I started with a Leadtek 6800GT PCI-E in a system full of mostly overclocked high-performance low-noise devices, so temps were reading high. In fact, they were so high I was beginning to suspect inaccurate temperature diodes.
Here's the pertinent componentry of the initial setup, followed by original spec, or "baseline" temperatures:
Case: Antec SLK3000B (120mm Nexus "Real Silent" exhaust @12v; VGA, TAC ducts open)
Mobo: Abit AN8 Ultra ("Silent OTES" heatpipe cooling)
CPU: AMD Athlon64 3200+ @ 2.41GHz (1.525v)
VGA: Leadtek 6800GT (1 slot copper OEM cooler w/ Arctic Silver 5, 410 core/ 1100 mem)
RAM: 1024Mb Corsair Value Select PC3200 DDR
PSU: Antec Neo HE500
CPU Cooling: Thermalright SI-120 (120mm Antec Tri-Cool @ medium 12v)
System: 41c idle/ 44c load
CPU: 34c idle/ 46c load
PWM: 44c idle/ 51c load
VGA: 61c idle/ 89c load
Hot, and intensely audible!
The Silencer itself is a well designed unit, stylish and substantial, with perhaps the exception of the clear plastic housing, which lacks a certain heft. I was relieved to see that the recent recall of all AC Rev.3 models was effective and fan polarity had been corrected on this example, although it doesn't matter in this case because the connection had to be modded to fit a proprietary Leadtek plug anyway. The packaging was impressive, providing an attractive presentation and adequate protection of the unit, but packaging is not why we purchase these things, and if you've read this far, you are probably waiting for juicy specifics, so on to installation.
Installation is similar to the other VGA Silencer units and is, in a word, painless. After replacing the fan connector plug with the proprietary Leadtek unit, I removed the OEM heatsink and removed the old thermal paste with 91% Isopropyl Alcohol. Preparation complete, it couldn't have taken more than five additional minutes to install the NV Silencer. Installation really is that simple.
Following the included directions very carefully, however, it's easy to see how some users are getting inadequate contact with their memory chips. Unlike the original NV Silencer 5 and the Rev.2 model the copper base utilizes thermal pads to contact the RAM rather than a machined surface, improving board-to-board compatibility. The thermal pads for the front mounted memory modules are pre-installed on the back of the integrated heatspreader, at the base of the front-side heatsink, in a somewhat haphazard manner. If the included directions are followed verbatim
, half of the memory modules will be adequately cooled, while the other half are barely covered. The simplest way to circumvent this is to remove the included thermal pads from the base of the cooler, apply them firmly and directly to the memory modules themselves and follow directions from there.
Installation complete, I couldn't wait to get up and running.
If first impressions were lasting ones, this one would have been a product killer. Upon re-starting the system a moderately loud, and very
noticeable, "whooshing" sound, cleaner and lower pitched, but similar in volume to the stock Leadtek unit (i.e. loud) was evident. Obviously, this is not what a quiet PC enthusiast is hoping for.
However, ten minutes with RivaTuner tones the noise down to acceptable levels. Final settings in this particular system are 30% at idle (desktop), 35% low-level 3D performance (heat throttling), and 43% during heavy 3D use (gaming, benchmarking). The fan is nearly inaudible outside the case to the finely tuned ear to about 45% in a very quiet carpeted room. While the room used for testing is not on the level of an SPCR test chamber, shallow breathing is noticeable from 3 feet away, so it's adequate.
Despite the reputation Arctic Cooling has earned, this particular fan is very smooth, not exhibiting the same bearing noise that characterized earlier VGA Silencer designs. It is obviously designed for low noise operation; no matter how slowly this sample spins, there is no perceptible clicking, rattling, humming or any other sound impurity. Close up it is very similar in quality, though higher pitched, to the much larger Nexus "Real Silent" fan, especially when the latter is undervolted.
So, how does the Silencer perform at these levels? Very well, but since numbers are better than hype:
System Setup Changes:
Fan speeds are lowered to enhance "calibrated-ear" measurements of the NV Silencer, and to contrast temperature readings.
Comparison to the system at previous settings has been added to provide an "apples to apples" understanding of the system in question.
120mm Nexus "Real Silent" run @7v (*low speed)
120mm Antec Tri-Cool fan run on low setting @9v (*low speed)
Arctic Cooling NV Silencer 5 Rev.3 installed
Video clock speeds increased to 414 core (+4MHz) and 1147.5 mem (+47.5MHz effective), sorry overclockers
Temperatures With AC Silencer:
System: 33/35c load (8/9c cooler)
CPU: 26c idle/36c load (8/10c cooler)
PWM: 37c idle/42c load (7/9c cooler)
VGA: 51c idle/69c load (10/20c cooler)
Temperatures With AC Silencer (*Low Speed):
System: 34/37c load (7c cooler)
CPU: 29c idle/ 41c load (5c cooler)
PWM: 37c idle/ 42c load (7/9c cooler)
VGA: 51c idle/ 72c load (10/17c cooler)
The entire system benefits with the NV Silencer. Everything operates cooler, safer and quieter with it installed. Even after reducing system fan speed, there is marked improvement overall system temperatures. Most interesting to our small collective of quiet crazies is the minimal impact reduced fan speed has upon component temperature readings with the NV Silencer installed.
At low speed, fan noise is no longer noticeable from as close as 6-9" from the open case. It's only within about 4" that fan motor and wind turbulence becomes audible at idle. With the case closed at peak speed (43%), the Silencer is audible only from the direct rear within 4". That's progress!
In the kitchen, where the computer is used, fan noise is completely drowned out by a refrigerator 18 feet away, with the system fans undervolted as noted above. Prior to installing the Silencer, video card fan noise was identifiable over the refrigerator, a washer and a dryer in an adjacent room on the other side of a wall. There is clearly a large improvement in real and perceived sound with the NV Silencer 5 Rev.3.
Nothing is Perfect:
As good as it is, there are a few things to address in the next revision.
The plastic housing feels a bit cheap. Maybe a slightly thicker gauge will improve this, maybe not, but as it is, it hurts perception of an otherwise well-designed product. It is difficult to make a case this affects cooling performance significantly, but it may have a subtle impact on acoustic properties.
Default fan settings should be fine-tuned or manual adjustments added to appeal to a wider range of enthusiasts. Not everyone likes their computer loud, or quiet for that matter. Regardless, a third party program, like RivaTuner, should not have to be used to tweak settings and attain tolerable noise levels. Perhaps the easiest solution is to simply package a utility such as ABit's "uGuru" with the product to allow the end user to easily adjust fan speeds to their liking.
The 6800 GT's power circuitry, barely cooled by the OEM heatsink fan, now operates without any
direct airflow. I haven't observed any ill effects thus far, but wonder about the long-term impact (see edit below). This may be a non-issue for some, given the excellent cooling over the rest of the card, and the fact that power module temperatures rise only slightly (from 44 to 46c) further reduces the likelihood that this is a significant oversight. Nonetheless, this is easily accomplished by venting the bottom of the fan shroud, and should be considered for the next update.
edit After removing the power module heatsink to allow for added clearance, the power circuitry quickly overheated, causing artifacting and instability. Although still substantially overclocked, lower clock speeds (400/1100) were necessary to maintain stable operation. Research the size and shape of the heatsink contacting the power circuitry of your 6800/7800 series card prior to installation to determine if this cooler best fits your needs. Oversized sinks, like Leadtek's 6800 series, may require modification or removal to ensure compatibility, reducing the overall effectiveness of the cooling solution.
Quibbles aside, the Arctic Cooling NV Silencer 5 Rev. 3 is an excellent product, that accomplishes its lofty goals admirably, greatly reducing both temperature and noise. Highly recommended.
Further tweaking resulted in increased video clock speeds. The "Performance 3D" setting in RivaTuner was increased to 75% (barely audible from ~3 feet, closed case) because the only thing that utilizes this setting is gaming, and either headphones or 5.1 surround speakers are in use, so the extra noise is not distracting (yet another example of why the NV Silencer should include fan adjustment...). Predictably, temps improve under load as well. If you're after a few extra FPS, this may prove advantageous. Results noted with system fans undervolted as noted above.
6800GT @ 420.75 core; 1166.75 mem. (+10.75 core (OEM cooling OC), +70.75 core (stock); +66.75 mem.(OEM cooling OC), +166.75 mem. (stock))
idle temp: 51c idle (-10c OEM cooling, same as prior settings); 66c load (-23c OEM cooling; -6c prior settings)
Testing Notes: Load temps for initial setup were taken performing prime 95 (CPU test) and a custom DOOM3 benchmark(GPU test) independently, each for 24 hours. The post-install results were obtained running Prime 95 and the ATI Tool artifact tester simultaneously for 24 hours, fully loading the CPU, GPU and memory subsystems. Because the second method puts more stress on more components (and logically should create more heat), results may appear less impressive than they actually are. I don't know that my system could have completed 24 hours of the second test prior to installation of the NV Silencer without crashing.
Room temps remained a consistent 24 C (75 F) throughout testing.