Overclocking focused on pushing our BFG 6800GT OC as far as possible while achieving total stability in 2D/3D performance. BFG sells this card
as part of their OC series, in this case the core speed runs 20MHz above the usual 350MHz 6800GT core speed. Since Ultra 6800's run at 400MHz core, and the chip is purportedly the same animal this isn't such a great achievement, however; BFG's 6800 Ultra OC runs at 425MHz.
In every case, albeit air-cooled or water-cooled this card easily met 6800 Ultra speeds, and then BFG 6800 Ultra OC specs which is, at the very least indicative of BFG finding some decent GPU-lots. To save space I've excluded 3d benchmark scores per-se, although I did run/record these test results including all Futuremark, Doom-3, and Far Cry to ensure the card was stable under all conditions. This is, however; a water block review, not a graphic card review, hence our primary concern will be temps.
While it's well known among the Overclocking and Enthusiast community’s temperature always affects performance, it would be presumptuous to think this is common knowledge.
For this reason I've chosen several 3DMark03 scores exemplifying how lower temp's can improve actual scores. Below I've included several thumbnails each corresponding to a 3DMark03 result. First (left) 3DMark03 result screenshot represents the BFG 6800GT OC running at its default speeds (370MHz/1000Mhz) with its stock-air cooler. The second (center) screenshot represents 3DMark03 result running the BFG 6800GT at default speeds water-cooled by the NV-68 in a closed (proprietary) system. The final (right) screenshot represents 3DMark03 result overclocking the BFG at 440MHz/1150MHz, once again NV-68 water-cooled on a closed (proprietary) system.
In each case BIOS settings remained the unchanged, including Vcore, VDIMM, CPU speed, FSB and memory timings (9x300HTT = 2700MHz).
Since there's no way to record temps while 3DMark03 is running, temps as indicated by RivaTuner Hardware Monitoring (GUI included in the screenshots) are not representative of temps during the benchmark itself. Although the screenshots were taken just moments after the benchmarks completed, as you may be aware, micro-chip temps will stabilize (drop) in milliseconds back to their prior level. This is primarily attributable to the effectiveness of your cooling system, other reasons would require several lengthy papers on thermodynamics and micro-electronics, for which I'm often criticized of delving in-too deeply in my reviews. Regardless our goal in this particular section was to show the end-result, being the 3DMark03 score. In fact during temp testing I discovered Rthdribl places much more stress on the card then any Futuremark, DOOM-3, or Far Cry benchmark as I discovered hap-hazardly after running the first series of Futuremark benchmarks with the card overclocked, only to have it fail when I ran Rthdribl.Overclocking the BFG 6800GT with the NV-68
As found on the previous page, I've included the screenshots (via thumbnails) so here's no doubt these temp/results were ran and recorded. While I find it hard to believe anyone would falsify graphs I'm sure it has happened, however; this is not the reason I supply the screenshots. The additional information contained therein may be useful. For example Smart Guardian also displays chipset temp, and PWM-IC
(Power Circuits on the motherboard). Depending how in-depth one would venture, seemingly superfluous data can be extrapolated into valuable theory. BFG @ IDLE 411MHz/1105MHz stock-air cooling LEFT/RIGHT - LOAD BFG @ IDLE 440MHz/1127MHz Single Feed (PolarFLO TT) - NV-68 LEFT/RIGHT Dual Feed IDLE BFG @ LOAD 440MHz/1127MHz Single Feed (PolarFLO TT) - NV-68 LEFT/RIGHT Dual Feed LOAD BFG @ IDLE 440MHz/1150MHz Proprietary H20 LEFT/RIGHT LOAD Proprietary H20 EpilogueDanger Den
released their venerable NV-68 nVidia
water block almost a year ago, and it sat in my room for several months waiting to be strapped to a PCIe 6800 version. When I'd first tested this GPU/memory cooler it was with an AGP based eVGA 6800GT and the temps were simply amazing. I achieved some exceptional overclocking, and best of al the temps were so low I could have ran the card forever at these speeds. That is the test of a true performer.Conclusion:
Danger Den lists the NV-68 at $125 MSRP, it can probably be found less somewhere other then their store. FrozenCPU is now offering an NV-68 SLI kit
for $232, and I can't think of a better H20 SLI solution at the moment. If I had one complaint about this block it would be the weight which is most likely un-avoidable if one wants the same performance, and the 1/2" ID which is an entirely separate issue. I simply feel 1/2" ID system are not worth the absence of ergonomics for the purported performance gain to be had with the additional water volume. Insofar as 1/2" GPU/memory coolers are concerned, this has to be one of, if not the best on the market. Danger Den was one of if not the first to incorporate memory and GPU water cooling into a single solution; they are pioneers in many respects. I proudly bestow my highest rating for the NV-68 a pH7
. I'd like to thank Dan (if he still remembers me) at Danger Den
for being so understanding with my postponement of this article.
Questions/Comments: forum thread