To get a faster dual-core or slower quad-core? That is the question...

[robogeek]

Originally Posted by TorontoSage

Hi. I am going through the same exercise right now and putting together my first HTPC which is going to be a headless server with SageTV, 1 analog and ATSC tuner card, 1 HD-PVR driving an HD cable box, 5 HD-200's, 1 500 GB HDD for the OS and 2 1.5 TB HDD's for media storage.

I am wondering if there is any reason why you went with the UD3R as opposed to the UD3P. They are almost identical boards with the only difference being one of the card slots. They both look like great boards at a good price and have good reviews.

I liked the slot configuration with 3 PCI and 3 PCIe...I was looking for something that had at least 3 PCI slots.

Also, how did you choose the Q8300 over all of the other quad Intel CPUs? I think I am going to go with a quad core CPU too, as a slower Quad core beats a faster dual core I have found from reading reviews. Also, I want to be prepared for other multi-threaded applications. Do you know if WHS or SageTV uses all four threads btw?

I was going to get the Q8200, but I needed another $10-$15 on my order to get the 6 months with no interest from newegg.com, so I bumped up to the Q8300. I can't say with any certainty that SageTV or WHS processes are using multiple cores on a per process basis, but the Task Manager shows each core usage on the Performance tab and all 4 cores seem to be more or less evenly utilized.

Also I hear more L2 cache improves performance quite a bit. The Q8300 has the smallest cache of all of the Intel Quad CPUs. How did this affect your decision, or did it?

Yeah, more L2 cache is better. My wallet dictated the decision. If I had another $150 to throw at the server I probably would have went for a Q9650 with 12MB L2, 3GHz, and virtualization support.

Finally I understand that of all of the Intel Quad CPUs, the Q9450, Q9400 and Q8200 are the only ones that have the lower power consumption (65w vs 95w). It looks like this might have not been a factor in your decision, but I am wondering what your thoughts are about it, as the Q8200 is almost the same clock speed (2.33GHz) as the Q8300 but lower power consumption, while at the same time also having a 4M cache, 1333 MHz bus speed and 45nm size.

When I ordered from newegg.com, the Q8200 and the Q8300 were only available in 95W versions and I think that was the lowest wattage of the quad core Intel processors they had available. If the Q8200 was available in the lower wattage version I may have considered it, but it wasn't a high consideration on my list.

[TorontoSage]

Originally Posted by robogeek

I liked the slot configuration with 3 PCI and 3 PCIe...I was looking for something that had at least 3 PCI slots.

The 3P only has 2 PCIe slots, but has 3 PCI Express XI and 3 PCI Express X16. Can you tell me where having 3 PCI slots is important? I will only be putting in 1 or 2 tuner cards, no modem and just a basic video in order to install the OS. (This is a SageTV server running WHS).

Originally Posted by robogeek

Yeah, more L2 cache is better. My wallet dictated the decision. If I had another $150 to throw at the server I probably would have went for a Q9650 with 12MB L2, 3GHz, and virtualization support.

I've decided on the Q9550 as the Q9650 is only a marginally 170 MHz faster, but is $332 vs $257, or 30% more cost. Both have the same specs and both have virtualization, although it seems to this non-techie that that has something to do with running more than one type of OS.

Originally Posted by robogeek

When I ordered from newegg.com, the Q8200 and the Q8300 were only available in 95W versions and I think that was the lowest wattage of the quad core Intel processors they had available. If the Q8200 was available in the lower wattage version I may have considered it, but it wasn't a high consideration on my list.

It turns out the prices for lower watt CPUs is sky high. I figured it would take over 5 yrs of electricity cost saving to make up the difference in price!

[robogeek]

Originally Posted by TorontoSage

The 3P only has 2 PCIe slots, but has 3 PCI Express XI and 3 PCI Express X16. Can you tell me where having 3 PCI slots is important? I will only be putting in 1 or 2 tuner cards, no modem and just a basic video in order to install the OS. (This is a SageTV server running WHS).

What was important to me (but may not be for you) was the fact I had 3 PCI cards including 2 PVR-500's in my old server. And paying slightly less for the UD3R motherboard and not having to shell out more money to replace one of my PCI cards with a PCI Express card was the main reason...saved me about $70.

I've decided on the Q9550 as the Q9650 is only a marginally 170 MHz faster, but is $332 vs $257, or 30% more cost. Both have the same specs and both have virtualization, although it seems to this non-techie that that has something to do with running more than one type of OS.

Good choice...I guess the 170MHz isn't really worth the price difference. I sometimes play around with virtual machines and the processor virtualization technology can help with performance.

It turns out the prices for lower watt CPUs is sky high. I figured it would take over 5 yrs of electricity cost saving to make up the difference in price!

The UD3R and UD3P (and some other Gigabyte motherboards) have Gigabyte Energy Saver Technology (GEST). You can enable it with their Energy Saver application in Windows and it can dynamically decrease and increase clock speed and voltage depending on CPU usage.

[TorontoSage]

There's an interesting article posted on the web that examined and did some tests to determine the increase in performance in going from dual to quad core. The conclusion in insightful:

The results seem encouraging, until you take a look at the applications that benefit from quad-core-- the ones that aren't purely synthetic benchmarks are rendering, encoding, or scientific applications . It's the same old story. Beyond encoding and rendering tasks which are naturally amenable to parallelization, the task manager CPU graphs tell the sad tale of software that simply isn't written to exploit more than two CPUs.

and

... only rendering and encoding tasks exploit parallelism enough to overcome the 25% speed deficit between the dual and quad core CPUs. Outside of that specific niche, performance will actually suffer for most general purpose software if you choose a slower quad-core over a faster dual-core.

Note that the author chose chips that had a clock speed deficit between them. So if the two you are comparing don't then this is a non-issue.

The article is posted here

I don't know if SageTV does much rendering or not but know I will be doing a lot of encoding of the H.264 stream from the 2 HD-PVRs I will have and so this makes the quad core a better choice for me. The video streams coming from the HVR-1600 tuner card are already encoded in MPEG-2, so I don't think the quad core will be of no help there.

[Slipshod]

Originally Posted by TorontoSage

I don't know if SageTV does much rendering or not but know I will be doing a lot of encoding of the H.264 stream from the 2 HD-PVRs I will have and so this makes the quad core a better choice for me. The video streams coming from the HVR-1600 tuner card are already encoded in MPEG-2, so I don't think the quad core will be of no help there.

The HD-PVRs do all the h.264 encoding themselves, so that's not really a reason to go to a quad core. The only common things I can think of that it would really make a difference with are placeshifting, transcoding, and commercial detection (and that's only if you are running more than 1 comskip instance at a time).

Future-proofing is also a reason, though it's pretty vague and may not be worth the money for your situation.

As far as L2 cache, I tend to go for the cheapest CPU with the most L2 cache. The X3350 is the xeon version of the Q9450, which at the time was the cheapest 12MB L2 cache CPU in that family.

[TorontoSage]

Originally Posted by Slipshod

The HD-PVRs do all the h.264 encoding themselves, so that's not really a reason to go to a quad core. The only common things I can think of that it would really make a difference with are placeshifting, transcoding, and commercial detection (and that's only if you are running more than 1 comskip instance at a time).

Future-proofing is also a reason, though it's pretty vague and may not be worth the money for your situation.

As far as L2 cache, I tend to go for the cheapest CPU with the most L2 cache. The X3350 is the xeon version of the Q9450, which at the time was the cheapest 12MB L2 cache CPU in that family.

Ok well that removes one reason for me choosing the Intel Quad Core Q9550 CPU. But, I will be doing comskip on all of my files so it will help there (and for future-proofing)

I am a newbie so excuse this question: I know that transcoding is going from one video format to anothe. Is this only required if I want to copy the video to an iPod/iPhone/iTouch or similar device or place it on the web? Is transcoding something someone can avoid if all they do is watch broadcast video and rips BluRays/DVDs? I'll be running 5 - HD200's off an SageTV server running WHS (no client HTPC's).

[Djc208]

Originally Posted by TorontoSage

I am a newbie so excuse this question: I know that transcoding is going from one video format to anothe. Is this only required if I want to copy the video to an iPod/iPhone/iTouch or similar device or place it on the web? Is transcoding something someone can avoid if all they do is watch broadcast video and rips BluRays/DVDs? I'll be running 5 - HD200's off an SageTV server running WHS (no client HTPC's).

By "on the web" I'm assuming you're talking about watching over a placeshifter connection? It woudn't be legal to post copywrited material on-line

Otherwise yes, if you want to convert it to a different format or quality (for remote use, archival, etc.) or you're watching via placeshifter over the internet you have to transcode the video first. Otherwise the HD200s will decode just about anything Sage can record, and even a client will do it's own playback.

I almost never use transcoding other than for placeshifting content. Otherwise I just copy the base file if I want to watch on my laptop somewhere I don't have internet access.

[TorontoSage]

Originally Posted by Djc208

By "on the web" I'm assuming you're talking about watching over a placeshifter connection? It woudn't be legal to post copywrited material on-line

Of course I wouldn't copy anything to the something like Youtube on the web, as I don't want to be the first person that puts a TV show clip on that site in violation of copyright

[TorontoSage]

I know that it goes without saying that you can only transcode on the fly for placeshifting live TV, but I was wondering, can you transcode in advance for placeshifting already recorded programs?

Also, this may be a silly question but does a faster processor necessarily result in a lower % of cpu usage over a slower processor, assuming the same number of cores in each processor? I understand that the transcoding process in SageTV is multi-threaded.

[QueOnda]

Good read: http://www.tgdaily.com/content/view/41371/135/

Originally Posted by UPDATED: February 10, 2009 - 10:06am CST

Sandia National Laboratories recently conducted simulations on the benefits and pitfalls of multi-cores. Their simulations show that moving from 2 cores to 4 cores shows a big increase in performance.

[TorontoSage]

and from the article referenced by that article:

Sandia used a simulation involving "key algorithms for deriving knowledge from large data sets." The simulation showed a significant increase in speed in going from two to four cores. However, there was an insignificant increase when moving from four to eight.

Is it implicit in this statement that the software has to be written to take advantage of the multiple cores or else there won't be any increase in performance? Also, I don't think that Sandia's conclusions necessarily apply to SageTV because it is a different type of application, but it strongly suggests that it will as SageTV can take advantage of multi-threaded processors. I went with the Intel Quad Core Q9550 because SageTV now takes advantage of all cores (but doesn't max them out, but I don't know if this is because there is not enough processing going on to max them out or because they would max them out if they could but can't write the software to do so). EDIT: SageTV itself only uses all of the cores when transcoding or placeshifting, third party programs such as comskip and showanalyzer may also take advantages of more cores in multi-core processors.

Or, with software that takes advantage of multi-threading, if the clock speeds are the same, does simply using a quad cored instead of a dual core increase performance for some other reason?

The article I referenced above also stated that 'it would be better to have several individual dual- or quad-core CPUs each connected to their own physical memory along with NUMA (Non-Uniform Memory Architecture) logic employed in the operating system'. So, this means there is still a bottleneck in multi-core processors, because of the fact the cores have to use the same bus to address the memory, and once they get there they have to wait for the memory to deal with the other cores that got there first. (I know this is simplistic, but I hope I am understanding it correctly.).

I understand that recording only uses about 5% of the processor power and also, given that it's the Hauppage and not Sage TV that does the H.264 encoding inside the HD PVR (I don't think it needs to do any in the OTA HD tuner as it merely passes the MPEG-2 stream), that a faster or multi-threaded chip won't make any difference with recording and not with reading either. But for running comskip and transcoding it should make a difference, as comskip is multi-threaded and transcoding is also multi-threaded because SageTV does the transcoding and it is multi-threaded. Is this all correct?

[Slipshod]

Originally Posted by TorontoSage

I know that it goes without saying that you can only transcode on the fly for placeshifting live TV, but I was wondering, can you transcode in advance for placeshifting already recorded programs?

Sort of... You could probably transcode to a more "friendly" format, at a lower resolution. H.264 is pretty CPU intensive, MPEG2 is much less so. I'm not sure if it would be worth the effort, or if the UI would be convenient and usable for it though. You'd probably have to transcode and dump into a video directory instead of the normal TV interface, but that's just a guess.

Also, this may be a silly question but does a faster processor necessarily result in a lower % of cpu usage over a slower processor, assuming the same number of cores in each processor? I understand that the transcoding process in SageTV is multi-threaded.

It's not a silly question; there are several variables that can confuse it. Here's some hypotheticals:

If the process you are running is limited by something other than CPU (such as disk access), then the % of CPU utilization from running that process will go down as the processor speed goes up. So something that takes 25% of one core on the core2duo at one speed will take less than 25% on a faster CPU. It would run for pretty much the same amount of time as well.

If the process is "CPU bound" (i.e. limited by the processing of the CPU), it will be running at 100% CPU util, and as you increase the CPU you will remain at 100% CPU utilization (with a shorter time to completion) until you run into another bottle-neck like Disk access. At that point it will behave as described above.

Other factors that comes into how CPU utilization is reported is how threaded the process is... If it's a single thread it will only run on one CPU, and could therefore only consume 25% of the total CPU time according to how Windows measures overall CPU utilization. That thread would be CPU bound, and the second case above would apply until another bottleneck showed up. This is why windows shows you the overall and individual CPU utilization for each core.

Most programs don't fall cleanly into one or the other case as different parts of the program will have different bottlenecks. Add that to threading and multiple cores, and it gets pretty hard to give a firm answer without knowing more.


As a general "rule of thumb", if you're not doing a lot of different things (threads or processes), you might consider 2-core instead of 4-core. It's kind of a short-term vs. long-term play though, as eventually more of your software will be more threaded, and you'll start using the PC for more things. My choice was pretty obvious because of the number of things I use the server for. If it was only for Sage, I'm not sure I'd notice a difference between 2 and 4 cores in how I use Sage, especially if the dual-core was a faster CPU. Later could be different, especially if I started doing a real amount of transcoding or placeshifting.

[TorontoSage]

Thanks for that very informative post.

Originally Posted by Slipshod

H.264 is pretty CPU intensive, MPEG2 is much less so.

Just what is H.264 pretty CPU intensive for, as the HD PVR has already encoded into H.264 before it arrives at the server. Is it the decoding of H.264 that is intensive? EDIT: I just found out that the HD100/200 extenders do all of the decoding of the H.264 stream, not the server, but even so is H.264 intensive for the server CPU for any other reason? I guess one reason would be where the server is used as an HTPC also (which mine won't be).

Originally Posted by Slipshod

If the process is "CPU bound" (i.e. limited by the processing of the CPU), it will be running at 100% CPU util, and as you increase the CPU you will remain at 100% CPU utilization (with a shorter time to completion) until you run into another bottle-neck like Disk access. At that point it will behave as described above.

But as you increase the CPU and assuming you don't run into another bottle-neck, then doesn't the utilization eventually drop below 100% because you eventually get to a speed that doesn't require 100% cpu utilizatoin to handle the process?

Originally Posted by Slipshod

Other factors that comes into how CPU utilization is reported is how threaded the process is... If it's a single thread it will only run on one CPU, and could therefore only consume 25% of the total CPU time according to how Windows measures overall CPU utilization. That thread would be CPU bound, and the second case above would apply until another bottleneck showed up. This is why windows shows you the overall and individual CPU utilization for each core.

But how can it be CPU bound if it is only using 25%, not 100%, of the CPU? Or do you mean that if it runs single-threaded on a quad core CPU, and is CPU bound on that thread, then it is using 25% of the total CPU time. So if the process was multi-threaded and it used all four cores then it would consume 100% of the total CPU time. Also, is there such a thing as a multi-threaded process that can only use some, but not all, of the available threads?

Originally Posted by Slipshod

As a general "rule of thumb", if you're not doing a lot of different things (threads or processes), you might consider 2-core instead of 4-core.

I'll be only using the headless WHS server for SageTV with extenders (EDIT: So I guess that means no decoding on the server as that is done on the extenders), almost no transcoding, maybe some rare placeshifting when I travel, but I do know that I will be using comskip all of the time. Future-proofing aside, would a dual core instead of quad core be more advisable in this situation, assuming the clock speeds were the same? I am most concerned about comskip being faster when it can take advantage of a multi-core cpu. I know that SageTV is multi-threaded now but I am not sure what part of the program benefits from this most. Do you?

Originally Posted by Slipshod

It's kind of a short-term vs. long-term play though, as eventually more of your software will be more threaded, and you'll start using the PC for more things. My choice was pretty obvious because of the number of things I use the server for. If it was only for Sage, I'm not sure I'd notice a difference between 2 and 4 cores in how I use Sage, especially if the dual-core was a faster CPU. Later could be different, especially if I started doing a real amount of transcoding or placeshifting.

What else do you use our server for? I am looking for ideas given this will be my first SageTV and server installation.

[TorontoSage]

Hey Slipshod, it's almost a month later. Are you still around. I'd love your feedback on my questions if you have the time! Thanks.