Aquamark 3 in B3D Reviews
- Thread starter Dave Baumann
- Start date
horvendile
Regular
IMHO, Serious Sam can still be used to extract quite a lot of information, even if it's only about old-school rendering. And framerates are not ridiculous with AA and a mid-range card, even at medium resolutions (10x7). As far as I can recall Wolfenstein is not as flexible and versatile in its benchmarking.
What I would like to see is B3D using as many benchmarks as possible. I know that time might not allow this but perhaps there is a practical solution. I feel that B3D is beyond reproach with respect to its reviews and is in this day and age it seems that it is in increasingly small company. However, there are some sites that also do a very good job. While they may not have the technical savvy of B3D they are honest and more than competent. NORDIC HARDWARE and TECH REPORT come to mind. If B3D and sites like these can pool benchmarks, preferably using the same or similar test bed, it would give readers more information while alleviating workloads.
Ok guys, everyone that says Quake3 based sucks, why don't you go out and get excel and make a graph of cards versus Q3 performance, and then compare it to other things so we can actually see how bad it is in regards to performance predictions. Not to mention many games use the engine and not benchmarking with it seems counterintuitive.
By otherthings I guess perhaps futuremark performance or something might work... I am not sure but I think it would be worth a try.
Alternatively just turn on shadows in Q3 based games, and turn up the partions of polygons and what not, insteda of using it on 16bit mode getting 400fps or something.
By otherthings I guess perhaps futuremark performance or something might work... I am not sure but I think it would be worth a try.
Alternatively just turn on shadows in Q3 based games, and turn up the partions of polygons and what not, insteda of using it on 16bit mode getting 400fps or something.
Quake3 scales very nicely. This is becuase it is of relatively low poly counts and CPU/system performance (for its poly requirements) has scaled well with the power of 3D graphics. The problem is that Q3 doesn't use much, if any, hardware vertex processing (T&L) making it very dependant on the CPU performance. With later Q3 based games they have continued to increase the poly count in line with the increases of CPU power, which means that many Q3 games are totally CPU dependant - you can see that RtCW:ET is now around the 60FPS mark whereas RtCW (1 or 2 years before it) is about 120FPS but they are both Completely limited by the CPU on modern boards.
Reverend said:
I believe that part of a reviewers job is to verify that a graphics card (and drivers) operate under DirectX and OpenGL. Unless a driver processes requests from DirectX and OpenGL using the same code path, I would continue to include a DirectX and OpenGL application in a test suite. I'm pretty sure that NVIDIA has two driver development teams and their drivers have separate dll's for DX and OGL.
No. First off, sound isn't used in any of my reviews, so thats a non issue. The demos are also multiplayer demos so there would be no AI involved either - besides, even in the single player game I doubt there would be any AI changes between RtCW and RtCW:ET.
AFAIK T&L was only ever used in the areas with subdivisions on the Q3 engine (i.e. the curved bits) and titles like RtCW probably don't even use this. The increase in CPU requirements is almost certianly down to increasing poly counts.
AFAIK T&L was only ever used in the areas with subdivisions on the Q3 engine (i.e. the curved bits) and titles like RtCW probably don't even use this. The increase in CPU requirements is almost certianly down to increasing poly counts.
Dave what about the other thing I said? About using a non-standard config that pushes the engine much farther, I find it slows down my system appreciably, I am sure that somewhere we could find a config that is tweaked enough to bring down the newer cards to reasonable levels.
Or is it too CPU bound to be worth trying, as in even tho it will slow it down appreciably it is only the CPU slowing down and not the GPU?
Or is it too CPU bound to be worth trying, as in even tho it will slow it down appreciably it is only the CPU slowing down and not the GPU?
Hi Dave,
I think you need to decide what kind of information you want to pass on to the reader. In the case of splinter cell, does the inability to benchmark AA matter to you? Will RTCW provide the kind of data that you want? Are you trying to focus on newer games, older games, ogl, dx, FSAA, aniso, or all of them? Having a sample which provides atleast something of each (which it seems is what you are trying to do) can be nice, so I guess you need to figure out what each benchmark brings to the table. What kinds of things will AM3 provide, and will it augment the other titles in your suite in ways you like? Are you planning on replacing another title soon when max payne 2 comes out? How about HL2 and D3? Think about how your benchmark suite would look in 6 months, and think about the order in which you'd remove older titles. What do you want to eventualy end up with?
Nite_Hawk
I think you need to decide what kind of information you want to pass on to the reader. In the case of splinter cell, does the inability to benchmark AA matter to you? Will RTCW provide the kind of data that you want? Are you trying to focus on newer games, older games, ogl, dx, FSAA, aniso, or all of them? Having a sample which provides atleast something of each (which it seems is what you are trying to do) can be nice, so I guess you need to figure out what each benchmark brings to the table. What kinds of things will AM3 provide, and will it augment the other titles in your suite in ways you like? Are you planning on replacing another title soon when max payne 2 comes out? How about HL2 and D3? Think about how your benchmark suite would look in 6 months, and think about the order in which you'd remove older titles. What do you want to eventualy end up with?
Nite_Hawk
OpenGL guy
Veteran
I believe Q3 always used the OpenGL transform pipe, hence, it's using HW transform when available.DaveBaumann said:
I don't really appreciate Unreal engine benchmarks for the simple fact of negative LOD bias (in default shipping config). It kills texture caches. It wastes memory and bandwidth. UT2k3 isn't the primary playground for texture filtering ch... *cough* optimizations for nothing.
As nice as it is as a game (and popular, too), it's just not representative of any rendering workload except itself. It isn't a sane workload at all. The sooner it leaves the spotlight, the earlier this filter optimization madness will stop. It probably just can't be ditched because there's too much consumer interest :?
/me crosses fingers that UT2k4 finally gets it right.
Quake3:
The good thing about it is that it's nice and easy to push any card to its limits, given high enough resolution. It has really become some sort of real world textured fillrate test.
IMO Villagemark is a good enough replacement, and more stressful. So here's my vote (well, sort of).
As nice as it is as a game (and popular, too), it's just not representative of any rendering workload except itself. It isn't a sane workload at all. The sooner it leaves the spotlight, the earlier this filter optimization madness will stop. It probably just can't be ditched because there's too much consumer interest :?
/me crosses fingers that UT2k4 finally gets it right.
Quake3:
The good thing about it is that it's nice and easy to push any card to its limits, given high enough resolution. It has really become some sort of real world textured fillrate test.
IMO Villagemark is a good enough replacement, and more stressful. So here's my vote (well, sort of).
Well then, I don't know the real answer as to why all post-Quake3 (the game) that uses the Q3-engine has all been CPU limited.DaveBaumann said:
I still do not believe that the higher poly counts is the reason though because :
1) There is no way in hell that the poly counts, and poly counts alone, are high enough that makes these games CPU limited.
2) Quake3 engine (or rather Carmack) did not mess with the default/built-in OpenGL matrix transformation support. I don't know if the engine licensees messed with it, but if they did, they're idiots but that could be the reason (can anyone confirm this?). Well, maybe that's too harsh a desciption since there are so many types of transform commands and, well, developers have been known to have strange reasons for what they aim for
.I particularily like both the benchmark and the company, Massive have gone out of their way to deliver and excellent benchmark - answer most (not all) demanding questions, and deliver a really useful performance database and search engine. I have found AM3 exceedingly useful in setting up and tuning my new rig and detecting problems.
Beauty is in the eye of the beholder. I like the fact the Krass engine has been around for a while since it first showcased the Geforce 3's capabilities. I like the fact its a balanced CPU / GPU measurement tool and it has shader fallback strategies (like Half Life 2) for older video cards.
Beauty is in the eye of the beholder. I like the fact the Krass engine has been around for a while since it first showcased the Geforce 3's capabilities. I like the fact its a balanced CPU / GPU measurement tool and it has shader fallback strategies (like Half Life 2) for older video cards.
Reverend said:
I should imagine its the number of and detail of the character models, as these wouldn't go through the Transformation pipe I guess.
