Creative 3DBlaster GeForce Annihilator - Page 4

Turn Up My Engine And Burn, Baby!!
First my system specs used for testing:

  • P3 450 @ 558MHz (4.5 x 124MHz)

  • 128MB LGS RAM


  • 01 x 13GB Quantum CR Ultra-ATA66 HDD (5400 rpm); 01 x 10.1 GB Seagate Ultra-ATA66 HDD (5400rpm); 01 x 6.4GB IBM HDD (5400rpm)

  • 01 x Creative CT6940 GeForce 256 (32MB SDRAM onboard) latest Nvidia drivers; Default settings with core/mem settings at 120/166; BIOS version

  • 01 x Sony 6X DVD-ROM drive

  • 12MB CL V2 SLI

  • SBLive!! Deluxe (Liveware 2.1 drivers)

  • VirataLink 1025 ATM card

  • 17” Mitsubishi DiamondScan 70 monitor (0.28 dpi)

  • Win 98SE

  • Ambient Temperature of 27 degrees Celsius.

As per my previous reviews, I first tested its overclockability in 2 aspects: AGP bus-speed and Core/Mem clock speed, to tune it to its optimum speed. In regards to overclocking the AGP bus, I’m happy to say that the Creative runs just fine on my 124 MHz bus speed. As you may know, the Abit BH6 does not provide a /4 divider for the AGP bus at higher clock speeds. Hence, I was running the GeForce without any problems at 82.67MHz. 

Secondly, I used the Creative overclocking utility and cranked up the memory setting all the way to its 183MHz limit. Note that the tool supplied by Creative only allows alterations to the memory clock and not the core clock. However, at 183MHz I had no problems whatsoever in Q3Test. Then I ventured further and noticed that the changes were correctly reflected in Powerstrip’s overclocking tool, implying that perhaps I could use Powerstrip to push it even further and change the core timings as well. The following table summarizes my results:

Core Clock Speed Mem Clock Speed Q3Test Results
120 183 Passed Q3demo1 timedemo
120 192 Passed Q3demo 1 timedemo
120 195 Q3Demo 1 started showing texture breakages and an interlacing effect, although it still runs for about ¾ way through 1 timedemo cycle.
140 192 Passed Q3demo 1 timedemo
142 192 Passed Q3demo 1 timedemo
145 192 Received lockups after about 1 min into Q3demo 1 Timedemo
150 192 Immediate lockup on launching Q3demo 1 timedemo

Looks like 142/192 is the breakpoint for my card. Still, I decided to perform all subsequent benchmarks utilizing 140/190 as a safer bet to ensure stability. It works fine so far even after an hour’s worth of benchmarking Q3 and 3DMark 99 at various resolutions…

If you do the maths, the 142 core max theoretically implies 568Mpixels/s fill-rate. Pretty decent in my view (as one can’t compare a TNT2 clock speed with a GeForce clock speed) owing to differences in chipset design complexity that would render different speeds attainable.

I was more concerned with the relatively low overclockability of the onboard memory. With my old, trusty 32MB V770 card non-Ultra TNT2 card, I could attain like 192 on memory speeds reliably with its 6ns Eltrontech RAM. With a second generation board whose mem bandwidth is probably much more important to achieve better scaling in performance, being stuck at 192 MHz just doesn’t do the GeForce chip justice. This may prove to be the bottleneck to the graphics performance (as we’ll see later). Note that one can attempt to provide better cooling to achieve a higher core, but the mem speeds are generally cast in stone ex-production.

So I guess this is a major consideration for anybody contemplating buying the board. This also reinforces my doubts of the SDRAM being rated at 5ns (1/5ns = 200MHz in specs) as it would be slightly below spec in that regard.  Guess the “–5T” marking may not infer a 5ns speed afterall, but rather a 5.5ns variant as most mem are usually capable of a 5-10% tolerance beyond spec. However, perhaps this could be a purely design issue and I guess Nvidia (and Creative) must have foreseen this – as they must have had a reason to set their default speed mem settings to 166, significantly below the stock speed of 183 on the Ultra-TNT2 boards.


< Previous

Next >