Jon Peddie Research - a technically oriented marketing, research, and management consulting firm in the graphics industry.

Revisiting some previous results

Posted by Jon Peddie on June 18th 2007 | Discuss
Categories:
Tags:

We have two follow-up tests to share with you this week: the Toshiba docking station and the R600-Radeon HD2900XT.

Toshiba Dynadock—it’s totally powered

If you read our evaluation of the Dynadock (see TechWatch, June 4), you may recall we had, or thought we had, a problem with the rear panel USB connectors not having power, and were forced to use the front USB slots for items that required powered USB. Well, that’s just not the case and we want to set the record straight.

We have verified that the Toshiba Dynadock does have powered USB ports in the back and front of the unit. The problem (for us) is that the nifty little 80-GByte HP Pocket Media external disk drive that we use for minute-to-minute backup (using NTI’s wonderful Shadow program) is a finicky little thing and very particular about where it’s plugged in. If, for example, we plug it into a Belkin USB 2.0–powered external dock, almost everything we have will work in that hub, but not the HP Pocket Media drive. The HP drive will only work when plugged into the front jacks of the Dynadock, or one—and a particular one—of the three USB ports on the HP Compaq nw8240 laptop, and that’s the port we use for the Dynadock. So we were incorrect in our assumptions about the Dynadock; it’s just fine, as are the Belkin USB hub and the other ports on the HP laptop.

AMD R600 AA, CF, and not

While visiting Taipei last week we spoke to some guys who had spoken to some guy who had told them that AMD’s R600’s clever big and little tent AA technique was a performance decelerator. Now, AA is naturally a degrading process in that extra processing steps that need to be run. However, this guy who told the guys who told us said that one of the reasons the R600 doesn’t perform better, especially given all the processors it has, is because they didn’t get the AA right. Well, we wondered so we ran some tests.

We have two Radeon HD2900XTs on a Vista Ultimate system here at MTTL powered by an AMD Athelon dual-core AMD 64X2 5200+ 2.6 GHz, and we ran the tests as shown in Table 1 (below) to see if we could support the conclusion. We ran three resolutions with AA off and AA on, set at 8x and display quality = 0. Our results are shown below.

	3D Mark R600	HD 2900 ST	Catalyst 7.5
Resolution	1024 x 768	1280 x 1024	1600 x 1200
Crossfire AA (8)	9961	9632	8704
Crossfire AA-off	10566	10500	10139
Diff	6.1%	9.0%	16.5%
No Crossfire (8)	7379	6132	5026
No Crossfire AA-off	9446	8884	8329
Diff	28.0%	44.9%	65.7%
Crossfire on diff AA	35.0%	57.1%	73.2%
Crossfire on diff AA off	11.9%	18.2%	21.7%

Table 1. AMD AA and Crossfire on/off comparisons

As the table shows, the maximum differences are found when Crossfire is not enabled, and presumably reveals the intrinsic limitations of the R600’s AA efficiency.

However, just to be sure, we also ran some tests on an Nvidia G80, GeForce 8800GTS system, which has an AMD Athlon 64 FX 60 2.6 GHz dualcore CPU (Table 2).

3D Mark R600	Nvidia GeForce 8800GTX		ForceWare 158.22
Resolution	1024 x 768	1280 x 1024	1600 x 1200	2580 x 1600
SLI on AA on 8x	10107	9549	8680	5120
SLI on AA off	10396	10214	9935	9237
Diff	2.9%	7.0%	14.5%	80.4%
SLI off AA on 8x	8531	7211	5892	3012
SLI off AA off	9651	9302	8828	6773
Diff	13.1%	29.0%	49.8%	124.9%
SLI on dif AA off	18.5%	32.4%	47.3%	70.0%
SLI on dif AA on	7.7%	9.8%	12.5%	36.4%

Table 2. Nvidia AA and SLI on/off comparisons.

This system has the Dell 30-inch 2580 x 1600 display on it so we were able to test for higher resolutions.

To evaluate the claim that AMD didn’t get the AA trick right we compared the two AIBs (HD2900Xt and GeForce8800GTX) to see if any major difference would appear and, as Table 3 (below) shows, it did.

	1024 x 768	1280 x 1024	1600 x 1200	Average Dif
Dual AIB AA on	1.5%	–0.9%	–0.3%	0.1%
Dual AIB AA off	–1.6%	–2.7%	–2.0%	–2.1%
Single AIB AA on	15.6%	17.6%	17.2%	16.8%
Single AIB AA off	2.2%	4.7%	6.0%	4.3%

Table 3. AMD–Nvidia comparisons.

Nvidia GeForce beats the AMD Radeon by an average of almost 17%. However, on all other comparisons the HD2900XT holds up well, and in some cases beats the GeForce8800GTX, as Table 3 shows.

Call of Juarez DirectX 10 benchmark

Techland (Wroclaw, Poland, http://www.techland.com.pl/) made their final, official version of a stand-alone version of their “Call of Juarez DirectX 10 Benchmark” available to us and we ran it on the above-mentioned Vista system with AMD Radeon HD2900XT AIBs. Running the test at three resolutions with MSAA set to 2x, we got the results shown in Table 4 (below).

You can also get a demo of the game at http://techland.pl/pl/produkt.php?readmore=80 (click on COJ SP demo in the box on the right of the page).

---