Welcome to Joe-Pinions, where the author puts to binary code his thoughts, anecdotes, and observations on just about anything under the sun. 
Addressing Compromises in Design
Before I majored in English in college, I was a mechanical engineering student. It used to be my dream to become a race car designer and work in Formula One. But once I was in college, I very quickly discovered that I had a fatal flaw: I did not have the drive to overcome my weaknesses in mathematics. Although I believe I had the creativity required to be an effective engineer, I simply didn’t have enough desire to make my dream come true by knuckling down and really working hard on my mathematics.
Despite my short time as an engineering student, I believe I learned enough to know that one of the immutable principles of engineering is that every design’s advantages must be balanced by some kind of penalty. For every gain made in any one area of a design, there is a consequent loss in another area.
All designs are creatures of compromise. This is the primary reason why I'm a little miffed whenever something in a particular product line is declared by consensus as "the best." Reading reviews and evaluations of these products tend to emphasize their quantifiable virtues, yet don't give enough weight to the inevitable compromises that balance out the design.
Take CPU air coolers as an example. PC enthusiasts and overclockers these days parrot Anandtech's current conclusion that the Thermalright Ultra 120 Extreme is THE KING in the realm of air coolers. Anandtech even goes as far as to say that blow-down coolers such as the Thermaltake Big Typhoon, the Enzotech Extreme-X, or Thermalright's own SI-128SE occupy a place "a notch below the top-performing heatpipe towers." Moreover, other PC enthusiast websites such as FrostyTech and MadShrimps place the Ultra 120 Extreme very high on their lists of recommended CPU air coolers, which naturally enhances the Ultra 120 Extreme's reputation even further.
With such glowing endorsements, it's all too easy to justify the purchase of an ultra-expensive product such as the Ultra 120 Extreme. I'll admit that I bought into the hype myself; I used to run my gaming rig (based on a S939 Asus A8N32-SLI Deluxe) with the Ultra 120 Extreme sitting on top of a naked Opteron 170.
The Ultra 120 Extreme performed really well for me, despite my needing to mod how it mounts onto my motherboard. But then I got to analyzing my machine and how it's put together, and I got to thinking: How are my motherboard's PWMICs/VRMs getting any air to cool them off? Just before summer ended, in fact, I had to get my A8N32-SLI Deluxe replaced (Asus took excellent care of me and provided a brand-new replacement board).
That's when it hit me: Very few (none that I can recall easily, actually) of the professional reviews that evaluated the Ultra 120 Extreme addressed the major compromise that a tower-style cooler design has, and that is any HSF that has its fans blowing parallel to the motherboard plane consequently does not cool the components located on the motherboard.
The spectacular performance on the CPU cooling masked the shortcoming of the design, and may have proved fatal to my original motherboard. The excellence of the design in one area made the compromise a little difficult to see; it was as if the brilliance in the CPU cooling performance blinded evaluators and prevented them from seeing that the great performance came at a steep price.
Machines, like organisms, depend on the smooth integration of many parts and components. The best designs don't tip the scales of compromise too heavily in favor of one side, just as the best people tend to be strong in many areas of appraisal. There is a reason a man with many abilities, talents, knowledge and understanding is lauded as a "Renaissance man." Better is the man who knows something about lots of things than the man who knows everything about just one or two things, in my opinion; specialization is not the best way in all things.
In the world of designs, nothing exists in isolation. Compromise is part of the natural order. Statistics that reflect too strong a bias in a certain area, and reviews that don't address a design's inevitable shortcomings, simply do the consumers who fork out their money trying to buy "the best" of something a big disservice.
Sometimes "the best" actually isn't.
That is, if you understand how it works, and therefore see the compromises in its design.
Despite my short time as an engineering student, I believe I learned enough to know that one of the immutable principles of engineering is that every design’s advantages must be balanced by some kind of penalty. For every gain made in any one area of a design, there is a consequent loss in another area.
All designs are creatures of compromise. This is the primary reason why I'm a little miffed whenever something in a particular product line is declared by consensus as "the best." Reading reviews and evaluations of these products tend to emphasize their quantifiable virtues, yet don't give enough weight to the inevitable compromises that balance out the design.
Take CPU air coolers as an example. PC enthusiasts and overclockers these days parrot Anandtech's current conclusion that the Thermalright Ultra 120 Extreme is THE KING in the realm of air coolers. Anandtech even goes as far as to say that blow-down coolers such as the Thermaltake Big Typhoon, the Enzotech Extreme-X, or Thermalright's own SI-128SE occupy a place "a notch below the top-performing heatpipe towers." Moreover, other PC enthusiast websites such as FrostyTech and MadShrimps place the Ultra 120 Extreme very high on their lists of recommended CPU air coolers, which naturally enhances the Ultra 120 Extreme's reputation even further.
With such glowing endorsements, it's all too easy to justify the purchase of an ultra-expensive product such as the Ultra 120 Extreme. I'll admit that I bought into the hype myself; I used to run my gaming rig (based on a S939 Asus A8N32-SLI Deluxe) with the Ultra 120 Extreme sitting on top of a naked Opteron 170.
The Ultra 120 Extreme performed really well for me, despite my needing to mod how it mounts onto my motherboard. But then I got to analyzing my machine and how it's put together, and I got to thinking: How are my motherboard's PWMICs/VRMs getting any air to cool them off? Just before summer ended, in fact, I had to get my A8N32-SLI Deluxe replaced (Asus took excellent care of me and provided a brand-new replacement board).
That's when it hit me: Very few (none that I can recall easily, actually) of the professional reviews that evaluated the Ultra 120 Extreme addressed the major compromise that a tower-style cooler design has, and that is any HSF that has its fans blowing parallel to the motherboard plane consequently does not cool the components located on the motherboard.
The spectacular performance on the CPU cooling masked the shortcoming of the design, and may have proved fatal to my original motherboard. The excellence of the design in one area made the compromise a little difficult to see; it was as if the brilliance in the CPU cooling performance blinded evaluators and prevented them from seeing that the great performance came at a steep price.
Machines, like organisms, depend on the smooth integration of many parts and components. The best designs don't tip the scales of compromise too heavily in favor of one side, just as the best people tend to be strong in many areas of appraisal. There is a reason a man with many abilities, talents, knowledge and understanding is lauded as a "Renaissance man." Better is the man who knows something about lots of things than the man who knows everything about just one or two things, in my opinion; specialization is not the best way in all things.
In the world of designs, nothing exists in isolation. Compromise is part of the natural order. Statistics that reflect too strong a bias in a certain area, and reviews that don't address a design's inevitable shortcomings, simply do the consumers who fork out their money trying to buy "the best" of something a big disservice.
Sometimes "the best" actually isn't.
That is, if you understand how it works, and therefore see the compromises in its design.
Total Comments 18
Comments
-
I believe that specialization can lead to great things as long as you understand that there is not just "one" thing (component) that needs to perform in order for a complex organism/machine as a whole to perform at "it's best".Quote:Machines, like organisms, depend on the smooth integration of many parts and components. The best designs don't tip the scales of compromise too heavily in favor of one side, just as the best people tend to be strong in many areas of appraisal. There is a reason a man with many abilities, talents, knowledge and understanding is lauded as a "Renaissance man." Better is the man who knows something about lots of things than the man who knows everything about just one or two things, in my opinion; specialization is not the best way in all things.
Thanks for sharingPosted 03-31-08 at 05:51 PM by mega_option101
-
This attention to detail and big picture thought process really would make you a great engineer. Math is a huge boundary to cross, that's for sure.Posted 04-02-08 at 12:25 AM by ted 
-
This might sound as a super off-topic but the human being body is simply perfect. Interesting read nevertheless.Posted 04-02-08 at 01:54 PM by SZayat
-
Thanks, ted, for your generous and kind comments.Posted 04-02-08 at 03:26 PM by txtmstrjoe
-
SZ, interesting, thought-provoking comment.
Worthy of more contemplation, it is.Posted 04-02-08 at 06:52 PM by txtmstrjoe
-
@SZayat: The human body is only perfect in a vacuum. Once you introduce enviromental elements (time, wear and tear, etc.), the human body breaks down quickly, based on the elements working against it. If the human body was perfect, it would never succumb to old age and sickness.Posted 04-04-08 at 06:03 PM by Sekigahara
-
Excellent point Sekigahara, but I was thinking about the human body more of a THE 'Machine'. Sure we are all certainly affected by all these elements, but till this very day there are no robot that would replicate human body movements successfully or that would live even that long under all these environmental circumstances.
I wasn't suggesting that we are immortals, well, I admit, I misused the word 'perfect' in my first post though .... but I still think we are.
edit: LOL... I hope you don't bring up animal bodybuilding!Posted 04-05-08 at 01:47 AM by SZayat
Updated 04-05-08 at 04:07 AM by SZayat -
If I interpreted that correctly, the point you're making is that nothing that we have created or will be creating in the foreseeable future will ever match the exceptional mechanics found in living organisms, which I agree with.
And... there's... animal... bodybuilding? ***?Posted 04-05-08 at 11:46 PM by Sekigahara
-
My bad again.... ;o, sorry amigoPosted 04-06-08 at 02:09 AM by SZayat
-
Actually the ACF7P has fins that bend to try and direct air onto the mobo parts.Posted 06-03-08 at 01:36 PM by lattyware
-
IMO if you're relying on your CPU HSF to cool other components you're doing it wrong. If a component needs to be cooled - cool it. When other heatsinks cool other components as a side effect, that is a bonus. It does not, IMO, reflect badly on the Ultra 120 Extreme (one of the most horrible names in computer parts) that it does not cool components other than the CPU.Posted 07-18-08 at 05:38 PM by rabidgnome229
-
I don't think it's too off-base to say that motherboards as presently laid out are designed to accept airflow from a down-blowing CPU cooler. Just look at the stock coolers provided by both AMD and Intel; all of them are down-blowing designs. It's not by accident that the hardware manufacturers have completely complementary designs to the products they work with, if they are components of an entire system.
Every designer chooses and prioritises certain aspects of the performance targets they want to hit. All I'm saying is that, in the case of most tower-style coolers, cooling the CPU socket area is ignored completely in pursuit of just the CPU cooling side of the equation. As you say, rabid, the end user can then add fans and other cooling methods to compensate for the inadequacies of a tower-style heatsink. But then, by definition, this testifies to the fatal weakness in the tower-style design.Posted 07-18-08 at 06:32 PM by txtmstrjoe
-
i dont think having hot air blow over my mosfets will improve the life of my system to be honest. thats what heatpipes are for: to carry the heat away to places where it WILL be cooled down.
its called a rear exhaust fan
and a top fan if you've got a good enough case for it :PPosted 01-30-09 at 05:59 AM by T3h_Ch33z_Muncha
-
wow well written and so true, i try not to buy into hype, as it can hurt your pocket a lot.Posted 01-30-09 at 08:25 PM by maxextz
-
Very helpful entry joe. I forget if I told you, but I had been considering a TRUE, but I'm almost completely set on a SI-128 SE thanks to this.
If you've done some research on the SI-128 SE vs a TRUE as far as performance goes, let me know. If I do go with a TRUE, I'm thinking i'll have to set up a spot cooler. (for an SI-128 SE, i'd be using a San Ace 1011.. much thanks to alexgheseger btw)
Chipset's all figured out now, by the way. Bought an HR-05 SLI/IFX to replace my Evercool VC-RE. Finally took your and thlnk3r's advice.Posted 01-30-09 at 09:43 PM by timxirish
Updated 01-30-09 at 10:00 PM by timxirish -
Well, I've used both an Ultra 120 Extreme and an SI-128SE. In terms of raw performance, the Ultra 120 Extreme on my machine is better by no more than 2 degrees C at the same ambient temperatures at maximum load.
The SI-128SE is what sits on my processor now, so you can see where my priorities reside.Posted 01-30-09 at 10:39 PM by txtmstrjoe
-
Heh, very exciting news to hear. Consider my CPU heatsink searches are over, so thank you. As you know, I don't find many tower heatsinks attractive.
Perpendicular versus parallel is one of my main reasons for that.. With the SI-128SE, the fan is parallel to the mobo and case window, making LEDs more noticible.
Although the San Ace is a black fan, I'm seeing a bit of modding potential. There's lots of room for at least four 3mm LEDs. I'll see what I can do with it when the time comes. :PPosted 01-30-09 at 11:30 PM by timxirish
-
Props for a great blog entry. A lot of the follow up posts in here are great as well.
Can't believe this was blogged almost a year ago. I must of missed this one sheeeshPosted 02-23-09 at 07:18 PM by thlnk3r
