FIC Radeon 9800 Pro - Benchmarking Part 3
(Page 5 of 6 )
FIC ATI Radeon 9800 Pro (R98P) Review
(Benchmarks continued...)
3DMark2001 (Build 330):
3dMark2001 (Build 330) has obviously been around a while, as you can see by the 2001 name. That said, it still scales nicely with the newer cards and offers us if nothing else a fun synthetic benchmark.
While there is no denying the obvious advantage of the R9800, the 9700 performed better than I anticipated coming in only 700 marks behind the 9800 in performance testing. Again however, once we leaned on quality at all, even in the balanced settings, the Radeon 9800 showed it's older brother who carries the mail. Impressive default runs at the optimized settings runs of the 9800.
3DMark2003 (Build 330):
I realize that even including this benchmark is going to raise some eyebrows. It is no secret that Futuremark has shot themselves firmly in the foot with this application and may have forever tarnished their reputation. Whether the allegations about "tweaked drivers" is true or not, the seed has been planted and it will be up to Futuremark to come from under the cloud on their own.
So why include this controversial benchmark you wonder? It's simple.. Take a cruise around the OCA forums (or any other tech site's) and you will notice a large amount of end users still proudly proclaim their scores in this application in their signatures and conversations. Let's face it, "need" is rarely addressed when it comes time to purchase a new graphics card, "want" dominates the decision. Everyone WANTS to rule the roost and have the most braggin' rights. That considered, it only seemed logical to include the application that whether we approve or not, end users will still be running with their new FIC Radeon 9800 Pro. So let's do it...
In this application that features some DX9 coding, the Radeon 9800 Pro clearly emerged as champion. Winning easily at the lower settings, and downright stomping the 9700 Pro at the "quality" setting.
OVERCLOCKING:
Overclocking is a topic I hate to discuss in a video card review. Even more so than in CPU's I see a even larger variance in the overclocking performance of video cards from board to board. I am going to discuss what we achieved with this particular FIC Radeon 9800 Pro, but it in no way indicates what you will achieve with an identical card. You may get drastically better or drastically worse results.
I can tell you right now that with memory rated at 350MHz the success rate should GENERALLY go up with these FIC 9800 Pro's assuming FIC uses the same memory modules on every card. The core of these 9800 Pro's (and the 9700's as well) get VERY hot have no doubt. That leads us to believe that with increased cooling efficiency we should be able to achieve greater results, but that will be for you to determine should you choose to purchase and mod one. :) In the near future we plan to place a Danger Den Radeon chipset water block on this thing and see what happens, we'll keep ya informed.
Alright, what did we achieve with THIS card once we installed the Rage3d.net tweak application? A relatively impressive 423Mhz core and 380Mhz memory. Considering that the default speeds of this card is 380/340, that's better than a 10% increase on both core AND memory, all with stock cooling and default AGP voltage. Now take into count that FIC has no cooling on the memory at all, and a rather lame reference design cooler on the chipset, these numbers are even more impressive. I can also let you know from experience (several four letter words shouted to prove it) that the core on this thing gets more than hot, and the memory chips warmer than "warm to the touch". In the past video memory heat sinks have been more about appearance and less about performance. I think we've definitely reached a stage that the heat sinks could offer actual performance gains, or useful cooling if nothing else.
Alright, all of that is fine and dandy but what about the performance right? Well let's take a look at a brief sampling of scores when we overclocked the core and memory to the stable speeds of 423/380. I emphasize the word "stable" because there's a lot of confusion and interpretation in that word. When I say stable I'm not talking about "it made it thru a 3dMark01 run", I'm talking about "It ran 3dMark01 in an endless loop overnight without any visual anomalies". Let's do this..
Unreal Tournament 2003:
DM Antalus Map:
Now this is a rarity, in my opinion at least. We achieved a 10% overclock, and the results increased, you guessed it, almost exactly 10%. I'm used to things scaling but rarely this closely.
Alright, for fun we've also overclocked in 3dMark, hit "Next" and let's wrap this thing up.
Next: Conclusion >>
More Video Cards Articles
More By Jim Miller
/ 6 
