core2 duo vs quad core

[Matt Vanecek]

Originally Posted by David Chilson

I have been working on my new workstation and the I7 is the way to go. After a couple of days I have the 920 up to 3.4GHz running 12gb Corsair ram at 1700. I have attached a SiSandra screen grab to show the comparison against a stock I7 and several other quad and dual core processors. If you are going the workstation route, I can highly recommend it.

I keep looking for answers amid the myriad solutions available, but what do you use to cool the system running it OCed that way, and what's your average CPU temperature when rendering? I'm deathly afraid of burning mine up, so I'm just using the stock speeds, with a Zalman CPU fan, and have Q-Fan turned on in the BIOS to speed up/slow down the fan. I see temperatures in the 60's when rendering, at times. I'd really like to keep the CPU temp in the 40's, if possible, when OC'ed.

thanks,
Matt

[Christopher Jensen]

Originally Posted by Dave Blackhurst

For video applications, the Quad cores usually shine over dual core, it's very noticeable in EVERY test I've run across. If you're editing HDV or AVCHD, a quad is almost a necessity, and the i7 appears to be the best bet by far, but not available in a laptop as of yet...

Originally Posted by Brian Boyko

Both Sony Vegas and Adobe Premiere can take advantage of multicore solutions. I think you'll get faster render times with the 2.0GHz Core 2 Quad.

Tom's Hardware was one of many sites back in the fall to tackle some of the issues regarding the merits of higher speed dual core versus lower speed quad core. At the time, they compared a higher clock speed Intel 3.16 Core 2 Duo to an AMD quad Core 2.0Ghz.

In general, they showed the higher clock speed duo outperformed the lower clock speed quad core. The graphs on page 7 of the review will give you an idea of how this dual core versus quad core debate evolved with some of the common video applications that people on DVinfo use. Performance: Video Benchmarks - Review Tom's Hardware : Dual-Core Versus Quad-Core: Part 2

Some of what has changed since that article was written , is that the Intel i7 quad core chip became available and some of the applications (like Vegas and Premier) now do a slightly better job of utilizing multiple cores.

The Intel i7 920 2.66Ghz Quad Core CPU has become extremely popular for video editing solutions in part because it is relatively inexpensive, delivers measurable performance gains, and is relatively easy to overclock at modest levels into the 3.6GHz to 3.8Ghz range. Higher speeds are possible but that introduces a whole different set of variables with more advanced cooling and stability.

While the Intel i7 920 is a great value in terms of price and performance, the same cant be said for the i7 940 (2.93Ghz) and i7 965 (3.2Ghz). The price premiums for the i7 940 and the i7 965 (about $270 and $720 respectively) over the price of an i7 920 are difficult to justify in terms of performance gains and added value.

PC World ran an article awhile back about one of the vendors, Eurocomm, getting ready to release an i7 based laptop. Battery life and heat dissipation would be a real engineering challenge to cram a current i7 into a laptop. It will be interesting to see how they handle that and at what price point if they bring it to market sometime in May.
Retailer to Ship Laptops With Intel's Core I7 Chips - PC World

Dave makes a good point in general that quad cores can be beneficial in editing the more computationally demanding AVCHD and HDV formats. A distinction that might be helpful is that the older quad cores that run at the lower clock speeds (2Ghz and below) are not always worth paying a premium for.

The i7 core 920 is a great value but I also wouldn't characterize it as the "best bet by far".
For anyone who is building a new video editing station today, there are multiple price/performance/design/expandability reasons to skip the i7 entirely.

There are newer motherboards available that support 2 physical CPU chips and more memory than the typical i7 setup. Four cores on each chip x 2 chips enables 8 physical cores available to tackle the video editing challenges. These newer motherboards are often referred to as workstation or server boards. Some of the more technical folks will refer to them as Dual Xeon 5500 series boards.

As you evaluate any core i7 based solutions, you may find it helpful to also consider a Dual Xeon based setup. More details about the price/performance benefits of the Dual Xeon 8 core setup will become available over the next few months.

--Christopher Jensen

[David Chilson]

Matt,

Your temperatures are fine. The I7 processor runs much hotter than what you are used to. It doesn't start to throttle down on it's own untill it reaches 100c. To be safe, keeping it below 80 is a good rule of thumb.

If you call intel and ask they won't tell you what the maximum temperature is. It's a trade secret. It's their contention they perform just fine at stock speeds with the stock cooler.

I have a Coolermaster V8 on my I7 920, running at 3.57 GHz. I ran prime95 for 24 hours and the temperature never went above 74c, with 0 errors. If you haven't done so, download that program and run a test or two. It will show all 8 cores running at 95 to 99 percent. This is also with overclocking 12GB 1600 MHz ram to 1700 MHz.

Where people run into problems is when the overvolt their system. One must be very careful manually entering voltages because some Bio's will let you enter high enough voltages to cook it.

Being conservative, I stuck with the stock voltages and adjusted the BLK core and memory frequencies only. If you have an Asus board with the latest bios, you can manually adjust the core multiplier from 20 to 21, giving you the turbo feature in the manual mode.

My I7 before overclocking was very similiar to yours in temperatures, so even though I am have no experience with your exact cooler, it seems like it is similiar and you could easily and safely replicate these results.

Christopher,

In reference to that comparison test you referenced it is apples to oranges. And if you can come up with a "multiple price/performance/design/expandability reason to skip the i7 entirely" I would like to hear it. (Other than the fact it was not in your budget)

2 quad Xeon Chips running at 2.66 will set you back over $1200 plus another $500 for a board. Using it as a video editing workstation would show no appreciable increase in productivity but cost about $1100 more than an I7 chip and board. Not to mention the fact that installing anything but stock coolers would be pretty much out of the question.

IMHO if you are buying a new computer for video editing, you best price performance lies with the I7 920 over the Xeon or any other combination out there at this time. I ran a SiSandra test comparing my processor to a stock I7 and a few other Intel Cpu's and have attached it.

Dave

[Bill Koehler]

Originally Posted by David Chilson

Matt,

Your temperatures are fine. The I7 processor runs much hotter than what you are used to. It doesn't start to throttle down on it's own untill it reaches 100c. To be safe, keeping it below 80 is a good rule of thumb.

If you call intel and ask they won't tell you what the maximum temperature is. It's a trade secret. It's their contention they perform just fine at stock speeds with the stock cooler.

I'm sorry, but this is nonsense. For normal commercial grade semiconductors, normal max temperature spec. was, is, and likely will remain ~70 degrees C. and this has been true for decades, it is not some recent development. And max. temperature information has always been standard data sheet info. I should know as this has been my business and profession. It has to be for manufacturer's to know what they should be designing for.

In the case of the Intel i7 processor, take a look at this technical document:

http://download.intel.com/design/pro...hts/320834.pdf

Take a look at page 79. This took me all of five minutes to find.

If you really are regularly running your processor over 70C, you *are* shortening the life of your machine. If it was otherwise, a boatload of datacenters all around the world could save a fortune in cooling costs. If you look at what I just referenced, Intel shows max processor case temp at just under 70C.

[David Chilson]

There has been quite a bit of confusion over this. I think this article will help clear it up:

Intel Core i7 Temperatures | Puget Systems Blog

[Bill Koehler]

As I have been a computer hardware designer, I take that perspective.

If you thoroughly read section 6 of the Intel document, they lay out that the activating of TCC (Thermal Control Circuit) should be momentary and infrequent. This is not about defining the upper level of safe performance - it is a progressively staged last ditch effort to avert meltdown because your cooling system, whatever it may be, is grossly inadequate to keep temps under control.

I am always going to take the numbers of the major manufacturer, in this case Intel, over a third party claiming to know Intels product better than Intel. If you were in the position of being either a manufacturer or designer, liability+warranty issues would guarantee that view.

[Harm Millaard]

Dave,

I agree with you that the "best-bang-for-the-buck" is currently the i7-920. However if you are into OC, you may consider the Noctua NH-U12 SE1366 push-pull CPU cooler. It will cool your system far better than your current setup. Using Prime95 stress test I did not see temperatures higher than 69 while running at 3.6 GHz.

When using the PassMark test, even at 4.2 GHZ, I never saw temperatures of 70 or higher. In my i7 system I have 15 internal HD's and 2 BR burners, so it is pretty crowded, but my disk temperatures are max 23 for 7200 disks and 31 for 10000 disks and my normal CPU temperature for the CPU is around 40 while running at 3.6 GHz.

BTW my PassMark rating is currently at 4733.76. I would like to hear your results.

[David Chilson]

Harm,

I had to manually adjust the fan on the cooler to get it to go above 70 degrees. When it runs with the fan attached to the board or controller turned up it's below 70. I was testing it for worst case scenario in case ambient temps rose.

Bill,

As far as the nonsense crack and this statement:

"If you thoroughly read section 6 of the Intel document, they lay out that the activating of TCC (Thermal Control Circuit) should be momentary and infrequent. This is not about defining the upper level of safe performance - it is a progressively staged last ditch effort to avert meltdown because your cooling system, whatever it may be, is grossly inadequate to keep temps under control."

What you are referring to here does not happen at 70C. Do you even know at what temp this occurs at?

I was trying to address Matt's concern about feeling the need to lower his temps from the 60C to 40C range. Referring him to an article that has some actual background and my own first hand experience dealing with the same concerns and actually testing them at higher temps.

So what he's come up with through searching the Internet he has convinced himself that lowering the temperature an additional 20C is appropriate. He’s looking for more cooling and the computer isn’t.

Now you’re classifying my suggestions as nonsense when your research consisted of a 5 minute search, secure in the knowledge that technology hasn’t changed since you have been at the helm, combined with your obviously zero experience taking this processor over the factory 70C "limit".

As if your 68C well running machine that goes up 5C all of a sudden will take exteme measures to advert an impeding meltdown because of one's now grossly inadequate cooling system. Sorry, it's not going to happen, and your response did nothing to clear up the confusion.

[Christopher Jensen]

Originally Posted by David Chilson

And if you can come up with a "multiple price/performance/design/expandability reason to skip the i7 entirely" I would like to hear it. (Other than the fact it was not in your budget)
Dave

Dave – thanks for sharing a response – your insights are most appreciated.
>In reference to that comparison test you referenced it is apples to oranges.

The authors of that article elaborated at length on the rationale of the components they selected. There was no pretense of a apples to apples comparison. They attempted to assess the basic issue of whether a higher clock speed duo would outperform a lower clock speed quad.

Instead of dismissing the article on a superficial analysis, it would be more constructive to review what lessons can be learned from the article and if the info will help people save money, select better components or build a better system.

The pretense of the article (to compare somewhat dissimilar systems) is still highly relevant today as many people are evaluating whether an i7 920 type setup would work best for them – or if they should skip the i7 and do a dual Xeon setup instead.

The recent Mac Pro is an example of a Dual Xeon setup. The $3299.00 introductory price tag caused quite a stir. However, the apple site presents a decent summary of why the dual cpu architecture has some architectural, design and performance advantages that might appeal to the video editing crowd. Apple - Mac Pro - Features - Processor - The fastest and most powerful Mac ever.

Many people who have previously built pc based systems correctly assessed that something comparable or “better” could be built for less than $3299.00

Before we get too far down that trail, it helps to assess the basic question of whether 2 physical CPU’s Xeon (4 cores each for a total of 8 physical cores) has advantages over a single CPU Quad core like the i7 920.

Last year , many of the comparisons were focused on Core 2 Duo versus Quad core. Now there are enough shipping products in the *8 core” category to shift the focus to comparisons of 8 core versus quad core.

>And if you can come up with a "multiple price/performance/design/expandability reason to > skip the i7 entirely" I would like to hear it. (Other than the fact it was not in your budget)

Your quirky comments about what is or is not in my budget won’t help anybody on DVinfo build a better pc. So lets not further dilute the credibility and value of the good insights that you do have by any more of that.

The more substantive issue that other people would actually care about is the expanding availability of Dual Xeon motherboards that may help them build a better system.

> 2 quad Xeon Chips running at 2.66 will set you back over $1200 plus another $500 for a > board.
No one recommended Dual Xeon 2.66 Ghz in much the same way that no one is likely to recommend the spending $1000 on a i7 965 .

Asus offers an entry level Dual Xeon Board that currently retails for about $260.00 and not the $500 that you indicated. The Tyan entry level Xeon Board is about $250.
Newegg.com - ASUS Z8NA-D6C Dual LGA 1366 Intel 5500 ATX Server Motherboard - Server Motherboards
Newegg.com - TYAN S7002G2NR-LE Dual LGA 1366 Intel 5500 Tylersburg SSI CEB Dual Intel Xeon 5500 Series Server Motherboard - Server Motherboards

There are more expensive Xeon Boards beyond $500 but they have other features (more memory slots, pci slots, sas support that may or may not be important to some customers.

The entry level Xeon Nehalem 5500 processors start at $235 for 2.0Ghz and go up from there. That leaves us with entry level Xeon solutions starting at about $ 720 for 2 CPU’s and a motherboard.

A common i7 setup would start around $280 for the i7 920 CPU and anywhere from about $150 to $440 for a motherboard. That puts the starting point for an i7 setup somewhere in the $430 to $720 zone.

The point here is not to spilt hairs on the nuances of the motherboards. The point is that the price of viable Dual CPU Xeon systems can be configured within a few hundred dollars of a similar i7 based setup.

If the prices are within a few hundred dollars of each other, it brings us back to the original issue of whether a Quad Core i7 920 works best for you or if you would be better served by an 8 core dual Xeon setup.

> Using it as a video editing workstation would show no appreciable increase in productivity >but cost about $1100 more than an I7 chip and board.

There are many potential benefits that the 8 core Dual Xeon setup can bring to video editing. Generic conjecture statements like “no appreciable increase in productivity” are more emotional than factual and don’t really help anyone objectively evaluate options for a new system.
>Not to mention the fact that installing anything but stock coolers would be pretty much >out of the question.

If you zoom in of the product images for many of the popular Dual Xeon boards from ASUS, Tyan, Supermicro and some of the others – you will see there is not much validity to your statement regarding coolers. Many of these motherboards are larger (12” x 13”) than the traditional ATX (12” x 9”) boards and do not necessarily have the same cooler design constraint.

More information about the SSI EEB 3.61/2008 form factor and CPU placements are available at http://ssiforum.oaktree.com/pdfs/SSI...2008%201_0.pdf

Pictures and Specs for some of the Xeon Boards are:
ASUSTeK Computer Inc.
Super Micro Computer, Inc. - Products | Motherboards | Xeon Boards | X8DT3
ftp://ftp.tyan.com/img_mobo/S7016_2D.jpg

There are the usual challenges of whether a given cooler will fit one of these motherboards. Additional design consideration has to be given to ensure the heat exhausted from one CPU is not fed to the other and vice versa.

>IMHO if you are buying a new computer for video editing, you best price performance lies >with the I7 920 over the Xeon or any other combination out there at this time. I ran a >SiSandra test comparing my processor to a stock I7 and a few other Intel Cpu's and have >attached it.

SiSandra can be a useful utility to assess some comparative system stats. The common problem with Sisandra and many of the other technical benchmarks is that the results are meaningful to a very small percentage of the audience that they are published to.

Stats like the inter-core bandwidth and inter-core latency won’t mean much to most people.

It would be much more interesting to hear in a less technical way about how some of the improvements you have made in your system have impacted everyday video editing related applications.

Sometimes telling us things in a simple way like the video encoding that used to take an hour now takes 52 minutes as a result of spending $$xx etc might be more useful than yet another SiSandra screenshot of inter-core latency in nano seconds.

Overall, no one is arguing whether the i7 920 represents great value and performance.
The issue is that the performance capabilities of a Dual Xeon system are becoming mainstream and may deliver better results at comparable or slightly higher price points.

Over the coming months we will have the benefit of comparisons from many more sources on the Dual Xeon 8 Core verus I7 Quad core to help people make more informed decisions on their video editing system investments.

I hope Dave and the others will continue to contribute their insights even if there are a few rocks thrown during otherwise constructive exchanges.

--Christopher Jensen

[Matt Vanecek]

Christopher's points are all well-made, I think, in the last post.

For a point--the i7 is virtually (as in, "virtual machine" virtually, and not "next best thing to.." virtually...) 8-core if you have HT turned on. And so far, I'm really appreciating HT. Probably not as good as 8 physical cores, but it is pretty awesome.

Were I 100% making a living from video, I would probably go dual-CPU. I'm curious if the Xeon Nehalem offers HT--would that make 16 virtual cores?? Guess I can look it up at some point.

For the work I do, the i7 is proving to be quite a blessing. I'm pulling in CFHD AVIs into CS4 as Windows VFW (or was, until Friday when PHD v4 beta was released), with no discernable hiccup in playback in multi-camera mode (2 audio/video streams from 1 eSATA RAID0). Also, M2Ts are similarly smooth--and I applied a color balance filter to one of those streams. I think if I added more effects/filters, that I would need to render some of the timelines to get smooth playback, but so far, rendering has not been required.

Capturing M2T & CFHD AVI via HDLink seems to be problematic (simultaneous)--the resulting AVIs were corrupt near the end. Recapturing the M2T and converting did the trick. I still need to play with that and make sure it wasn't a Firewire or camera issue...

Oh, speaking of which, performing the HDLink is blazing fast compared to my previous system (Athlon X2 64 Opteron).

That's my experience so far. My only concern is cooling. The NZXT case I have performs great and quiet, even with the CPU cooler running at full RPM. But I am a little concerned about CPU cooling in general, so I'm looking at some of the fanless liquid cooling options (only 1 available, it seems--the Zalman Reserator thing). I'd probably have the same concerns in a dual-CPU setup.

I did look at dual-CPU setups, but just can't justify the extra money right now (Production Premium CS4 + new computer and some new hard drives wrung me dry...). The cheaper dual-CPU motherboards generally don't provide me all the slots I need, and the more expensive ones are, well, expensive....

Just my experience so far...

Thanks,
Matt

EDIT----
Ok, I got home and reconverted some M2Ts that I needed to reconvert anyhow, using Cineform's HDLink. On my old system, it seemed like it was 1:1. I unfortunately don't have specific numbers, but conversion time was usually close to video's length, or maybe 3/4 of the video's length (empirically speaking). On this new system, 105 minutes of M2T took about 20 minutes to convert, on the High quality setting. The difference is VERY noticeable, and I'm quite happy. :)
---EDIT

[David Chilson]

Let me clarify, for my workflow 8 virtual cores are more than enough, so the I7 fits me just fine. I'm getting so old if I'm doing three things at once that's one too many.

A dual Xeon system may be better for running many more processes, and software may be changed to utilize these features in the future, but for me, spending that much extra cash at this time wasn't prudent.

The way I see video editing processes going is to pass off more and more CPU duties to the GPU, especially in regards to rendering. It's already happening with the CX from Nvidia.

Matrox announced a H264 accelerator board at NAB that's supposed to be available in the summer for around $500 bucks for PC/MAC so there are more hardware options coming too.

What's important to me is CPU speed and the amount of memory and bandwidth for each program I may have open. At 3.57GHz and 12GB of DDR3 memory at 1700 MHz, I'm a happy guy. It's the fastest computer I've owned.

As far as a real comparison test on how the I7 compares to my old system, to be fair I will have to wait. It's running CS3 and the Matrox RT.X2 and until the upcoming releases from Adobe and Martox are available, it wouldn't be an apples to apples comparison. To be fair, many plug-in companies are having problems with 64bit/CS4.

The limited rendering tests I have been able to run show the stock I7 920 averaged just under 6 minutes to render what my Q9650 needed almost 13. The overclocked version did it in 4 minutes, 30 seconds. This chip looks pretty promising. But I had to do this in CS3 we will see what happens in CS4.

[John Peterson]

I went from a Core 2 Duo E6600 (2.4GHZ) to a Quad Core Q6700 (2.6GHZ) and using Vegas 8 Pro configured to use all four cores I did notice a slightly faster rendering speed, however I don't think the difference was very significant in terms of practical time savings.

When I render a project I do something else while it is rendering so ten minutes or even a half hour doesn't make that much difference from my viewpoint.

I guess it all depends upon what you are doing, but speaking for myself anyway, the upgrade wasn't a big leap and probably not worth the cost for me.

John