Intel vs. AMD for Pro Tools

[sunburst79]

Originally Posted by mjslakeridge

Howdy Scott,
The information you provided here is exactly what I was looking for in the discussion. I don't fully understand all of the technical points, like sharing a single floating point unit, or power gating, but the L2 and L3 cache size makes sense to me. I looked up the cache size for my Phenom II 1045T (3,072 Kbytes and 6,144 Kbytes) and my FX 6300 (6,144K and 8,192). The Phenom II outperforms the FX 6300 in the 2 tests I have run-Super Pi and Passmark CPU test. The best Passmark CPU score I could get with the FX 6300 was 3,875 compared to Passmark average of 6,541. Must be something about my specific setup. The Phenom II got a score of 4,383 compared to the avg of 4,889.

I'm not going to worry too much about scores, both processors work fine with my PT setup. I may put the Phenom II back in the PT machine as it seems to run somewhat cooler than the FX. Up until about 18 months ago I was still using the Intel Pentium D (dual core) that Sweetwater put in the original build. Then the motherboard failed, so someone gave me a used mobo that i put a new Intel Core 2 Duo E 4400 @2.00ghz. That was a big upgrade. Went with the AMD and Gigabyte board this past summer, and am not looking back.

I guess the takeaway from your comments is that if you want and need the most horsepower, then at the present, Intel is the way to go. For the rest of us, a current day budget AMD build still far exceeds the performance of the top end Intels from a few years ago.

Pretty much any modern Dual Core proc is going to be better that a older P4 or pre Core Intel. The cache size theory was put forth by my buddy Bob Nagy who used to work for Digi and was the first guy to get it running on PC. I did some Dverb testing a couple years ago when the 99.00 Athlon X4 came out and while it ran PT just fine it was very weak on processesing power. How weak? How about less than a 99.00 intel E7400 Dual Core. I suspect this was due mainly to the fact the Athlon X4 had very small caches. The Intel was no barnburner either at 2.5GHz and as I recall 3MB of cache. When we went from the Intel dual cores to the quad cores we saw roughly 75-80% of a power increase. Not quite double but pretty close. Pretty damn good bang for the buck. I was hoping the AMD quad core would provide the same boost but it didn't happen. I stuck a 199.00 X6 1090T in it and with the larger caches the performne scaled more like the Intels had. The AMD X6 used the same size cache as the X4. I suspect had they had the room on the die or power envelope to bump the cache up to the per core equivalent of the X4 965 that the X6 would have been a real barnburner on a budget. Still a great chip. They do run hot though. Hot enough that I couldn't used the X6 in a favorite micro ATX case I've used on many other builds. Even with all of the power saving features disabled. The power supply fan would detect the extra heat and spool up like a turbo. I ended having to stuff it in a full size ATX tower to dissipate the heat. Still the machine ran PT fine. Integrated Graphics an all.

Since we are discussing this type of stuff I'm going to express my opinion on integrated graphics and PT. And its just that. An opinion. I have no hard evidence only empirical data gathered from the forum and my builds. Early integrated graphics were pretty bad and often required some assistance from the CPU to help render screen redraws. Obviously PT would not like the interuption and throw out a error since it had to stop processing audio. Couple this with the fact that most desktop machines of the era came with gimped out binned chipsets that often had performance features disabled. Combine this with little or no graphics memory and and sharing slower system memory this made for both dismal graphics performance and required more CPU intervention than should have been required. Add in buggy drivers for Intel Graphics and these systems were not good at much other than running MS Office. Some other problematic chipsets were the AMD RS series and most of the Nvidia 61xx series. Where am I going with this.... Modern integrated graphics are plenty powerful and the drivers are good. There's notebooks that run PT and plenty of Macs that used integrated graphics. My suspicion is that its not the integrated graphics as much as it is the fact that some of them dynamically allocate memory to the graphics from the main memory. PT goes to use a memory address that it had previously used only to find it unavailable. Pro Tools had issues with dedicated video cards that used dynamic memory allocation like ATI HyperMemory or NVidias TurboCache. Modern integrated graphics are powerful and the boards no longer have second tier chipsets. As long as the amount of memory allocated to the graphics is fixed in BIOS than your chances are pretty good its going to run PT OK. This isn't any specific recommendation for integrated graphics. In fact I would very much say stick with a conventional chipset and video card format simple because its potentiialy less prone to have issues since PT is coded more towards a workstation type of architecture.

I'm mentioning it for a reason. I know there's some if the forums sharper eyes in the thread. I've seen more than a few posts were users with lower end AMD laptops that the the A Series APUs (Accelerated Processing Units) in them. These are different from regular X86/X64 chips in that the processing cores and onboard graphics are now sharing the same caches on the die. And theoretically all of this dynamically shifting allocation of memory between graphics and applications should be seamless if AMD has gotten their architecture right. I however doubt that Pro Tools is "aware" of this architecture since its likely compiled with a traditional Intel environment in mind. There were a rash of posts around the holidays from new users with issues. And many of them were running entry level laptops with Pro Tools SE etc. And many of these types of users have problems with the software and their hardware anyway. So its possible that there's no correlation between APUs and issues. Just low end hardware, less than optimal software and user error. I'm just trying to determine if there's a trend.

I may have the Dverb test score from my experiment a few years ago. If I can find them on my server I will post them although they are a bit out of date and not entirely relevant.

One thing I would like to say is that having built, repaired and refurbished a LOT of computers some of the nicest running computers I've had have been AMD. You build enough machines and some of them just run smooth as glass. My apologies for the Wall O' text.

[mjslakeridge]

Lots of good info in your post. I don't like to fool around in the BIOS too much, but how would I determine if the amount of memory allocated to graphics is fixed? Currently I have selected "load optimized defaults". I am using Award BIOS version F3.

When I ran the Dverb test the other day, I noticed (using Task Manager) that the page file utilization never exceeded about 1.04 GB. I have the initial and max page file size set to the highest value allowed 4,987 MB-any problem with that? I have 4 GB DDR3 dual channel 1333 ghz Kinston memory installed.

Time to run some more Dverb tests, I maxed out using 48 audio tracks the last time, going to set up some aux tracks and load them up.

[sunburst79]

Originally Posted by mjslakeridge

Lots of good info in your post. I don't like to fool around in the BIOS too much, but how would I determine if the amount of memory allocated to graphics is fixed? Currently I have selected "load optimized defaults". I am using Award BIOS version F3.

When I ran the Dverb test the other day, I noticed (using Task Manager) that the page file utilization never exceeded about 1.04 GB. I have the initial and max page file size set to the highest value allowed 4,987 MB-any problem with that? I have 4 GB DDR3 dual channel 1333 ghz Kinston memory installed.

Time to run some more Dverb tests, I maxed out using 48 audio tracks the last time, going to set up some aux tracks and load them up.

The BIOS is my playground. In mine on the 785G chipset theres a setting for graphics reserved memory or something like that. I'd have to look farther.

Under Advanced BIOS Features I have a setting called UMA Frame Buffer size with the choices of AUTO, 128, 256 and 512. I set it at 128 figuring it was enough memory and I would rather it be a fixed value since their was no explanation of what AUTO entailed. Since I wasn't having errors before I can't say if it made any difference. Seems like the prudent choice.

I'm not sure I would have changed the Pagefile from Auto or 1.5 times the amount of system RAM. I've never seen a difference by changing virtual memory. From what I remember once theres a more than a 1GB of RAM the page file doesn't come into play much. As far as I know there no performace hit from having more page memory other than taking up a little more HDD space.

[mjslakeridge]

Thanks Scott, I will look into the BIOS settings. I do have the manual which describes all of the BIOS settings, just never had to dig into it much before.

With 4GB of Ram installed, I was a little surprised to see the pagefile usage at around 1 GB, but that was running Dverb. My PT disk utilization meter was at about 50%, while in my normal sessions, the meter stays very low. I may defrag the audio drive, it's been a while.

Looks like the DVerb limit for my machine is 271, using all 6 cores at 90%. I did not move all of the extra plugins to the unused folder as recommended, my goal is not to necessarily get the highest score, just to give someone wanting to do a budget AMD build (at least for PT LE) some idea that my build is no slouch.

Mike

[sunburst79]

Originally Posted by mjslakeridge

Thanks Scott, I will look into the BIOS settings. I do have the manual which describes all of the BIOS settings, just never had to dig into it much before.

With 4GB of Ram installed, I was a little surprised to see the pagefile usage at around 1 GB, but that was running Dverb. My PT disk utilization meter was at about 50%, while in my normal sessions, the meter stays very low. I may defrag the audio drive, it's been a while.

Looks like the DVerb limit for my machine is 271, using all 6 cores at 90%. I did not move all of the extra plugins to the unused folder as recommended, my goal is not to necessarily get the highest score, just to give someone wanting to do a budget AMD build (at least for PT LE) some idea that my build is no slouch.

Mike

At the end of the day (to me) its all about how it runs and if you have the horsepower to do the mix you want. The idea behind Dverb Test is its CPU cycle intensive and not so much memory sensitive. You can probably get the memory usage up there to PT's current limit - which is about 2.7 GB as I recall. Good way to do this is to creat a bunch of instrument tracks with MiniGrand. You should crash or have the system become unresponsive once your at or near the RAM limit. Your upper limit is likely to be less since you sharing that 4GB with Windows.

As far as the slow drive error goes I would ensure the drive is 7200RPM and has a healthy cache. Defragging it may help if its heavily fragmented. Opinions vary as to whether one should defrag the drive by using the Windows Defragger or the method of copying it to another drive. If your drives are not set to AHCI you may want to set them to that in BIOS. If you installed Windows using IDE it can be a little tricky enabling AHCI. Make sure you do all the Windows setup tips as suggest in the troubleshooting threads at the top of the forum and your getting started guide. My final piece of advice would be to A) Make backups. B) Take notes. C) only change one thing at a time.

[mjslakeridge]

Scott,
I found the UMA Frame Buffer size setting in my BIOS and set it to 128. When I tried to set the SATA drives to AHCI, WIndows XP would not start up, so I changed it back. My drive is 7,200 rpm. I wasn't really having a problem with the drive, I just noticed it was working very hard on the 48 track Dverb2 test. I changed the pagefile size to 1,500 initial value and 1,500 maximum. What is puzzling to me is that I have a small PT session up now (4 audio tracks), and the pagefile usage is a pretty steady 814MB. The disk usage meter in PT is almost zero (maybe 2-3%). Not a problem or anything, just surprising to me.

[sunburst79]

Originally Posted by mjslakeridge

Scott,
I found the UMA Frame Buffer size setting in my BIOS and set it to 128. When I tried to set the SATA drives to AHCI, WIndows XP would not start up, so I changed it back. My drive is 7,200 rpm. I wasn't really having a problem with the drive, I just noticed it was working very hard on the 48 track Dverb2 test. I changed the pagefile size to 1,500 initial value and 1,500 maximum. What is puzzling to me is that I have a small PT session up now (4 audio tracks), and the pagefile usage is a pretty steady 814MB. The disk usage meter in PT is almost zero (maybe 2-3%). Not a problem or anything, just surprising to me.

If you restart the computer and just lauch PT with out opening a session what does it show? 814 doesn't seem hign to me. I didn't realize your on XP which doesn't support AHCI natively. Getting AHCI going in XP can be a bear and I dont think its worth bothering with. The only reason I enabled it (and it was a huge pain and a lot of trial and error) was because I was dualbooting Win7 64 and wanted to gain the hotswap ability on my motherboard eSATA ports.

Your questions got me looking at my manual since the AMD machine's in storage and I did see one item of note regarding the UMA frame Buffer setting. If you have a BlueRay Disk player and are playing BlueRay disks the UMA buffer needs to be set to 256mb. Since I'm pretty sure you need Vista or above to play BlueRay I don't thing its a concern. Something you would want to know though if Win7 or Win8 and a Blueray upgrade were on your list of upgrades.

[mjslakeridge]

Launching PT without opening a session, the pagefile usage is about 600mb. Depending on how big a session is, the pf utilization ranges from 800 to 1,100. Even opening a program such as Windows Media Player (with nothing playing) the PF usage was around 300 mb. Again, not a problem for me, but I though PF only came into play after all of the available memory was in use. I could be wrong there.

I do not have a blue ray player on my PT rig. I try to keep this rig as simple as possible. On my Windows 8 computer I have all of the daily use programs like DVD Shrink, IP camera viewer, etc.

I did discover today that on the PT computer, I had antivirus software active. Must have turned on when I did a reinstall of XP and service pack 3 a few months ago. Wasn't causing any problems with PT, as far as I know, but it is now turned off.

Mike

[sunburst79]

Originally Posted by mjslakeridge

Launching PT without opening a session, the pagefile usage is about 600mb. Depending on how big a session is, the pf utilization ranges from 800 to 1,100. Even opening a program such as Windows Media Player (with nothing playing) the PF usage was around 300 mb. Again, not a problem for me, but I though PF only came into play after all of the available memory was in use. I could be wrong there.

I do not have a blue ray player on my PT rig. I try to keep this rig as simple as possible. On my Windows 8 computer I have all of the daily use programs like DVD Shrink, IP camera viewer, etc.

I did discover today that on the PT computer, I had antivirus software active. Must have turned on when I did a reinstall of XP and service pack 3 a few months ago. Wasn't causing any problems with PT, as far as I know, but it is now turned off.

Mike

The PF Usage reading is something of a misnomer IIRC. If I remember correctly its not indicating the actual PF usage or the amount of memory usage being used by applications and windows. To get that you need to subtract subtract Total Commit Memory from Total Physical memory. Total Commit should be the total of RAM+the actual virtual memory being used. As long as Total Commit is less than Total Physical memory you should be good.

If your OS was attempting to write a gig of info to the HDD all the time it would be thrashing around like crazy all the time.

I'm going from memory here so this information may not be exactly 100% accurate but it should be close.

I haven't thought about any of this stuff since 7.4 and the days when we used to edit our Boot.ini files to enable the /3GB switch

[sunburst79]

If your running XP and have 4GB of RAM and PTLE 8.0 its probably worth while to edit your boot.ini file for the /3GB switch. As detailed in this thread. http://duc.avid.com/showthread.php?t=212853

Make a backup copy of your existing boot.ini and save it to your desktop or a thumb drive "Just in Case". Its a simple notepad file.

The post linked to is probably my favorite posts of all time in the DUC. New guy wanders in and on his first post lays out a step by step method to increase the amount of memory PT can use by a GB or so. It helped a lot. Then Digi patched PTLE 8.0 so it was large address aware.