|
|
|
Diamond Viper V770 Ultra Review
Introduction
It seemed like only yesterday that the Diamond Viper V550, which is based on NVIDIA's award winning TNT chip, was the talk of the town.  The Viper V550 has served me well since last October as it was a key part of the Pentium II-450MHz based system I purchased from GamePC.  I held off purchasing a new system until the release of the TNT and Viper V550 and am glad I did.
Eight months have since passed by and now Diamond Multimedia has added two additional products to their arsenal of 2D/3D graphics accelerators: The Viper V770 and Viper V770 Ultra, both of which are based on NVIDIA's RIVA TNT2 chipset.
The standard Viper V770 features a 125MHz engine clock speed and 150 MHz memory clock speed while the V770 Ultra runs at 150MHz/183MHz. Both of these cards can be overclocked with the V770 Ultra being capable of reaching speeds up to a blistering 175MHz/200MHz.
In the first quarter of 1999, NVIDIA released their successor to the TNT chip and appropriately named it the TNT2.  Up until it's release, NVIDIA decided to keep information regarding the TNT2's specifications under wraps.  This decision was made due to lessons learned from the controversy surrounding the release of the TNT last year.  NVIDIA wisely responded to their customers by issuing the following statement in late 1998:
Dear Friends,
We've received a number of mails asking about clock frequency of the RIVA TNT chips.  NVIDIA's focus in developing the RIVA TNT was to deliver the best visual experience available anywhere, bar none.  Our focus has been on better realism and interactivity, both of which have been dramatically improved in RIVA TNT versus previous generations.
Early feedback from our OEMs and key developers was so positive regarding both performance and quality that we neglected to communicate the specification changes, deeming them as of relatively minor importance. It was a bad call on our part.
As background, the major improvements provided by the RIVA TNT are due to the dual texel pipeline, an architectural innovation which can more than double throughput, not the clock frequency. At the same clock frequency of the RIVA 128, the RIVA TNT delivers 2 to 4 times the performance of the previous generation.  This is due to architecture and software improvement, not clock frequency.
In a battery of tests that NVIDIA ran as we were determining final production clock speeds, we found that a 6% difference in RIVA TNT clock speed correlated with a change of anywhere from 0% - 3% on real applications.  Simply put, clock frequency is just one of several variables determining the quality of visual experience.
The design goal for the RIVA TNT architecture was to achieve 125 MHz. In order to meet high volume production for the fall, we made a decision to put the first implementation of the RIVA TNT into TSMC's high volume .35 micron process technology.  This required us to trade off clock frequency for volume production schedule.  The current operating specification is 90 MHz at 55 degrees Centigrade with 0 airflow, a conservative (most PCs exceed the minimum) airflow specification.
A couple of you have expressed disappointment that the first implementation didn't achieve the goals we communicated in February.   We regret that we didn't communicate our intentions sooner.  In any case, all of this is a long winded way of saying that we will do a better job of communicating our specifications and intent in the future.
NVIDIA's metric of success for the RIVA TNT was to deliver the fastest 2D and 3D processor in the market and we are proud to have achieved that goal.  We hope this helps clear up some of the confusion.
Best regards,
Dan Vivoli
VP Product Marketing
NVIDIA
The TNT2 has arrived and NVIDIA has nothing to apologize about this time around.  The TNT2 core is fabricated using .25-micron silicon wafer, which allows higher clock speeds to be obtained while keeping a lid on heat production.  The "suggested" core clock speeds of the ultra (150MHz) and non-ultra (125MHz) TNT2 chips were established by NVIDIA and passed on to board manufacturers including Diamond.
Similar to core clock speeds, memory clock speeds are ranging from 140MHz to 200MHz.  Ultimately, it's up to the board manufacturers to investigate options to squeeze out that extra "non-guaranteed" performance such as using faster rated memory, adding additional cooling, or optimizing drivers.
I purposely used the words extra "non-guaranteed" performance to indicate a recent trend with graphics accelerators.  Similar to overclocking processors higher than their intended speeds, graphics accelerators are following suit.  Based on yields received from a particular manufacturing lot, processors marked at one speed may actually be able to run at a higher speed.
For example, take the overclocking experience I had with the Viper V550.  After adding a standard Pentium cooling fan, the highest speeds I was able to attain was a 106MHz core and 120Mhz memory (up from the default of 90/110).  During the past 6 months, I've come across instances where other TNT owners have been able to surpass or fail to reach my speeds.  The same situation will exist with the Viper V770 or any other graphics accelerator that allows core and/or memory speeds to be "tweaked".
Next, let's take a look at the features of the Viper V770 Ultra.
Next: Viper V770 Ultra Features
|
Last Updated on September 4, 1999
All trademarks used are properties of their respective owners.
|
|