Tech-Hounds.com

The Radeon X800XL and GeForce 6800GT Compared - Part 2

PCI Express x16 graphics cards
Chipset Radeon X800XL 256 MB, 256 bit, GeForce 6800GT 256 MB, 256 bit

In the first part (which you can read here), we looked at performance numbers from both the Radeon X800XL and GeForce 6800GT. We conclude that the Radeon X800XL is the better buy, because it delivers more effective anti aliasing and it consistently has less of a performance hit with AA and AF than the GeForce 6800GT. If you look at the numbers, you can see the GeForce 6800GT experienced quite a significant hit when AA and AF is enabled. However, this also means the GeForce 6800GT has a far higher fps than the Radeon X800XL when running without AA and AF. Normally, this would not be a problem, but if you're getting very low frame rates like we're getting in Brothers In Arms, you might want to consider running without AA and AF. In this situation, the GeForce 6800GT is the 'better' card. Why? Because it offers higher average and minimum fps, something essential for fluid gameplay. A difference of 4 fps may not seem much, but when frame rates are dropping below 30 fps, your gameplay will suffer so even a 4 fps difference is noticeable (and significant).

Out of our suite of benchmarks, only one game (Brothers in Arms) exhibits very low average frame rates on the Radeon X800XL (and the GeForce 6800GT). Of course, this have left us wondering, will we see a similar trend in newer games? New games like F.E.A.R, Quake 4, Serious Sam II and Call of Duty 2 are made with next generation graphics in mind, so they can very well behave the same way. If we look back, both new games in our benchmark suite (Brothers in Arms and Battlefield 2) is really only playable at 1024 x 768 with AA and AF or 1280 x 960/1024 without AA and AF. So, we're putting the cards to the test again in F.E.A.R, Quake 4 and Serious Sam II.

At the same time, there's a few other things we want to look at. For one, we've managed to get our hands on a faster processor: the Athlon 64 3500+. Both ATI and NVIDIA also released new drivers, the Catalyst 5.10 and Forceware 81.85. So, in addition to testing the new games, we took some time to re test our usual suite of benchmarks to see just how much performance these cards offer with a 400 MHz increase in processor clock and newer drivers.

F.E.A.R

Gamers everywhere are in awe of F.E.A.R, which is definitely the heaviest fps on the market today. Unless you have a pretty high end system, you're not going to see the full image quality this game has to offer. Unlike other games that uses timedemos for benchmarks, F.E.A.R uses a performance test, you could try to tweak your settings to find the best compromise between frame rates and image quality. Even if you have a GeForce 7800GTX (or perhaps lucky enough to get the X1800XL or X1800XT), chances are you have to made some compromises - it's that heavy.



We've spent some time trying different settings and see how well they translate into the real world. Out of the various settings available, there are three things of interest: volumetric lights, soft shadows and texture / video resolution. Soft shadows is definitely the heaviest performance hit, making it a good choice for benchmarking. Unfortunately, it won't work with AA enabled. Since the image quality impact is minimal to say the least, we opted to leave soft shadows off in our benchmarks - this way we can test the game with both AA on and off. The second heaviest feature is volumetric lights, we also opted to leave this to off. From the entire performance test, the only noticeable difference in image quality can only be seen in the very beginning of the test (where the lights are shining through the moving ventilation fan). This feature does work with AA and AF, but the performance impact is too heavy for both the GeForce 6800GT and Radeon X800XL. The last option we took note is texture / video resolution. We took some screenshots that you can see below, comparing each settings (minimum, medium and maximum).



You can definitely see both the medium and maximum screenshots have more detail than the minimum screenshots. However, medium and maximum differences are not easily noticeable - an acceptable image quality compromise. While maximum settings provide the best image quality, it's also very hard on the graphics card memory - we'll have to wait for 512 MB cards to play at maximum settings. Performance differences between minimum and medium is non existent, so why not choose the better looking option? We chose to set the rest of the settings to maximum. Not only does it look better (take a look at the water screenshots below), but a quick test shows that there are not performance difference between medium and maximum setting.



To avoid any benchmark results discrepancies we ran FRAPS and compare the results we got with the built in performance test tool. The results are identical, within 1 fps of each other. We also found that F.E.A.R performance test do tend to fluctuate by 2 to 3 fps (especially for the minimum and maximum fps), so keep that in mind when you're looking at the benchmark results. To keep it graphics limited, we set all system options to minimum.

Quake 4

Quake 4 needs no introduction. The game uses the same engine as Doom 3 (but with larger environments) and also features a built in timedemo function. Unfortunately, as of the moment, timedemo replay is plagued by a bug causing some objects and textures to disappear during replay (you can read about it here). So, alas, we can't use a timedemo - the results will not be even close to gameplay. After much deliberation, we settle for the next best thing - playing the game in a repeatable manner and record the results with FRAPS. We settled on using the single player level - Data Processing Terminal. The results we're getting is consistent and there's lots of smokes, explosions and even some heat haze present. However, we have to remind you that this level is not really representative of the whole game. Basically, there's only yourself and the (huge!) tower guard that's casting dynamic shadows. If you remember, shadows is a big part of the Doom 3 engine. They're still present, but not in (over) abundance. Think of this as a temporary solution, until Raven (or id) release a patch to fix the timedemo replay.

Serious Sam II

The third newest game to our benchmark suite is no stranger to the fps crowd, or the general gaming crowd for that matter. Who doesn't remember the kamikaze bomber? We'll Sam is back courtesy of the people from CroTeam, with a new upgraded engine. Unlike the last two games, Serious Sam II uses Direct3D with some new effects such as parallax mapping and HDR rendering support. We've decided to use a customized 'High' settings - we turn off anti aliasing and set anisotropic filtering to 0 for default testing and set AA to medium and anisotropic filtering to 16x for AA and AF performance testing. We left HDR off, since this feature will only work on SM 3.0 cards. So, the settings should provide nearly identical image from both cards.



Serious Sam II features a built in benchmarking tool and several demos, we chose the Greendale demo since it has the heaviest graphics load (and of course lots and lots of enemies, explosions, grass and smoke). Just like F.E.A.R, we ran the built in timedemo function and compare the results to FRAPS,. Again the results are nearly identical, within 1 fps of each other.

Performance

Since all these games feature application controlled anti aliasing and anisotropic filtering, we enable these settings in-game, either through the menu or console command. With application controlled anisotropic filtering, the MIP levels used by both cards should be fairly close and any difference in image quality is kept to a minimum. We did took some screenshots, but they're pretty similar so we chose to not display then (now). Performance testing using older drivers (Forceware 66.93) does not seem to indicate any optimizations in filtering or shader optimization in Forceware 81.85 (in other words, the results are identical). Without further ado, here are the results.

Our test setup
AMD Athlon 64 3500+ socket 939
2 x 256 MB Kingston KVR 3-3-3 PC3200 DDR-SDRAM
MSI K8N NForce 4 SLI motherboard
Maxtor DiamondMaxPlus9 80 GBs Serial ATA 8 MB buffer
ASUS E-616 DVD-ROM
450 watts ATX power supply

Windows XP Professional with Service Pack 2 installed
ATI Catalyst 5.10 reference driver
NVIDIA Forceware 81.85 reference driver
NVIDIA NForce 4 6.66 reference driver
Creative SoundBlaster Live! 24 bit 5.12.1.512 driver.
DirectX 9.0c

The results:

As you can see, performance between the game's medium and our custom settings is very close, and of course not noticeable in gameplay. The Radeon X800XL seems a little bit slower, but its still within the standard deviation of 2 fps between each run. The GeForce 6800GT is the better card here, partly because it can provide higher fps without AA and AF. The Radeon X800XL is still able to provide decent frame rates, but it's undeniably the slower card. With AA and AF enabled, both cards are neck to neck, but the frame rates are too low to have an enjoyable experience. A shame really, penalty with AA and AF on the GeForce 6800GT is just too much.

	GeForce 6800GT		Radeon X800XL
	Default	4x AA 16x AF	Default	4x AA 16x AF
<30 fps	0	23	0	22
30-45 fps	3	21	16	24
45-60 fps	23	5	19	5
60-90 fps	19	6	14	4
90-120 fps	6	0	6	0

By looking at the fps progress graph and the table above, we can see where each card spends the most time in the test. Without AA and AF, the GeForce 6800GT spends the most time between 45-90 fps while the Radeon X800XL spends most of its time evenly spread out between 30 to 90 fps.We can see, the lower frame rates occurred during the firefight sequence, thankfully the Radeon can still provide playable frame rates. Now, if we look at performance with AA and AF, we can actually see the Radeon spending more time (3 seconds worth) in 30 - 45 fps compared to the GeForce 6800GT. However, frame rates are too low, so we don't recommend using AA and AF with this game on both cards.


Test are run under High Quality settings, with AA and AF set through the menu or console command. Remember, Quake 4 uses the Doom 3 engine on top of OpenGL and not Direct3D. Understandably the Radeon X800XL is at a disadvantage here. So, it's quite surprising to see the Radeon X800XL came very close to the GeForce 6800GT - a mere 2 fps difference with AA and AF at 1024 x 768. Again we see too huge a performance hit on the GeForce 6800GT. Since both cards can still provide very playable frame rates at 1024 x 768 with AA and AF, we recommend this setting for best gameplay experience with both cards. However, you don't want to play Quake 4 with AA and AF above 1024 x 768, since minimum frame rates is already at the 30 fps mark on both cards.

	GeForce 6800GT		Radeon X800XL
	Default	4x AA 16x AF	Default	4x AA 16x AF
<30 fps	0	0	0	5
30-45 fps	0	19	0	14
45-60 fps	2	29	6	26
60-90 fps	21	22	20	21
90-120 fps	21	1	19	2

It's obvious that the cards are actually evenly matched in this game - both spent their time mostly between 45 to 120 fps - despite the higher average frame rates from the GeForce 6800GT. Even when AA and AF is enabled, they stay pretty close to each other (the X800XL does drop to 29 fps for at least 5 seconds while 6800Gt is slightly better at 31 fps). Looks like higher maximum fps from the GeForce 6800GT is again the culprit here - causing higher average frame rates..


We've included HDR test results (without AF) to give you some idea of what kind of performance you can expect with HDR enabled. Interestingly enough, HDR's performance penalty is smaller than AA and AF on the GeForce 6800GT. Even on default settings, the Radeon X800XL offers performance very close to the GeForce 6800GT. With AA and AF enabled, the X800XL leaves it's rival behind. But it's not a clean getaway - minimum fps are lower than the GeForce 6800GT. With higher resolutions, the GeForce 6800GT drops more performance, even in default settings, but its still maintain a higher minimum fps.

	GeForce 6800GT		Radeon X800XL
	Default	4x AA 16x AF	Default	4x AA 16x AF
<30 fps	1	2	1	1
30-45 fps	1	62	0	3
45-60 fps	14	0	42	59
60-90 fps	48	0	20	1
90-120 fps	0	0	0	0

Wow - what happened here? The GeForce 6800GT is now trailing behind - it can push more than 45 fps with AA and AF enabled. Without AA and AF, it's certainly the faster card before. So, it looks like the X800XL is a better match for this game - look at the minimum frame rates. Our concern over minimum frame rates was unfounded since it's just 1 second. It's more likely caused by disk access when first time loading the demo.

Out of these three titles, only F.E.A.R behaves in similar manner to Brothers in Arms, but F.E.A.R is actually a very different beast of a game. Quake 4 and Serious Sam II is more in line with our previous test results - the GeForce 6800GT is significantly faster without AA and AF, but the significant performance hit it has makes it slower (or about the same) than the Radeon X800XL when AA and AF is enabled. On both cards, you're likely only use 1024 x 768 with AA and AF or 1280 x 960 / 1024 without AA and AF.

There will be games that's just too heavy with AA and AF (on both cards). F.E.A.R is one such game. If you play games like F.E.A.R, you're better off with the GeForce 6800GT. However, we don't feel most games will behave much like F.E.A.R or Brothers in Arms for that matter. Although it's tempting to speculate that the reason for the GeForce 6800GT higher frame rates in F.E.A.R is shader related, we do not think so. The performance difference in F.E.A.R between the two cards are in line to differences in other games. And in all honesty, don't buy both cards to play F.E.A.R - you're better off with an even faster card such as the GeForce 7800GTX/GT or the Radeon X1800XT/XL if you want to max out all settings in that game.

Scaling Test

After looking at F.E.A.R, Quake 4 and Serious Sam II results, we like to discuss how both cards scale - just how much frame rate do we gain when we paired them with a faster processor. Traditionally, scaling test are limited to average fps, but this time, we want to go deeper by looking at the gains on the minimum, average and maximum side of the fps band.

Minimum fps gains are an indication that the cards are system limited, while maximum fps gains means the graphics card can still offer higher frame rates if you pair it with an even faster processor. Ideally, a card (or a benchmark) that scales well will have a maximum fps gain higher than it's minimum fps gain. If maximum fps gains are low, that means the card have already hit the fps ceiling - it will go no faster even with a faster processor. If minimum fps gains are low, then our platform is graphics limited - it would be better to choose an even faster graphics card (with a faster processor of course) to achieve a higher fps in that game / benchmark.

The test setup for our scaling test are the same as before, running our usual suite of benchmarks with both the Athlon 64 3000+ and 3500+.However, we used older drivers (Catalyst 5.9 and Forceware 78.01/66.93) on these tests. In some games (Brothers in Arms, SW: KOTOR, Battlefield 2), there will be negative values. While it may seem strange, it's actually quite normal. Variations between runs (since we're not using a replay or a battle record) may cause a lower frame rate, particularly when running without AA and AF, So these games do have a higher margin of error (around 5 - 10 % between each run). Don't look at the precise numbers, but try to spot trends - that's why we've included the difference for all resolutions with and without AA and AF.

Call of Duty

177.25

X800XL
Athlon 64 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	85	83	76	87	68	60
Avg	202.74	184.19	169.61	181.45	153.34	128.06
Max	443	327	326	364	317	275
Athlon 64 3000+
Min	72	74	71	74	64	57
Avg	164.26	153.12	162.64	140.82	120.99
Max	438	295	287	324	280	255
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	18.06%	12.16%	7.04%	17.57%	6.25%	5.26%	11.06%	12.42%	9.69%
Avg	14.38%	12.13%	10.77%	11.57%	8.89%	5.84%	10.60%	12.43%	8.77%
Max	1.14%	10.85%	13.59%	12.35%	13.21%	7.84%	9.83%	8.53%	11.13%

8.68%

6800GT – Quality
Athlon 64 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	92	92	83	93	95	70
Avg	228.54	210.58	184.05	181.07	132.39	96.14
Max	537	384	313	286	185	128
Athlon 64 3000+
Min	78	78	78	79	80	69
Avg	198.83	188.1	169.04	168.33	128.11	95.2
Max	439	365	288	281	185	128
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	17.95%	17.95%	6.41%	17.72%	18.75%	1.45%	13.37%	14.10%	12.64%
Avg	14.94%	11.95%	8.88%	7.57%	3.34%	0.98%	7.94%	11.92%	3.96%
Max	22.32%	5.21%	1.78%	0.00%	0.00%	6.33%	12.07%	0.59%

Call of Duty – High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	93	95	94	94	83	61
Avg	227.57	207.16	180.28	155.61	112.84	87.77
Max	538	365	310	268	175	122
Athlon 3500+
Min	81	77	81	81	80	61
Avg	199.6	186.98	166.37	148.01	110.76	81.59
Max	537	416	286	261	175	122
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	14.81%	23.38%	16.05%	16.05%	3.75%	0.00%	12.34%	18.08%	6.60%
Avg	14.01%	10.79%	8.36%	5.14%	1.88%	7.58%	7.96%	11.06%	4.87%
Max	0.19%	-12.26%	8.39%	2.68%	0.00%	0.00%	-0.17%	-1.23%	0.89%

In default settings (without AA and AF), we gain the most in minimum fps with the GeForce 6800GT under High Quality settings. But it's only under Quality settings we gain additional maximum' fps. So, it's very likely we can achieve higher fps with the GeForce 6800GT under Quality. However, in average fps, we can see the Radeon X800XL actually gain more fps since it's achieve both a higher minimum and maximum fps. So, it's pretty likely we can have a higher fps as well with the X800XL. Even more surprising, we can still gain more fps under AA and AF, while the GeForce 6800GT is definitely has hit the fps ceiling.

Splinter Cell

X800XL
Athlon 3500+
	1024	1280	1600	1024 AF	1280 AF	1600 AF
Min	59.28	45.55	42.19	58.17	44.14	36.87
Avg	85.34	68.46	58.55	85.04	57.96	50.51
Max	144.78	116.54	97.03	140.22	108.34	92.78
Athlon 3000+
Min	58.72	45.9	40.34	53.94	40.48	36.88
Avg	81.36	68.02	58.55	73.7	57.96	50.51
Max	127.74	116.9	93.8	126.57	108.36	92.62
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AF Difference (Averaged)
Min	0.95%	-0.76%	4.59%	7.84%	9.04%	-0.03%	3.61%	1.59%	5.62%
Avg	4.89%	0.65%	0.00%	15.39%	0.00%	0.00%	3.49%	1.85%	5.13%
Max	13.34%	-0.31%	3.44%	10.78%	-0.02%	0.17%	4.57%	5.49%	3.65%

6800GT - Quality
Athlon 64 3500+
	1024	1280	1600	1024 AF	1280 AF	1600 AF
Min	60.1	40.87	31.36	56.67	38.68	29.68
Avg	87.34	66.96	55.78	80.58	61.62	51.24
Max	151.2	125.73	97.25	151.81	112.99	94.75
Athlon 64 3000+
Min	47.88	40.47	32.11	58.1	39.57	30.59
Avg	86.08	70.18	59.4	83.31	65.85	55.33
Max	130.25	126.55	104.98	133.09	126.72	98.26
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AF Difference (Averaged)
Min	25.53%	0.99%	-2.34%	-2.46%	-2.25%	-2.97%	2.75%	8.06%	-2.56%
Avg	1.46%	-4.59%	-6.09%	-3.28%	-6.42%	-7.39%	-4.39%	-3.07%	-5.70%
Max	16.08%	-0.65%	-7.36%	14.07%	-10.83%	-3.57%	1.29%	2.69%	-0.11%

6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AF	1280 AF	1600 AF
Min	56.22	40.79	31.29	54.64	37.53	28.83
Avg	86.86	66.68	55.58	75.72	58.15	48.51
Max	152.12	114.67	105.66	150.41	111.31	91.53
Athlon 3000+
Min	60.61	41.51	32.07	55.97	38.58	29.82
Avg	85.77	70.19	59.17	79.92	62.58	52.62
Max	133.44	123.14	118.38	133.19	123.14	99.28
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AF Difference (Averaged)
Min	-7.24%	-1.73%	-2.43%	-2.38%	-2.72%	-3.32%	-3.30%	-3.80%	-2.81%
Avg	1.27%	-5.00%	-6.07%	-5.26%	-7.08%	-7.81%	-4.99%	-3.27%	-6.71%
Max	14.00%	-6.88%	-10.75%	12.93%	-9.61%	-7.81%	-1.35%	-1.21%	-1.49%

Looks like the GeForce 6800GT can go no faster, both under High Quality and Quality settings. We even see a slight slower score, but they are normal and still within the margin of error with this benchmark. Remember, we're not using Forceware 81.85 here, but 66.93. The Radeon fares a little better, but in all honesty its already close to it's maximum fps potential, particularly with AF enabled.

F1 Career Challenge

X800XL
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	67	67	67	62	58	45
Avg	100.47	100.72	99.96	96.21	79.85	63.16
Max	124	124	124	120	100	80
Athlon 3000+
Min	60	61	61	60	59	46
Avg	91.38	91.76	91.31	90.35	79.9	63.53
Max	113	114	113	112	100	81
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	11.67%	9.84%	9.84%	3.33%	-1.69%	-2.17%	5.13%	10.45%	-0.18%
Avg	9.95%	9.76%	9.48%	6.49%	-0.05%	-0.57%	5.84%	9.73%	1.95%
Max	9.73%	8.77%	9.73%	7.14%	0.00%	-1.23%	5.69%	9.41%	1.97%

6800GT - Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	61	59	60	61	60	46
Avg	91.02	90.97	90.72	90.91	87.54	69.35
Max	113	113	113	113	107	87
Athlon 3000+
Min	53	53	53	53	53	46
Avg	80.86	80.31	80.27	80.25	80.07	69.12
Max	101	100	100	100	100	87
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	15.09%	11.32%	13.21%	15.09%	13.21%	0.00%	11.32%	13.21%	9.43%
Avg	12.57%	13.27%	13.02%	13.29%	9.33%	0.34%	10.30%	12.95%	7.65%
Max	11.88%	13.00%	13.00%	13.00%	7.00%	0.00%	9.65%	12.63%	6.67%

GeForce 6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	61	60	60	60	52	40
Avg	91.61	91.07	90.63	90.18	75.14	58.52
Max	114	113	113	112	93	73
Athlon 3000+
Min	53	53	53	53	52	40
Avg	80.35	80.33	80.52	80.14	74.58	58.47
Max	100	100	101	100	92	73
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	15.09%	13.21%	13.21%	13.21%	0.00%	0.00%	9.12%	13.84%	4.40%
Avg	14.00%	13.37%	12.56%	12.53%	0.76%	0.09%	8.89%	13.31%	4.46%
Max	14.00%	13.00%	11.88%	12.00%	1.09%	0.00%	8.66%	12.96%	4.36%

The Radeon X800XL is again already very close to its fps ceiling. We see that more clearly in AA AF situations, the card can go no faster. The GeForce 6800GT under both Quality and High Quality settings still has room for improvement, since it gain more than 10 % - just about the the difference between itself and the Radeon X800XL.

Homeworld 2

X800XL
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	45	51	43	52	50	38
Avg	145.86	131.82	111.83	138.84	122.37	100.86
Max	355	351	350	358	353	354
Athlon 3000+
Min	27	30	28	31	30	25
Avg	136.25	125.5	110.34	131.77	117.33	98.7
Max	329	327	326	335	331	329
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	66.67%	70.00%	53.57%	67.74%	66.67%	52.00%	62.77%	63.41%	62.14%
Avg	7.05%	5.03%	1.35%	5.36%	4.29%	2.19%	4.21%	4.48%	3.95%
Max	7.90%	7.34%	7.36%	6.87%	6.65%	7.60%	7.29%	7.53%	7.04%

6800GT - Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	63	65	60	52	38	27
Avg	196.98	191.89	169.85	157.65	114.58	80.57
Max	386	385	382	334	217	140
Athlon 3000+
Min	33	37	36	34	25	17
Avg	173.65	170.12	159.51	155.64	114.26	80.41
Max	347	346	346	331	217	140
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	90.91%	75.68%	66.67%	52.94%	52.00%	58.82%	66.17%	77.75%	54.59%
Avg	13.43%	12.80%	6.48%	1.29%	0.28%	0.20%	5.75%	10.91%	0.59%
Max	11.24%	11.27%	10.40%	0.91%	0.00%	0.00%	5.64%	10.97%	0.30%

6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	66	65	58	51	37	26
Avg	198.22	189.64	163.16	150.26	107.52	76.4
Max	391	386	382	334	217	140
Athlon 3000+
Min	37	37	36	32	24	16
Avg	175.18	169.52	155.92	146.22	108.41	77.09
Max	353	346	343	331	216	140
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	78.38%	75.68%	61.11%	59.38%	54.17%	62.50%	65.20%	71.72%	58.68%
Avg	13.15%	11.87%	4.64%	2.76%	-0.82%	-0.89%	5.12%	9.89%	0.35%
Max	10.76%	11.56%	11.37%	0.91%	0.46%	0.00%	5.84%	11.23%	0.46%

Since Homeworld 2 is an RTS, it's understandably a system limited game. We gain so much just by increasing the processor clock by 400 MHz. The GeForce 6800GT has pretty much hit the ceiling when AA and AF enabled, but it can still go faster without AA and AF. The Radeon XL can still gain ground on both, but not as much as the 6800GT without AA and AF.

SW: KOTOR

X800XL
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	31.33	21.33	13.67	23.67	15.67	8
Avg	60.88	59.31	46.29	54.64	47.5	26.17
Max	74	69.33	62.33	68	61	35
Athlon 3000+
Min	31	21.33	15	20.33	18	15.33
Avg	54.71	51.59	44.17	47.56	46.2	38.07
Max	65	64	59	60	58.66	51
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	1.06%	0.00%	-8.87%	16.43%	-12.94%	-47.81%	-8.69%	-2.60%	-14.78%
Avg	11.28%	14.98%	4.81%	14.89%	2.80%	-31.27%	2.92%	10.36%	-4.53%
Max	13.85%	8.33%	5.64%	13.33%	3.99%	-31.37%	2.29%	9.27%	-4.68%

6800GT - Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	35.66	21.33	17.33	19	12.66	9
Avg	69.89	60.73	58.89	62.52	46.9	35.4
Max	93	90	88.66	90.5	65	48.67
Athlon 3000+
Min	33.66	23.66	17	17.67	12	10.67
Avg	78.33	58.71	51.8	52.38	41.81	34.19
Max	59.62	79.67	77	77	58	43.33
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	5.94%	-9.85%	1.94%	7.53%	5.50%	-15.65%	-0.76%	-0.65%	-0.87%
Avg	-10.77%	3.45%	13.69%	19.36%	12.17%	3.52%	6.90%	2.12%	11.68%
Max	55.99%	12.97%	15.14%	17.53%	12.07%	12.32%	21.00%	28.03%	13.98%

6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	30	21.66	16.66	18.33	13.66	9.33
Avg	66.82	61.6	60.81	56.41	43.62	33.06
Max	92.66	89.66	87.33	79.33	57.33	43
Athlon 3000+
Min	31	23	17	18	12	12
Avg	59.58	56.38	54.8	52.18	41.19	35.64
Max	80	79	78	78	58	43
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	-3.23%	-5.83%	-2.00%	1.83%	13.83%	-22.25%	-2.94%	-3.68%	-2.19%
Avg	12.15%	9.26%	10.97%	8.10%	5.91%	-7.23%	6.53%	10.80%	2.26%
Max	15.83%	13.49%	11.96%	1.71%	-1.16%	0.00%	6.97%	13.76%	0.18%

Contrary to what people thought, the minimum fps is not always caused by system limitations. We see here that we actually lost some fps or getting roughly the same level of performance even when we're using a faster processor. The negative scores are more likely caused by the testing method, but they are still within the standard deviation between runs in this game. Most interestingly, we gain some fps on average on all cards, quite possibly because of a higher maximum fps. Since the Radeon X800XL is less sensitive to AA AF than the GeForce 6800GT, it does not gain as much even when compared to the 6800GT under High Quality settings.

Brothers In Arms

X800XL
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	24.33	19.66	16	26	19.66	16.5
Avg	44.51	44.15	35.52	43.66	34.37	19.1
Max	74.33	65.33	56.33	74	60	49.5
Athlon 3000+
Min	27	22	16.33	24.33	16	14
Avg	50.88	42.3	34.44	41.83	33.18	18.52
Max	75.67	63.33	54	72	58.67	47
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	-9.89%	-10.64%	-2.02%	6.86%	22.88%	17.86%	4.18%	-7.52%	15.87%
Avg	-12.52%	4.36%	3.14%	4.38%	3.60%	3.13%	1.02%	-1.67%	3.70%
Max	-1.77%	3.16%	4.31%	2.78%	2.27%	5.32%	2.68%	1.90%	3.45%

6800GT - Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	35.66	24.66	18.33	31	21	14.5
Avg	90.16	64.93	47.31	71.39	50.49	24.42
Max	146.33	110.33	88.33	134.33	100.66	79.5
Athlon 3000+
Min	36	22.33	14.67	32	18	11.5
Avg	84.51	63.48	47.42	70.73	48.55	36.66
Max	140.67	113.33	88.67	132.66	103.67	81
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	-0.94%	10.43%	24.95%	-3.13%	16.67%	26.09%	12.34%	11.48%	13.21%
Avg	6.68%	2.28%	-0.23%	0.94%	4.00%	-33.38%	-3.28%	2.91%	-9.48%
Max	4.02%	-2.64%	-0.38%	1.26%	-2.90%	-1.85%	-0.42%	0.33%	-1.17%

6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	39.66	24	13.66	25.66	17	12.5
Avg	87.44	59.55	45.33	57.38	39.55	30.45
Max	143.33	112	89	130.66	98	79
Athlon 3000+
Min	39.33	22.33	15	27	17.67	11
Avg	83.52	61.36	45.23	58.09	39.99	29.79
Max	139.66	110.33	89	129.33	96	81
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	0.84%	7.48%	-8.93%	-4.96%	-3.79%	13.64%	0.71%	-0.21%	1.63%
Avg	4.69%	-2.95%	0.22%	-1.22%	-1.10%	2.21%	0.31%	0.65%	-0.04%
Max	2.63%	1.51%	0.00%	1.03%	2.08%	-2.47%	0.80%	1.38%	0.21%

The GeForce 6800GT is definitely nearing its fps ceiling here. It can only a higher fps on the minimum side. Interestingly, the Radeon X800XL gain more than the GeForce 6800GT under Quality and Higher Quality settings when AA and AF is enabled. Be careful though, fluctuations of scores is a little high with this benchmark. The X800XL seems 'stuck' and can not gain more fps except on less intensive parts of the game, perhaps a driver bug is interfering with its performance.

Battlefield 2

X800XL
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	65.66	58	43	45.33	36.75	27.66
Avg	78.59	66.03	51.26	53.8	42.15	32.74
Max	87	71.66	57	62.33	46.75	37
Athlon 3000+
Min	64.75	56.25	39.75	47.5	39.25	24.3
Avg	80.67	65.49	52.27	55.83	45.52	32.19
Max	90.25	73.25	61.75	64.25	52	36.67
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	1.41%	3.11%	8.18%	-4.57%	-6.37%	13.83%	2.60%	4.23%	0.96%
Avg	-2.58%	0.83%	-1.93%	-3.63%	-7.40%	1.72%	-2.17%	-1.23%	-3.11%
Max	-3.60%	-2.17%	-7.69%	-2.99%	-10.10%	0.90%	-4.27%	-4.49%	-4.06%

6800GT - Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	52	50.5	29.5	37.75	24	10.33
Avg	79.2	63.09	46.96	45.43	32.85	23.6
Max	90.8	74	57	51.75	38.33	28
Athlon 3000+
Min	54	51.33	36.33	28.5	27.5	14.25
Avg	68.13	64.72	47.19	43.84	33.14	22.88
Max	77.42	73.66	56	51	37.5	28.75
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	-3.70%	-1.62%	-18.80%	32.46%	-12.73%	-27.51%	-5.32%	-8.04%	-2.59%
Avg	16.25%	-2.52%	-0.50%	3.63%	-0.89%	3.14%	3.19%	4.41%	1.96%
Max	17.28%	0.46%	1.79%	1.47%	2.21%	-2.61%	3.43%	6.51%	0.36%

6800GT - High Quality
Athlon 3500+
	1024	1280	1600	1024 AA AF	1280 AA AF	1600 AA AF
Min	65.5	46.5	32.66	25	18	12
Avg	79.63	59.68	43.82	29.9	22.68	16.88
Max	89.75	69.75	52	34.75	27	21
Athlon 3000+
Min	60	49.5	24	24.5	17	8.5
Avg	75.46	61.38	49.64	29.45	22.7	19.26
Max	85.33	72.75	60.33	35.25	27	25.5
Difference							Total Difference (Averaged)	Default Difference (Averaged)	AA AF Difference (Averaged)
Min	9.17%	-6.06%	36.08%	2.04%	5.88%	41.18%	14.71%	13.06%	-16.37%
Avg	5.53%	-2.77%	-11.72%	1.53%	-0.07%	-12.35%	-3.31%	-2.98%	-3.63%
Max	5.18%	-4.12%	-13.81%	-1.42%	0.00%	-17.65%	-5.30%	-4.25%	-6.36%

Things are not looking good for the GeForce 6800GT in this game. It can only gain more fps on maximum and average settings. It can go no faster than this, although there's still the possibility of a higher minimum fps under High Quality settings. While the Radeon X800XL has also pretty much hit the ceiling here, it has way higher frame rates than the GeForce 6800GT, especially with AA and AF.

	X800XL			6800GT Quality%			6800GT High Quality
	Total	Default	AA AF	Total	Default	AA AF	Total	Default	AA AF
Min	11.52%	11.71%	11.33%	14.27%	16.56%	11.98%	13.69%	15.57%	11.81%
Avg	3.70%	5.13%	2.27%	3.77%	6.02%	1.52%	2.93%	5.64%	0.22%
Max	4.01%	5.38%	2.64%	6.70%	10.46%	2.95%	2.21%	4.66%	-0.25%

You can see the average total gains we see from both cards in the table above. The Radeon X800XL definitely can offer you higher gains with a faster processor if you play mostly with AA and AF. For those looking for 'pure' performance, the GeForce 6800GT is a better match. It can only gain as much as the X800XL with AA and AF under Quality settings. Of course, by using Quality settings, you get a slightly lower image quality (and unfortunately, texture shimmering under Quality settings in Battlefield 2 is still present in Forceware 81.85),

Other Issues

Power

One issue we encounter when testing various GeForce 6800GT is power. Since it draws power from the 12 volt rail not to mention the PCI Express slot, it's very sensitive to power fluctuations. Those with lots of 12 volts peripherals (hard drives, DVD / CD-ROMs) would have to make sure their power supply can withstand the additional load and provide stable output (in amps, not voltage). We actually used two power supply in our GeForce 6800GT round up because of this. With the default power supply we've used, whenever we try to access more than two IDE devices (two hard drives and a DVD-RW dual burner) at the same time, the additional drain from the DVD-RW spinning up would cause a blue screen and forced a restart with all of the GeForce 6800GT. However, this problem did not occur even once during our long hours of gameplay and testing. None of the Radeon X800XL cards we tested is affected by this problem.

Noise

Unfortunately, we don't have a sound meter device, so we can't show you noise test results from these cards. However, overall the Radeon X800XL emits a more noticeable noise than the GeForce 6800GT. It's not loud, just noticeable in an open case environment, if you're using a closed case, the noise would not be as noticeable (except for the GeCube Uniwise - it did have a higher pitched noise).

Some Forward Looking Statements

If increasing processor clock speeds by 400 MHz (roughly between 1/4 to 1/5) or performance rating from 3000+ to 3500+ nets us an average increase of 11 to 14 % increase in frame rates only on the minimum side, we can safely assume that most games are system limited on these cards. The best processor to pair these cards with are those with 10 % higher clock or performance rating (an Athlon 64 or FX) running at 2.4 GHz or about 3800+ to 4000+. Of course, this makes us wonder - what's the point of buying a faster graphics card like the GeForce 7 or Radeon X1K series? Even with the fastest processor available today, these cards will be system limited, at least until new, faster clocked processors arrive. Unfortunately, both Intel and AMD will likely focus more on dual core and multi core solutions than higher MHz in the future.

Well, since we are undoubtedly system limited with cards such as the GeForce 7 and Radeon X1K series, the deciding factor would be higher frame rates with AA and AF enabled and / or higher frame rates with games that are graphically limited, quite possibly shader heavy. In this sense, AA and AF are 'free' because we can't push frame rates any higher (sans overclocking the processor). Developers are more aware than this, so newer games are making use of effects and shaders more heavily than they did a few years ago. That's why shader and AA / AF efficiency is more important than ever with these games. In respect to the GeForce 6800GT and Radeon X800XL, they are on opposite sides of this issue - the GeForce 6800GT with SM 3.0 support and the Radeon X800XL with less hit when running with AA and AF.

While this is good (new, more detailed graphics from games) a long term solution is badly needed to address system limitations on next generation graphics cards. Not only do they have to be very shader efficient and have less performance hit with AA and AF, they must also scale with dual and multi core processors. Thankfully, both ATI and NVIDIA have begun researching ways to multi thread their drivers on dual core processors.Both vendors have release beta versions of multi threaded drivers. However, note that these drivers are beta and it will take some time before they are ready for widespread use and provide tangible, significant performance compared to current drivers.

Conclusion:

Both cards offer the same level of performance, each with it's own advantages and disadvantages. If you don't like playing with AA and AF, the GeForce 6800GT should be your choice. Even if you decided to change you mind, the 6800GT can still provide playable frame rates with AA and AF with new games. For gamers who fuss more about image quality, compatibility and frame rate consistency, the Radeon X800XL is more appropriate. You get more effective anti aliasing - thus overall better image quality and with less of a performance hit. An added bonus is the tendency to scale even with AA and AF - something the GeForce 6800GT also lacks.

Go to top