- Joined
- Mar 12, 2006
- Messages
- 327 (0.05/day)
- Location
- TX
Processor | Ryzen 7 5800x3d |
---|---|
Motherboard | Asrock b550m riptide pd |
Cooling | Thermalright Peerless Assassin |
Memory | 32gb (4x8) G.Skill RGB DDR4 @ 3200 |
Video Card(s) | XFX Speedster QICK319 RX 7800 XT 16gb |
Storage | 1x Hynix P31 1tb (os) 1x 4tb SP UD90 |
Display(s) | 1x Acer Predator X34, 2x 13.3 mobile USBC wings |
Case | Jonsbo D31 mesh |
Audio Device(s) | EVGA Nu Audio |
Power Supply | Corsair SF750w |
Software | Windows 11 |
The biggest problem lies with the architecture path that ATI has chosen. Nvidia's Scalar MADD+MUL architecture is nothing like the Superscalar MADDx5 architecture ATI is using. On paper ATIs can look stronger, but comparing the two architectures even on paper is not a safe thing to do because they are nothing alike. ATI may have more ALUs for example, but NVidias run at twice the speed, but then again the each unit in the superscalar architecture can do more independent instructions at a time.
ATIs weak point right now is really that its architecture requires highly optimized drivers, and your games need to be designed with this type of architecture in mind. If your application and driver are not constantly generating as many independent instructions as possible all the time, the execution units in this architecture are being left idle, and you have terrible efficiency and this impacts performance. Nvidias scalar architecture does not have this limitation, because its more flexible because in a lot of applications your next command often relies on the previous one, and you dont always need or can generate 5 independent operations. You can look to this to understand why ATI is often improving performance with every driver release instead of just improving compatibility with games. And it can also explain why at its best, the hd2900xt was able to beat the 8800GTX, yet at its worst the hd2900xt wasnt even able to beat the x1950xtx.
Another thing to remember is that with these architectures ATI is basically focusing all their power on raw shader power for which it is significantly faster than even the 8800 ultra for shader operations per second, yet it neglects things like pixel fillrate and texture fillrate. R600 only has 16 ROPs and 16 TMUs, compared to G80s 32 TMUs and 24 ROPs.
Its always going to boil down to case by case comparison, some games will run better now on G80/G92 others right now may run better on R600/RV670...who knows about tomorrow, maybe in a year ATI's architecture will show its full power, maybe not. Maybe ATI's will prove stronger with the professional segment where AMD wants it, and us gamers will have to make due with acceptable performance, or go to the competition.
What you have to remember is the difference between the fastest, and whats fast enough. Does an hd2900xt play most if not all games at high resolutions while keeping playable FPS? Yea. Will the new HD38x0 series? Yea. Is the 8800 series faster? More often than not yea... Will people pay more for a faster card? You bet your ass, but should ATI expect strong sales from something slower? No reason why they shouldnt. If they can get close to HD2900xt performance from a $150-$200 card range than thats a huge plus. The HD2900 pro sold VERY well at $250 for just this reason.
If ATI can make some further optimizations to the RV670 and related architectures, such as improving their shader recompiler, driver/application optimization, improving their texture dispatch processing, etc then they stand to gain solid ground. How many people believed the 8800GT would get close to the GTX performance? It wasnt hard when you know how to make a good thing better.
ATI and Nvidia have always, and will always trade the performance crown back and forth. One company will have the superior performance for one or two generations, then the other company will take that lead. If you look here and now your wasting as much time as you are money. Buy the cheapest card you can that remains "fast enough" and upgrade it two or more times a year. Buying 1 big expensive card and hoping it lasts you 2 years might work, but its usually gunna hit its peak long before then. Premium cards never stay on their pedestal for more than a year, and high end cards are lucky to get half that. Upgrade to the best value and upgrade often.
*edit*
And to fuel the fire a bit, take a look at some other numbers, namely the 3870:
twrococ.com
ATIs weak point right now is really that its architecture requires highly optimized drivers, and your games need to be designed with this type of architecture in mind. If your application and driver are not constantly generating as many independent instructions as possible all the time, the execution units in this architecture are being left idle, and you have terrible efficiency and this impacts performance. Nvidias scalar architecture does not have this limitation, because its more flexible because in a lot of applications your next command often relies on the previous one, and you dont always need or can generate 5 independent operations. You can look to this to understand why ATI is often improving performance with every driver release instead of just improving compatibility with games. And it can also explain why at its best, the hd2900xt was able to beat the 8800GTX, yet at its worst the hd2900xt wasnt even able to beat the x1950xtx.
Another thing to remember is that with these architectures ATI is basically focusing all their power on raw shader power for which it is significantly faster than even the 8800 ultra for shader operations per second, yet it neglects things like pixel fillrate and texture fillrate. R600 only has 16 ROPs and 16 TMUs, compared to G80s 32 TMUs and 24 ROPs.
Its always going to boil down to case by case comparison, some games will run better now on G80/G92 others right now may run better on R600/RV670...who knows about tomorrow, maybe in a year ATI's architecture will show its full power, maybe not. Maybe ATI's will prove stronger with the professional segment where AMD wants it, and us gamers will have to make due with acceptable performance, or go to the competition.
What you have to remember is the difference between the fastest, and whats fast enough. Does an hd2900xt play most if not all games at high resolutions while keeping playable FPS? Yea. Will the new HD38x0 series? Yea. Is the 8800 series faster? More often than not yea... Will people pay more for a faster card? You bet your ass, but should ATI expect strong sales from something slower? No reason why they shouldnt. If they can get close to HD2900xt performance from a $150-$200 card range than thats a huge plus. The HD2900 pro sold VERY well at $250 for just this reason.
If ATI can make some further optimizations to the RV670 and related architectures, such as improving their shader recompiler, driver/application optimization, improving their texture dispatch processing, etc then they stand to gain solid ground. How many people believed the 8800GT would get close to the GTX performance? It wasnt hard when you know how to make a good thing better.
ATI and Nvidia have always, and will always trade the performance crown back and forth. One company will have the superior performance for one or two generations, then the other company will take that lead. If you look here and now your wasting as much time as you are money. Buy the cheapest card you can that remains "fast enough" and upgrade it two or more times a year. Buying 1 big expensive card and hoping it lasts you 2 years might work, but its usually gunna hit its peak long before then. Premium cards never stay on their pedestal for more than a year, and high end cards are lucky to get half that. Upgrade to the best value and upgrade often.
*edit*
And to fuel the fire a bit, take a look at some other numbers, namely the 3870:
twrococ.com