ATI Radeon X1600 XT and X1300 PRO

[Dave Baumann]

<a href="http://www.beyond3d.com/reviews/ati/rv5xx/"><img border="1" src="http://www.beyond3d.com/reviews/ati/rv5xx/images/focus.gif" align="right" width="120" height="75"></a>Back in October '05 ATI introduced the initial, much belated, X1000 series product line, comprising X1800 (R520), X1600 (RV530) and X1300 (RV515). Although these were all introduced at the same time their releases were staggered a little, and even when they did finally start rolling off the production lines their further releases hardly appeared to go too smoothly.

Both X1300 and X1600, although around in various locations roughly when they were supposed to be, hardly appeared plentiful, which may partially have been to do with their pricing. ATI set fairly high MSRPs on both the X1300 and X1600 product lines, likely set for some high margins and to be competitive to where they were when they were introduced. However, when shipping is set to occur nearly two months after announcement, as was the case for X1600, a lot can happen in that time span, as indeed it did with NVIDIA's release of the 6800 GS. Now, though, following a reduction in MSRPs by ATI both products are becoming plentiful, vendors are coming out with numerous SKUs of both product lines, and street prices are more inline with where we would expect for their die sizes.

<a href="http://www.beyond3d.com/reviews/ati/rv5xx/">Click here to read our X1600 XT and X1300 PRO Review</a>

[tEd]

I mentioned it once already. Int16 blending is not supported on any r5xx chip , you may wanna correct that.

[Ali]

Very good article, but one point - You might want to do a search and replace for dROPs to drops.

Also, if you still have the X1600XT CF setup, could you look into its multi-monitor capabilities? Ie, can you run 4 screens off it, and then game off one screen with the speed bost of CF without having to reboot or anything?

Thanks

Ali

[AlStrong]

Page 1:

although in this instance RV515 has double Z capabilities

Should be RV530 that has double Z?

[Geo]

Appreciated as always. Thanks particularly for the AF/HQ/AAA testing.

This certainly isn't the only example of what I'm pointing at in your results, but it will do: http://www.beyond3d.com/reviews/ati/rv5xx/chart/image017.gif

Now, is it just me. . .or is there a line on that graph that is doing everything but waving and spelling out in semaphore "The driver team hates me!"? If so, is ATI going to do something about it?

[Blacklash]

Thank you for the review,

I am sure there are some interested in mainstream cards for crossfire simply to be able to use the antialiasing supersampling modes. I know adaptive AA adresses some of this and not where opaque textures are concerned. So 10X Crossfire AA may be of interest to some that game around 1280x. Hopefully there will be a slightly more powerful mid-high card in ATi's future.

Honestly, I am surprised many have bemoaned the performance of single X1600s yet haven't bother to try them in crossfire to see how they behave.

The only thing I miss when using a single ATi card, is SSAA on opaque "interior" textures.

Once you have gamed with ATi's HQ AF at 16x there is no going back. The higher filtering extended beyond horizontal textures to vertical and near vertical is wonderful. The competition do fine in the horizontal department and get blown away when comparisons are made outside of that arena, particularly when what I have qualified is combined with great depth of field.

Anywho, thanks again for taking the time to do this when many others have not bothered to see how the entry and mainstream cards do in crossfire at 1280x.

[mczak]

Do these cards really have dual dual link dvi? I'm pretty sure both the x1300 and x1600 only have one dual link capable dvi port (and good luck finding _any_ x1300 with more than one dvi port anyway). (ATI's website explicitly mentions one dual-link + one single-link for x1300, it is not accurate enough on x1600 (just mentions "dual-link dvi") and explicitly mentions dual dual-link on x1800.)

And btw I'd be interested in reviews of the low-end versions - as far as I can tell reviews always cover x1600xt and x1300pro whereas the numbers of cards sold with x1600pro and x1300 probably far exceed those (though I'd suspect it has something to do with card companies not wanting to have the cards they really sell in volume actually tested, and heaven forbid maybe even the 64bit mem interface / hypermemory versions so they never provide any samples).

Otherwise, excellent stuff as usual. Nice to see some crossfire numbers, though you really have to wonder what's the point of it on mid- and especially low-end cards...

[Mintmaster]

I mentioned it once already. Int16 blending is not supported on any r5xx chip , you may wanna correct that.
Really? What were all those ATI sell sheets (http://www.techpowerup.com/reviews/ATI/R5XX/images/X1300-2.jpg) talking about then?

Are you sure it's not just a driver bug? I'd sure like OpenGL guy to clear this up for me. Dave, do you think you could use your connections to find this out for sure? I'd really appreciate it, as I have an algorithm planned that sort of depends on it.

[Jawed]

See page 25 of the Radeon X1x00 Programming Guide in the October 05 SDK, which indicates that FX16 isn't blendable. But it is filterable.

Jawed

[DPGX2]

Awesome article as always guys, you consistantly have some of the best written tech articles on the internet. I just have a few questions if anyone has a moment. First I've heard rumours that most of the non pro x1300's being shipped are actually pro's with downclocked cores / memory (mine has the same memory as the pro I believe). Second Im just curious if anyones tried running either of these boards in crossfire mode on an Nforce4 SLI mobo. I dont see any hardware limitation that'd cause it to not function. Given the fact that it sounds more like a software solution than hardware I dont see a reason that any mobo with 2 * PCIE Slots shouldnt work with this. Reason being I plan on picking up a second x1300 if this is the case, I'd just hate to purchase it and find out for some reason my assumption is incorrect. Thanks!

Mike

[tEd]

Really? What were all those ATI sell sheets (http://www.techpowerup.com/reviews/ATI/R5XX/images/X1300-2.jpg) talking about then?

Are you sure it's not just a driver bug? I'd sure like OpenGL guy to clear this up for me. Dave, do you think you could use your connections to find this out for sure? I'd really appreciate it, as I have an algorithm planned that sort of depends on it.

that's seems to be a mistake in the sheet.

as jawed said in their programming guide it clearly saiys no int16 blending and it's also not exposed in the driver

[dizietsma]

that was very interesting. I was very surprised that X1600XT beat X700Pro in UT2004, I thought that would be an early victory for the older card and then for it to lose everything else.

Just looking at the low resolution scores ( to try and avoid bandwidth issues ) if the 1600 is the same core speed as the 1300 but has 3 times the shading and also that r580 is supposed to have 3 times the shading of the r520 then can you get a picture of how well the r580 is going to do ? That assumes no clock speed increases of course.

I think it is looking very good for X1900 at present.

Dave, do you have COD2 to test with ? I'd be interested to know how the 1300 and 1600 perform both at medium textures and "Extra" textures because at that level COD2 looks very much better. I asume here that textures mean textures and not some sort of shading going on.

[pc999]

Great article.

Will X1700 be a 8-1-3-2, right it should be a very nice part (at 250)considering this one at least in a ratio of performance/price/longevity.

[Tahir2]

I think this review pretty much confirms that pixel shading power is essential for increasing performance in the low-mid range.

[ERK]

The performance difference between the X1600 and the X700 is increased a little under these conditions[AA/AF], indicating that the shaders in F.E.A.R. are sufficiently long to enable the new batch processor on X1600 to effectively hide the extra texture demands of 4x AF.
Good insight, Dave.

[thana]

page 6, third paragraph:The X700, on the other hand, has a 1-to-1 pixel shader to texture / ROP mapping though, whereas X1600's is 3-to-3, so combined with the clock speed the X1600 XT ends up with twice the Pixel Shader performance than the X700 PRO. you mean 3-to-1 here, don't you?


page 9, third paragraph:The performance differences between the X1600 and X700 is similar as well, however what's interesting is that despite its much higher texture rate, only in one of the shorter shaders is the X700 actually faster than the X600. X1600 here


same page, last paragraph:When comparing the performance differences between the X1000 boards and the X700 and X600 we an see that many of the gains are greater than the theoretical differences,can


page 14, last paragraph:The relative performance differences between the X1600 and X700 are again extended by a little in these settings, with the X1600 having up to an 88% performance advantage.should be 78%?


page 19, second paragraph below SC:CT:Unsurprisingly X1300 is gaining more in Crossfire mode than X1600 is, which is probably due to it performance being lower in the first place and the X1600 in Crossfire mode is limited by the bandwidth over the PCI Express bus.it's


same page, last paragraph: With 4x FSAA and 8x AF enabled the relative performance gains are for Crossfire to single board rendering are very similar to the previous test scenario, albeit with the average FPS performances being a little lower. In this instance the X1300 PRO Crossfire is mostly performing above a single X1600 XT.


page 20, the last sentence lacks a .


last page, third paragraph:Although we haven't given it a full test here we can see that the Crossfire gains for X1600 in with this solution can yield fairly tangible results, with performance gains up to nearly 70% in Splinter Cell, at least, however in other games its clear that it would benefit from increased bandwidth between the two boards, which is widely expected to come with ATI's next chipsetIMHO the sentence should be split up in two parts. as it is now it's very confusing (at least for me).


next sentence:Of course, that raises other issues with Crossfire such as it being platform specific and to take benefits of it most users will likely need a new main board, pushing the costs up for these entry and mainstream level solutions.mainboard

[Fodder]

it'sNo apostrophe. Good ol' English, inconsistency strikes again. :???:

[chavvdarrr]

<table style="border-collapse: collapse;" border="1" bordercolor="#111111" cellspacing="0" width="96%"><tbody><tr><td class="articledate" colspan="2" rowspan="2" align="center" width="30%">GL_Reme (FPS)</td> <td class="articledate" rowspan="2" align="center" width="14%">Back to Front</td> <td class="articledate" rowspan="2" align="center" width="14%">Front to Back</td> <td class="articledate" rowspan="2" align="center" width="14%">Random</td> <td class="articledate" colspan="2" align="center" height="31" width="28%"> % Improvement from Back to Front</td> </tr> <tr> <td class="articledate" align="center" width="14%">Front to Back</td> <td class="articledate" align="center" width="14%">Random</td> </tr> <tr> <td rowspan="2" align="left" width="16%">X1600 XT</td> <td align="left" width="14%">Overdraw Factor 3</td> <td align="right" width="14%">825.6</td> <td align="right" width="14%">2296.7</td> <td align="right" width="14%">1316.4</td> <td align="right" width="14%">178.2%</td> <td align="right" width="14%">59.4%</td> </tr> <tr> <td align="left" width="14%">Overdraw Factor 8</td> <td align="right" width="14%">306.8</td> <td align="right" width="14%">1846.4</td> <td align="right" width="14%">834.4</td> <td align="right" width="14%">501.8%</td> <td align="right" width="14%">171.9%</td> </tr> <tr> <td rowspan="2" align="left" width="16%">X1300 PRO</td> <td align="left" width="14%">Overdraw Factor 3</td> <td align="right" width="14%">383.0</td> <td align="right" width="14%">1115.4</td> <td align="right" width="14%">619.9</td> <td align="right" width="14%">191.2%</td> <td align="right" width="14%">61.9%</td> </tr> <tr> <td align="left" width="14%">Overdraw Factor 8</td> <td align="right" width="14%">137.8</td> <td align="right" width="14%">791.8</td> <td align="right" width="14%">369.1</td> <td align="right" width="14%">474.5%</td> <td align="right" width="14%">167.8%</td></tr></tbody></table>
Any idea why 1600 is so much faster? Due to 5 vs 2 VS 's ?

[mczak]

Any idea why 1600 is so much faster?
Uneducated guess, due to the 2 times as many z compare units + much faster ram of the 1600?
That said, what's strange is the really low back-to-front performance of the x1300. It easily gets beaten by the X600XT, and since it has both higher core clock + higher memory clock that's a bit strange imho (I'd assume that the (in this case) useless hierarchical-z buffer shouldn't cause an additional penalty).

[Tahir2]

X1300 could have some items removed from its core.

[chavvdarrr]

X1300 could have some items removed from its core.
Well X1300 obviously has Hier-Z and X600 is supposed to be the less capable of all 4 chips in the comparison. Either drivers for X1300 are not optimised at all or...?

[Dave Baumann]

I mentioned it once already. Int16 blending is not supported on any r5xx chip , you may wanna correct that.
Confirmed and done.

Also, if you still have the X1600XT CF setup, could you look into its multi-monitor capabilities? Ie, can you run 4 screens off it, and then game off one screen with the speed bost of CF without having to reboot or anything?
Although I haven't checked this out explicitly, you should be able to. The 4 screen are deteched and displayed in CCC and to enable Crossfire its tickbox in a CCC panel which shouldn't need a reboot if the boards are well matched. The drawback is that you would have to select/deselect that every time you play a 3D app.

This certainly isn't the only example of what I'm pointing at in your results, but it will do: http://www.beyond3d.com/reviews/ati/rv5xx/chart/image017.gif

Now, is it just me. . .or is there a line on that graph that is doing everything but waving and spelling out in semaphore "The driver team hates me!"? If so, is ATI going to do something about it?
Having thought about it, I'm beginning to suspect that RoOobo may be right about that and it just runnng out of HierZ. 1024x768, or HDTV resolutions, would probably be the optimal target for a board of this ilk and the HeirZ size may be limited to around that region.

Do these cards really have dual dual link dvi? I'm pretty sure both the x1300 and x1600 only have one dual link capable dvi port (and good luck finding _any_ x1300 with more than one dvi port anyway). (ATI's website explicitly mentions one dual-link + one single-link for x1300, it is not accurate enough on x1600 (just mentions "dual-link dvi") and explicitly mentions dual dual-link on x1800.)
X1600's certianly have the hardware for dual dual-link DVI, its up to the vendor whether or not they support it. X1300 is a little fuzzier in that respect, but I think that it still has the capability for it as well.

Second Im just curious if anyones tried running either of these boards in crossfire mode on an Nforce4 SLI mobo. I dont see any hardware limitation that'd cause it to not function. Given the fact that it sounds more like a software solution than hardware I dont see a reason that any mobo with 2 * PCIE Slots shouldnt work with this.
Technically there is no hardware reasons why this shouldn't / couldn't happed, however there are business reasons. You'll find that this is blocked in the official drivers, much like SLI is on Crossfire platforms.

Thanks for the updates guys.