Can anyone recommend an up-to-date site that provides benchmarks, or similar quantitative data, that allow comparison of the various P4, Athlon64 and Opteron systems for digital media editing/encoding?

Thanks.

Hi Graham,

Your best bet is to probably do a google search. Or by Computer Video magazine.

I remember seeing a review in CV where they pitted a G4 apple, against the Athlon 64 and the P4 3.0GHz systems. When I'm at home I'll try and dig out the review, and tell you what issue it was in.

Cheers,

do a google search on "overlcokers au" thre is a whole forum dedicated to tweaking your machine :)

Peter I take it you mean 'overclockers'?

Thanks guys. I realize the info is indeed all out there in scattered form - google-able etc - but I was hoping there might be an updated version of something like this:

http://www6.tomshardware.com/cpu/20030217/cpu_charts-26.html

anandtech.com and xbitlabs.com are usually good for benchmarks, but their methodology isn't always the best, they're wrong sometimes, and what they benchmark usually has little to do with video editing (except for the main concept encoder test).

2- What exactly are you trying to do? I know you are looking for benchmarks, but you must be looking at benchmarks for a reason.

Hi Glenn,

I currently have an overclocked Barton 2500XP+ used for editing and rendering DV, and MPEG2 for DVD.

Just bought an HDV camera which needs quite a bit more grunt, (in fact Cineform's Aspect HD package which I'm considering currently doesnt run on Athlon XP). Cineform recommends 3GHz P4 as 'minimum spec', and dual Opteron as 'preferable'.

Since there's a considerable price gap between those two options I'm wanting to get a feel for likely cost:benefit of some intermediate-price alternatives (single Opteron, 3.4GHz P4, Athlon64 etc).

Tech Report seems to be on top of the latest info.

Latest benchamrks of current AMD vs. Intel shows pretty much an even playing field, with AMD having a number of near future advantages that Intel does not.

http://techreport.com/reviews/2004q2/opteron-x50/index.x?pg=1

Faster at MPEG encoding, Faster in 3DS Max, excellent memory bandwidth.

And this is without a NUMA aware OS or the benefits of 64 bit.

WinXP 64 gives AMD an even greater advantage.

Considering the similar cost and performance parity, there is littel reason not to go 64 bit.

Actually, there are reasons not to go 64-bit right now, but not to buy a processor that is capable of 64-bit. First amoung the reasons of not going 64-bit is that XP has not been released as a 64-bit OS. Yes, you can run the beta, but MS does not officially support it. Second is that any current 32-bit applications may run slower on a 64-bit OS because there is an extra application layer that has to convert all of the 32-bit addresses into 64-bit addresses (thunking).

XP 64 may be released this winter and Intel is releasing the 800MHz bus Xeon, Nacoma, which should be on the streets in August. There are interesting times going with the new processors and motherboard chipsets that may again level the playing field between the Xeon and Opteron.

<<<-- Originally posted by George Ellis : Actually, there are reasons not to go 64-bit right now, but not to buy a processor that is capable of 64-bit. First amoung the reasons of not going 64-bit is that XP has not been released as a 64-bit OS. Yes, you can run the beta, but MS does not officially support it. Second is that any current 32-bit applications may run slower on a 64-bit OS because there is an extra application layer that has to convert all of the 32-bit addresses into 64-bit addresses (thunking).-->>>


That is absolutely incorrect. Applications do NOT run slower on 64 bit and there is no conversion of 32 to 64 bit invvolved at all. Although your assessment would be correct for Itanium processors, which cannot run 32 bit apps without some interpolation.

Opteron and Athlon 64 bit processors run 32 or 64 bit applications natively.

The fact is, there is EVERY reason to buy 64 bit now. Changes in the general architecture and design of the 64 bit chips makes for a superior 32 bit design.
A lot of studios run Linux, and there are 64 bit Linux distro's for Opteron.

WinXP 64 does have a hard street date, and it is coming up well before the end of the year.

In almost all 3D animation applications, Opteron is faster than Xeon or P4, including the EE edition of the P4.

As well as MPEG encoding, Audio applications, Mental Ray and quite a few other applications.


<<<-- XP 64 may be released this winter and Intel is releasing the 800MHz bus Xeon, Nacoma, which should be on the streets in August. There are interesting times going with the new processors and motherboard chipsets that may again level the playing field between the Xeon and Opteron.-->>>

The problem with the Xeon and even with their newer 800mhz FSB is that it is still slower and has a number of design flaws that reduce it's efficiency. First off, even at 800mhz, Intel's bus is still slower than the bus Opteorn uses. Also Xeon has to rely on a separate memory controller which passes it's data to fand from the CPU over this 800mhz bus.

Opteron has an integrated memory controller on the the CPU's die making for less latency.

Opteron's bus is called Hypertransport. A direct link that avoids secondary controllers. It allows memory to operate at CPU speeds. Hypertransport is 800mhz - 1ghz. speed.

In most applications, Opteron is faster. Opteron is slightly faster at MPEG encoding since the 250 models came out last month. On the few tests Intel does win, it isn't by very much. Basically it is a generally level playing field. the choice is not a clear brand X is better. There are now more subtle technical points to consider.

And a final thought on 64 bitness, Intel scrapped almost every 32 bit CPU plan they had in the last month or two so that they could devote all their time and attention on converting Xeon into a copy of the Opteron. See HP's server page on 64 bit Xeons for proof of that. Xeon is using AMD's complete x86-64 implementation, although that is largely due to Microsoft dictating that they would not support two 64 bit branches.

As one of the administrators' of this site I am responsible of remarks posted in the forums. Potentially libelous remarks need to be backed up with a link to an article that provides some authentication to your claims. If this can't be provided, postings may be edited to protect the administrators and wranglers from legal liabilities for false claims.

<<<-- Originally posted by Steve Rogers :
That is absolutely incorrect. Applications do NOT run slower on 64 bit and there is no conversion of 32 to 64 bit invvolved at all. Although your assessment would be correct for Itanium processors, which cannot run 32 bit apps without some interpolation. >>>
This is incorrect. 32-bit applications run in the Windows on Windows subsystem, which handles the address translations.

As for performance, see this review and comparison.

http://www.anandtech.com/systems/showdoc.html?i=1961

WOW Implementation Details including descriptions of the thunking layers
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/win64/win64/running_32_bit_applications.asp

<<<WinXP 64 does have a hard street date, and it is coming up well before the end of the year.>>>

There is no hard street date for the OS. It is projected for the end of the year. MS reserves the right to change that.

<<<Opteron's bus is called Hypertransport. A direct link that avoids secondary controllers. It allows memory to operate at CPU speeds. Hypertransport is 800mhz - 1ghz. speed. >>>

While this is true, Hypertranport does not really gain a strong benefit until more than 2 CPUs are added.

***EDITED by GE per wardens note*** ;)

Additional reading -
Why XP 64 will be good (includes NLE benefits)
http://www.microsoft.com/WindowsXP/64bit/evaluation/overviews/extended.mspx

Sorry for the threadjack Ed.

Direct quote from the almost 6 month old article you provided a link to:


"It is far too early to reach any conclusions in this area, but there is a lot of driver optimization to be done to make up this kind of delta. With the CPU and memory providing faster 64-bit performance, we have to believe the drivers play a big part in this disappointing gaming performance."


In everything BUT gaming, 64 bit was an improvement, not a detraction.

So I say again, there is no reason not to buy a 64 bit based system.

Even if you only use a 32 bit OS on it, you still get equal or better performance to the top 32 bit CPU's out there at comparable prices.

There is no reason to go 64-bit *right now* except for the slightly elevated price. And I agree totally on Steve with this, the problem is that you won't be exactly getting your money's worth or the true capabilities of your hardware until the software is coded to run in a 64-bit environment natively.

Til then you won't be seeing the REAL numbers that these 64-bit babies should be crunching out. And the optimizations that these new processors have somewhat already addresses the shortcomings of the older processors such as on-chip memory controllers for a bigger bandwidth etc. which is enough reason to go 64-bit. Just reiterating what Steve is saying because I agree with him wholeheartedly.

64-bit is the next logical step to an already aging 32-bit environment.

When 64-bit becomes a reality processors might be 2X faster, in which case it's upgrade time :D

IMO the Pentium is a slightly better buy right now as it generally averages better in benchmarks, like the Main Concept MPEG2 encoder. However, the AMD64 processors are pretty damn close!

If you consider the motherboards, the Intel 865PE side is slightly better bang for the buck than AMD offerings. For example, the Abit IS7 is a nice board.

Both choices are running very close though. I don't think you can really go wrong with either.

2- For pro audio, AMD64 is definitely the better choice. It shows that things really depend on what you do.

<<<-- Originally posted by Glenn Chan :
IMO the Pentium is a slightly better buy right now as it generally averages better in benchmarks, like the Main Concept MPEG2 encoder. However, the AMD64 processors are pretty damn close! -->>>

Have a link? last few batches of reviews I have seen have AMD easily taking the majority of benchmarks.

sauch as this one:

http://techreport.com/reviews/2004q2/opteron-x50/index.x?pg=1

<<<-- If you consider the motherboards, the Intel 865PE side is slightly better bang for the buck than AMD offerings. For example, the Abit IS7 is a nice board.-->>>

Really? I didn't know that Intel started integrating SATA and Lan into their southbridge chips freeing up the PCI bus? I wasn't aware they had managed that yet.

I went and looked up the IS7, it does look like a decent budget board. A bit limited I thought though, max 2gb ram, only 2 SATA ports, 10/100 lan only, no hardware firewall, but I notice that it is only around 65.00

But in comparison I don't see any Intel offering that matches either the AMD chipset or the nVidia nForce3 250gb chipsets.

stability wise, they are on a par. AMD is just as capable of 24/7 operation as any Intel based system. And I have seen absolutely nothign to indicate that one is any better than the other. Although AMD offers more standard features.

Steve, the 865 and 875 southbridge also supports SATA RAID. My board, an ASUS PC-DL has both an ICH5R (Intel SATA RAID 0/1) and a Promise FastTrak 378 RAID (0/1) controller. Most of the 865/875 boards will host 1GB Ethernet off of the southbride. Usually it is Intel, but some mfgrs are using the Broadcom chipset. Since it is off of the southbridge instead of through the PCI controller, it does not have the bandwidth issues that it could.

The ASUS P4P800 boards are good too. If you want to look at a fairly reliable guide with prices, check out the weekly guides at Anandtech http://www.anandtech.com/guides/index.html. They cover entry, mid-level, and high-performance systems. These sheets cover both AMD and Intel offerings with current street prices for the components.

If you want to see benchmarks, they are spread out throughout the pages, but CPU ones are in the CPU section http://www.anandtech.com/cpu/index.html

Thanks George I'll take a look at those. I thought that most of that stuff was all off the southbridge so adding that info will certainly help.

Although with Intel changing their architecture to x86-64, I will probnably wait and se what board changes go along with that. According to the information available, they are apparently goign to try to release x86-64 before the end of the year, and that is coming pretty quickly.

Steve, check the review on Anandtech's main page today. The initial report is only 1%. I do need to complain to the reviewer though as the used a Northwood on the 478 and a Prescott on the LGA775 sockets. Not a like to like comparison.

Interesting article, but somewhat disappointing results all things considered.

In general, the 3.6 Intel didn't fare all that well to the FX-53 unless you consider that many of the benchmarks gained a 1 second or less lead.

It did take some benchmarks that AMD had been winning lately like mpeg encoding, but at an extremely high cost difference and not by any decisive amount. Personally I am glad that the actual speeds between the two are almost identical.

New motherboard, new very expensive memory, new case design. That 3.6 Prescott is going to be a very pricey upgrade.

Generally I don't like to comment on unreleased products or products that are coming soon if you can't buy them. Try buying a 3.4ghz Prescott. They are still listed as pre order only.

I purposely did not give reference to the FX-55 chips since those are only recently given out to reviewers and are not available in retail yet.

I'l have to go back and double check the figures, FX-55 was pretty much a clean sweep of the benchmarks comapred to the 3.4 and lower. I'll have to see how that stacks up against the 3.6 benchmarks.

Well I ended up going for a cheapo P4 system to tide me over until the dust settles a bit more on the relative merits of Athlon64 and the new intel chipsets, etc, etc .... I picked up a reconditioned ASUS P4S800D with SiS 655FX chip for $45 at Newegg for its SATA and RAID and I dont need GBethernet or onboard firewire. The SiS chip benchmarks well versus the Intel 625 for video stuff.

I'lll run it with an O/C P4 3.0 Northwood and a RAID pair of Hitachi sata drives for video capture playback. Will see how I go for heat.

Thanks for the .... err... vigorous debate. All points made were appreciated!

Sounds great Graham.

Steve, the techreport article you linked to (http://techreport.com/reviews/2004q2/opteron-x50/index.x?pg=6) did not contain many relevant benchmarks for video production.

There were 3d benchmarks, which may be relevant to some people who use those programs.

There was a benchmark for divX encoding (XMPEG), but that does not really relate to video production as few people will encode divX. The main concept MPEG2 encoder is a more useful benchmark as it is found in many mainstream MPEG2 encoding products (i.e. Vegas+DVD, ?encore?). Pentiums have an edge here AFAIK.

As far as NLEs goes, I have seen few benchmarks comparing them. Benchmarks exist for Vegas and After Effects. Don't know about Avid and Premiere Pro, although Pentiums and Xeons are recommended by the manufacturer.

My understanding is that the Mainconcept encoder is one of the first apps to implement the SSE3 instructionset .... so it may be most relevant as a benchmark for forthcoming, rather than current, video applications.

Hi Graham,

Don't know whether this is relevant any more. But in the April 2004 Edition of Computer video magazine they done a shoot out between, a dual Mac 2GHz, Dual Xeon 3.06GHz and Dual Opteron 2.2 GHz. All tests were done with editing in mind. I wont go into the details but this is what scores they gave them:

Mac: 84%
Opteron: 84%
Xeon: 80%

Marks were given on: Performance, features, ease of use, desireability and value for money.

Each having their merits and pitfulls. You might find the article on CV's web site www.computervideo.net.

Cheers,

Most of us who buy a computer this summer are going to have to live with it for a couple of years. Taking that into consideration is it wiser to buy a 64 bit processor for that reason alone? If one can wait til August would there be any advantage to buying the Intel version then?

PS--only slightly off topic, does anyone know why processor speeds have stalled out at a little over 3 ghz for about 18 months now? Based on the rate we were going 2 years ago shouldn't we be up to 5ghz or more?

64-bit computing allows larger numbers than 32-bit. You get numbers than can go up to 2^64 (that's HUGE) instead of 2^32 (also huge).

For nearly all tasks 64-bit computing is not necessary. For video it might be useful if you need slightly more rendering quality. Using large numbers can help avoid banding on things like color correction. Right now I don't think it's much of a problem at all as long as your NLE uses 32-bit floating point numbers for rendering.

On the audio side, any good audio app with use 32-bit floating point numbers for rendering audio. One Waves plug-in uses 48-bit float point numbers, which could benefit from 64-bit computing.

On the Pentium side, if you get the Prescott processors, they have the potential for gaining speed since the Prescotts support the SSE3 instruction set. I'm not too sure what it does exactly but it seems to be potentially faster for any video rendering involving floating point numbers. BUT programs have to be written to take advantage of the SSE3 instruction set.

The Prescotts are not a good buy right now IMO as they run hot and consume more electricity (which really adds up) compared to the last generation Canterwood processors. If a Prescott consumes 20W/hour more than the Canterwood processors (20W is not a very accurate figure) then it ends up as ~60W/h on your energy bill (PSUs lose a lot of electricity on the AC-DC conversion). Over the lifespan of your computer this will add up. On a side note, you may wish to consider a high efficiency power supply (i.e. Seasonic) as those will cut your energy bill for your computer tower in half (the monitor is not powered by the power supply, so ignore it).

Most of the Pentium processors out right now support hyperthreading, which gets 1CPU to do the work of 2. Right now the performance difference ranges from roughly -5% to +50% (averaging around 15-20% boost for most programs). As programs are written to take advantage of hyperthreading you will see the performance boost increase.

Betsy,

This link has some info on why MHz speeds are not increasing as rapidly these days, and why processors are starting to waste so much power:

http://arstechnica.com/news/posts/20040620-3907.html