsb or the P4 2.6 non HT 133(533) fsb ( based on a quick look at one of my regular suppliers) and in this app at least the athlon offers the best price/performance at the near top rating (and the 2800+ cost less than £100 or just over if retail)

This ones good from extremetechs 3400 review.


http://www.extremetech.com/article2/0,3973,1426436,00.asp

this shows the athlon 64 3200 comparing well to the P4 3200 except in after effects and the 64 3400 beating the P4 in all tesst except after effects.

This ones from [H] even the athlonXP here beats the P4 3200 on DivX encoding (don't ask me how!)

http://www.hardocp.com/article.html?art=NTc1LDI=

This is Anands only video benchmark:

http://www.anandtech.com/cpu/showdoc.html?i=1941&p=9

and they are pretty pro-AMD as well! but its this comment at the bottom which annoys me.

Hyper-Threading along with other features of the NetBurst architecture give Intel the performance advantage in video encoding as you can see by our DivX results above. The Athlon 64 3400+ performs respectably but it isn't the CPU that's best for these sorts of tasks.

They have decided the whole video encoding performance of these chips on one benchmark ARRRRGGGHH!!

The point I have been trying to make is that a lot depends on what app and codec you are using for video, for instance I never use DivX (although I do remember all night encodes with flask on my K63-450 LOL) and it appears that the popular benchmark is DivX but it may bear no reletionship to YOUR video work.

My feeling is (to all you 3D guys out there) that we have come about this impression in much the same way as if we just used one benchmark for graphics cards.

I am not saying that the AMD chips are better/faster for video just that the diffrence may not be what you think in the situations you use your PC.

Its taken me about an hour to post this and I would really like to hear peoples views. Do you think I have just come across Pro-AMD benchmarks? Do you have more comprehensive video benchmarks that disagree with the ones I've posted?


Sorry for the long post, I'm going to post my new system specs in another On-topic post in case people don't read all of this

 

[Vortigern_red]

Now for my upgraded system:

SKT 939 with prob the cheapest (or only) Athlon FX
Best motherboard available at the time.
ATI R420 (fave) or NV NV40 (if it is very much better)

I already have RAM ect and have been considering an athlon 64 3000+ but I think its prob worth waiting for SKT 939 and hopefully I can get a new vid card at the same time. I currently have a R9700 non pro at 360/315 and this is the first time ever I have not been wanting to upgrade my Video card within 2 months, I still have it now 12 months after I bought it and it is still one of the best cards you can buy! IMO! The next Gen cards will have to be something special but I am hoping to upgrade the whole lot at once and I will move to PCI express and BTX next birthday

 

[Bouncing Zabaglione Bros.]

Its taken me about an hour to post this and I would really like to hear peoples views. Do you think I have just come across Pro-AMD benchmarks? Do you have more comprehensive video benchmarks that disagree with the ones I've posted?

As a general rule, the P4s are better for applications that are dependent on lots of memory usage (due to the high FSB), but the AMDs are better for computational use (due to the larger number of instructions per clock). Of course this is a difficult distinction to make because so many apps use a combination of both memory and computation.

My understanding is that the P4 is good for video *editing* because of the high memory usage allowing the high FSB to come to the fore. For editing you are copying large streams of data in and out of RAM & hard disk.

However, when you are looking at something like DivX *encoding*, memory speed plays less of an influence, as the bottleneck (and the heaviest workload) takes place in the CPU while calculating the encoding on every frame.

So you really need to decide what kind of work you are going to do, and decide if you need more CPU power, or more memory access, and choose a system accodingly.

 

[epicstruggle]

Vortigern_red, great post. Im suprised AMD has tried to encourage developers to create encoding software that takes advantage of AMDs strength. Im not into encoding so maybe they have? Anyone know for sure?

later,
epic

 

[Vortigern_red]

BZB,

Where did you get that opinion from?

Have you tested both chips in popular video apps or do you know of somewhere reliable (IE the B3D of the video world) that has?

Some links would be really great.
Thanks.

(due to the high FSB)

I'm not being funny but they don't have a higher FSB. The Athlon has the same FSB (200 mhz) and the Athlon64/64FX has a FSB that runs at the same frequency as the chip (between the CPU and memory controller, or a hypertransport link to everything else at what I though was equivilent to the P4)

EDIT

My understanding is that the P4 is good for video *editing* because of the high memory usage allowing the high FSB to come to the fore. For editing you are copying large streams of data in and out of RAM & hard disk.

Sorry I sort of missed this. The amount of video that can be stored in ram in even my small projects is very small (with only 1gig of DDR400) and I find that hard drive speed is far more of a bottle neck.

 

[Vortigern_red]

Vortigern_red, great post.

WOW, thanks. I'm really glad I made the effort now.

 

[Bouncing Zabaglione Bros.]

BZB,

Where did you get that opinion from?

From general reading up when I built my last system. You have to do a lot of reading around to get an opinion and find things out (like parts of the P4 internally seem to be running at much lower than the actual CPU speed).

Have you tested both chips in popular video apps or do you know of somewhere reliable (IE the B3D of the video world) that has?

Some links would be really great.
Thanks.

If I had them, I'd give them, but I don't think you'll find anything than a few video creation benches hiding in amongst some office type benchmarks. I don't think anyone has done the kind of specifics you are after.

(due to the high FSB)

I'm not being funny but they don't have a higher FSB. The Athlon has the same FSB (200 mhz) and the Athlon64/64FX has a FSB that runs at the same frequency as the chip (between the CPU and memory controller, or a hypertransport link to everything else at what I though was equivilent to the P4)

P4s are now shipping at 800 mhz FSB - that's faster than AMD, and is the only thing that is keeping P4s competative. P4s do less instructions per cycle than AMD, but they get more cycles in per second.

My understanding is that the P4 is good for video *editing* because of the high memory usage allowing the high FSB to come to the fore. For editing you are copying large streams of data in and out of RAM & hard disk.

Sorry I sort of missed this. The amount of video that can be stored in ram in even my small projects is very small (with only 1gig of DDR400) and I find that hard drive speed is far more of a bottle neck.

Hard drive speed is actually the bottleneck for any system sooner or later, which is why so many video editing systems go for RAID, but you'll find that manipulating chunks of data in and out of memory should be faster on a P4, at least until AMD-64 gets full support.

If your primary use will be video editing, then a couple of fast disks running in RAID will make a more significant difference to your productivity than the difference between a P4 or AMD. If your system spend the majority of it's time reading and writing to disk, then making that faster will have the biggest impact.

While there's only so much video you can fit in RAM, halving the ram (also affecting your disk cache) or lowering the speed of getting data in and out of RAM will have a significant impact when all you are doing is basically reading and writing video data from RAM/disk.

 

[Vortigern_red]

I think:

1: you are assuming I know less about hardware than I do.
2: you are basing your opinions of P4 video performance on the very same factors I mention as being misleading in my first post.

for instance:

P4s are now shipping at 800 mhz FSB - that's faster than AMD, and is the only thing that is keeping P4s competative. P4s do less instructions per cycle than AMD, but they get more cycles in per second.

P4s currently run at 200mhz FSB speed!! This is what I already pointed out (I was not making it up ). You are confusing the quoted (maximum) data rate and FSB.

Its not the only thing thats keeping P4s competitive IMHO its a massive clock speed advantage and P4 optimized software, if your app is coded with the P4 in mind you can minimise the effects of its long pipeline.( I've have always presumed that apps optimised for SSE2 will also have taken the time to make more general P4 optimisations?) However the optimisation WRT video encoding seems to be less than the general perception with certainly some of the Apps/codecs.

While there's only so much video you can fit in RAM, halving the ram (also affecting your disk cache) or lowering the speed of getting data in and out of RAM will have a significant impact when all you are doing is basically reading and writing video data from RAM/disk.

You keep putting ram/disk as if they are one, but in the context of the current Intel/AMD conversation. They are not. Assuming that the two processors access the same amount of ram and have the same hard disk. The difference in ram/disk subsystem then becomes just the ram bandwidth. The current fastest Athlon (FX/Opteron) and P4 (EE/3.2) both have the same (maximum) memory bandwidth (2x DDR400). But keeping to "normal" chips the Athlon64 has about half the memory bandwidth of the P4 but does it make a huge difference in real life video work? I was trying to point out that I find that 1gig of ram very quickly becomes a bottle neck when you are dealing with files 10s of gigabytes. (or even scrolling through a (compressed) mpeg2 in order to put in chapter points)

This is what I'm interested in. The P4 has this "Video Processor" image based, as far as I can see, on rendering benchmarks but as I've pointed out (with links) this is perhaps not such a huge lead in some apps/codecs as people belive.


If you have some non-rendering benchmarks or personal use to back up your claims please link /state.

thanks

 

[Tim Murray]

A64s are too new to really change the perspective of the P4 as the Video Processing Lord and Master. Also, the constant flux, both for A64 motherboards and A64s themselves (with regard to the socket), makes going AMD an iffy proposition at this point. Plus, you have to get a Via board if you want full HyperTransport performance, and I know a lot of people don't like that at all; after all of the problems with their Athlon motherboards, it makes people go "Eeeehhhhh...."

 

[Vortigern_red]

the P4 as the Video Processing Lord and Master.

I know I'm repeating myself LOL but why do we all have that impression?

http://www.tomshardware.com/cpu/20030210/barton-16.html

This is a good chart with no athlon64. Both video encoding benchmarks show the P4 3.06HT outperforming all other chips by quite a margin. But then there is a whole load of athlon chips with the second fastest P4, the 2.8, coming in at around the XP2400/XP2500 mark.

This will have changed somewhat with the introduction of new chips(200mhz FSB P4, athlon64) but the point I want to make is that during the time these graphs were produced we all assumed (me included) that the P4 was just a faster video encoding chip "one of its strenths" so to say. But in reality only the most expensive P4 was (more than) competitive with the Athlon.

I'm not saying the Athlon is better choice for video encoding (chipsets are a good point) but that in many cases the Athlon is competitive with the P4 whereas every man and his dog thinks that all video encoding runs like those flask/DivX benchmarks.

 

[Bouncing Zabaglione Bros.]

You keep putting ram/disk as if they are one, but in the context of the current Intel/AMD conversation. They are not. Assuming that the two processors access the same amount of ram and have the same hard disk. The difference in ram/disk subsystem then becomes just the ram bandwidth. The current fastest Athlon (FX/Opteron) and P4 (EE/3.2) both have the same (maximum) memory bandwidth (2x DDR400). But keeping to "normal" chips the Athlon64 has about half the memory bandwidth of the P4 but does it make a huge difference in real life video work? I was trying to point out that I find that 1gig of ram very quickly becomes a bottle neck when you are dealing with files 10s of gigabytes. (or even scrolling through a (compressed) mpeg2 in order to put in chapter points)

Your one gig of ram is not the bottleneck. The bottleneck is when your ram is not enough and you go to disk access. Sure, if you had enough ram to hold your whole movie (in addition to OS tasks) you would never touch disk and it wouldn't be an issue, but this isn't the case.
As soon as you start disk access (as you do almost all the time when video editing) then disk access *is* the bottleneck, significantly so, as this is what you are spending the majority of your task doing.

You can *prove* this by running tests on a machine with a single disk, and then again using a multi-disk RAID setup. This is why I said a fast disk setup will make more of a difference to a task that involves constant disk access than the difference between P4 or AMD.

Trying to equate disk access with memory speed misses the point - your task is completly disk-bound, and improving that is what will give you massive speed increases. Working out which CPU gets bottlenecked slightly faster than the other before it has to wait on disk access doesn't make much difference to your task. Increasing the speed at your bottleneck point will make a big difference.

If you have some non-rendering benchmarks or personal use to back up your claims please link /state.

thanks

See my previous message. I haven't seen anything specific for what you're asking beyond some content creation stuff. You have to look at other performance figures and extrapolate towards what you want to do with your system. Unless you can run such benchmarks yourself and share them with us.

If you do want my anecdotal advice (I used to work on large, disk bound applications, on a much larger scale than PCs), the speed of the memory or CPU is very much secondary to the speed of the hard drives. If your tasks use the hard drive extensively (as video edititing does) than this is where your bottleneck is and this is where you can make the most improvements.

 

[epicstruggle]

If you do want my anecdotal advice (I used to work on large, disk bound applications, on a much larger scale than PCs), the speed of the memory or CPU is very much secondary to the speed of the hard drives. If your tasks use the hard drive extensively (as video edititing does) than this is where your bottleneck is and this is where you can make the most improvements.

What you say is true enough, thats why I think going for the best cpu based on price/performance is the smart thing to do. And i think most would agree AMD has the best price/performance numbers. And once we see win64 out and programs compiled for amd64 we might see another few percentage point gains.

later,
epic

 

[Joe DeFuria]

This is a good chart with no athlon64.

Um, without the A64, it's not a very good chart. (It's a year old.) No 4x200 FSB Intel chips?

Here's a recent one:

http://www.anandtech.com/cpu/showdoc.html?i=1946&p=5

Unforuntately, there's only one comparison at the bottom of the page for media encoding. I'm not interested in any "exotic" chip like the P4 EE or FX 51. I'm budgeting aout $300 for my CPU.

The A64 3400+ simply lags behind even the P4 3.0 Ghz. Roughly 20% behind the P4 3.0 Ghz.

 

[Joe DeFuria]

Some additional recent benchmarks:

http://www6.tomshardware.com/cpu/20040106/athlon64_3400-19.html

Start on that page and continue on. Note NON OVERCLOCKED performance. (The AMD 64 3400+ non overclocked scores are in Red) and compare to the P4 3.2. Some of the differences are slight, and some are very significant.

 

[Vortigern_red]

I posted the older benchmarks because the baron did not want to see benchmarks that contained the athlon64! and to make the point that even then (before mass availability of HT) the P4 had this "video encoder king" rep which was all based on one chip!

In the benchmarks you posted there are differences between the two platforms but not as great as is often assumed. The movie maker benchmarks are what a lot of people imagine the difference is from the old xmpeg/divx charts which looked very similar.


Wrt to the anand benches I have already posted a link to those myself (in anands part 1) and I concede that if you are using Xmpeg for DivX the P4 is faster, I have said that loads of times.

Whilst the difference is 20%ish in that benchmark the toms pinnacle and mainconcept charts show a difference about 10% and 2%. Which is my point really. It depends on what video work you are doing.

Of course Anand and Toms both have Xmpeg/DivX benches and both use different setting which changes the rankings somewhat. Tom only gets 5% difference!. This is where it gets fun! LOL In the {H}(i did not mean to give Kyle curly brakets there but now I think they look so much nicer ) review I linked to before he has the 3400 20% faster than the P4 and even the XP3200 is almost 5% faster!!! encoding DivX.

That is not using Xmpeg but again it shows what app you use (and settings) can have a huge difference on both chips.

I think I'll give up now, I only wanted to point out that often the "video Benchmarks" at some places are very limited and they make grand claims of one benchmark (Anands). And also, for me, the rendering time is a very small part of producing a video, as long as I get acceptable preview performance and I can make a cup of tea while rendering. I'm a happy guy.

Edit: I meant to add that obviously the Movie maker2 benchmarks are really good on the P4. So if anyone is considering doing A lot of work with this App you will be better off getting a 3.2 P4 of perhaps treat yourself to an EE.

 

[Tim]

A64s are too new to really change the perspective of the P4 as the Video Processing Lord and Master. Also, the constant flux, both for A64 motherboards and A64s themselves (with regard to the socket), makes going AMD an iffy proposition at this point.

Why? Even considiring the move to Socket 939, the upgrade path is still better with Socket 754 than it is with Socket 478.

Plus, you have to get a Via board if you want full HyperTransport performance, and I know a lot of people don't like that at all; after all of the problems with their Athlon motherboards, it makes people go "Eeeehhhhh...."

The real-life performance difference between the Via K8T800 and the nVidia nforce3 150 is non-existing so the less than full speed HT-connection with the nforce3 150 does'nt really matter. Plus, the Via K8T800 seems like a very solid solution.