Sapphire Toxic X800 PRO VIVO Review

[Dave Baumann]

http://www.beyond3d.com/siteimages/b3dsmall.gif (http://www.beyond3d.com/reviews/sapphire/toxic/)Sapphire made a name for themselves providing the total silent “Ultimate Editions” for the Radeon 9x00 series, however with the X800 series Sapphire are taking a slightly different route. Here we take a look at Sapphire’s new alternative high end board, the Toxic X800 PRO VIVO.<blockquote>"Beyond the cooling system, and the obvious PCB colour change, the rest of the board appears to adhere to the ATI reference design. Removing the heatsink from the board reveals Samsung 1.6ns GDDR3 RAM, which should be capable of a maximum operating speed of 600MHz, which well beyond the rates that even the "APE" overclocking runs it at."</blockquote>Read the full article here (http://www.beyond3d.com/reviews/sapphire/toxic/).

[Tokelil]

What would the reason that the memory "only" clocks as high as 545? 520 also seems very conservative at first glance for memory that is supposed to run at 600 MHz. (Without cooling?) Chip design issue? (Shouldn't be an issue since the XT PE is running at 560 MHz) Board design maybe? Any clues? :P

Btw. good review as always! The small gain when not using APE was also kind of interresting IMO...

[Richard]

Great review (as always duh - running out of compliments). Anyway, I especially liked the OC parts. I do have one suggestion though. Could you make it so that each card always uses the same colour in the charts? I find myself always checking the legend because some show the Saphire Toxic as bordeux others as yellow, etc.

[Seiko]

Another robust article although I'm not sure that the ATI reference comparison is really necessary? I'm finding sub 10 FPS differences of little use already let alone 0.1% from reference. Instead of graph after graph showing the tiniest deviances I’m hoping B3D could instead utilise the time with the card a little differently and move into slightly different areas.

One aspect I'm really hoping for from B3D in relation to this article is the X800 VIVO softmod. I appreciate this takes a slightly more hardware related “modding” direction but with such a buzz at the moment it would be a great article? I know a lot of people including myself swear blindly that unlocking the pipelines works and yields results but then some mentioned that although unlocking the pipes with a bios flash shows via tools like “atitool” that the pipes are unlocked certain tests show that they are not utilised without the XT bios. Although I'm using the XT bios anyway and therefore should be fine I'd love to read an in depth article explaining the myths and legends that gather on forums. I note the article currently skirts over this potential free gain?

Also, I think that a clear CPU bottleneck report should be derived against any of the high end cards as many people will be shocked at just how much the CPU becomes the bottleneck. I’m a dual TFT user and with a resolution of 10x7 the likes of UT2004 become comparatively CPU limited even with 6xFSAA and 16xAF when using my 2.8 P4. Again, I appreciate B3D is not a gaming site nor hardware “modding” site but I do think with such a solid and trusted set of reporters it’s a shame not to push out these findings.

With regard to the overclocking results etc I’d also like to see temperature reports too, i.e. heat rise from standard and FPS gain. These figures from a trusted source like Dave and the guys bring so much more and ensure I and others have a solid foundation to build from.

Just my two pennies so please don’t feel the need to take defensive positions.

:D

[AlphaWolf]

One aspect I'm really hoping for from B3D in relation to this article is the X800 VIVO softmod.

He already explained that this would not happen in a previous thread. Bios flashing was apparently not an option.

Also, I think that a clear CPU bottleneck report should be derived against any of the high end cards as many people will be shocked at just how much the CPU becomes the bottleneck. I’m a dual TFT user and with a resolution of 10x7 the likes of UT2004 become comparatively CPU limited even with 6xFSAA and 16xAF when using my 2.8 P4. Again, I appreciate B3D is not a gaming site nor hardware “modding” site but I do think with such a solid and trusted set of reporters it’s a shame not to push out these findings.

Look at the graphs.

[Dave Baumann]

What would the reason that the memory "only" clocks as high as 545? 520 also seems very conservative at first glance for memory that is supposed to run at 600 MHz. (Without cooling?) Chip design issue? (Shouldn't be an issue since the XT PE is running at 560 MHz) Board design maybe? Any clues? :P

Like anything else, memory speeds can be limited by many factors - even if a core is able to scale up to high speeds, that doesn't necessarily mean its memory will scale as well, even if the design allows for much greater speeds in some cases. Likewise, the board itself may limit speeds and there may be board level yields at different performances. We have seen some R420 cores scale up much higher.

Could you make it so that each card always uses the same colour in the charts? I find myself always checking the legend because some show the Saphire Toxic as bordeux others as yellow, etc.

Oops - thats the Excel defaults coming in. Because there is a different mix of useage in the OC and FSAA/AF sections to the main game and performance testing areas it picks up different default colours.

One aspect I'm really hoping for from B3D in relation to this article is the X800 VIVO softmod. I appreciate this takes a slightly more hardware related “modding” direction but with such a buzz at the moment it would be a great article?

As Alphawolf mentioned, I did discuss this in a thread beforehand, and a the moment it doesn't appear to be just a software trick, but a BIOS update - I wasn't keen on replicating something that may nor be able to return the board in its initial state, from both a reviewing perspective and an end user perspective. We've debated for a long time whether or nor to even include overclocking because the results are so variable board by board, and that's even more so with things like unlocking pipelines - thats why I put the warning in the conclusion; I don't want to give the expectation that people will actually get the extra pipelines.

That said, I may go back and look at this in the future.

Also, I think that a clear CPU bottleneck report should be derived against any of the high end cards as many people will be shocked at just how much the CPU becomes the bottleneck.

Part of the reason for using Fill-rate graphs, as we do, is to highlight CPU/system limitations.

With regard to the overclocking results etc I’d also like to see temperature reports too, i.e. heat rise from standard and FPS gain.

Might be an idea, but I don't have any external thermal probe stuff, so that would ave to come from the control panel and not all hardware uses temp probes on chip/board. We don't necessarily run in a normal way either, since its an open rig.

[Anonymous]

What would the reason that the memory "only" clocks as high as 545? 520 also seems very conservative at first glance for memory that is supposed to run at 600 MHz. (Without cooling?) Chip design issue? (Shouldn't be an issue since the XT PE is running at 560 MHz) Board design maybe? Any clues? :P

Like anything else, memory speeds can be limited by many factors - even if a core is able to scale up to high speeds, that doesn't necessarily mean its memory will scale as well, even if the design allows for much greater speeds in some cases. Likewise, the board itself may limit speeds and there may be board level yields at different performances. We have seen some R420 cores scale up much higher.

Could you make it so that each card always uses the same colour in the charts? I find myself always checking the legend because some show the Saphire Toxic as bordeux others as yellow, etc.

Oops - thats the Excel defaults coming in. Because there is a different mix of useage in the OC and FSAA/AF sections to the main game and performance testing areas it picks up different default colours.

One aspect I'm really hoping for from B3D in relation to this article is the X800 VIVO softmod. I appreciate this takes a slightly more hardware related “modding” direction but with such a buzz at the moment it would be a great article?

As Alphawolf mentioned, I did discuss this in a thread beforehand, and a the moment it doesn't appear to be just a software trick, but a BIOS update - I wasn't keen on replicating something that may nor be able to return the board in its initial state, from both a reviewing perspective and an end user perspective. We've debated for a long time whether or nor to even include overclocking because the results are so variable board by board, and that's even more so with things like unlocking pipelines - thats why I put the warning in the conclusion; I don't want to give the expectation that people will actually get the extra pipelines.

That said, I may go back and look at this in the future.

Also, I think that a clear CPU bottleneck report should be derived against any of the high end cards as many people will be shocked at just how much the CPU becomes the bottleneck.

Part of the reason for using Fill-rate graphs, as we do, is to highlight CPU/system limitations.

With regard to the overclocking results etc I’d also like to see temperature reports too, i.e. heat rise from standard and FPS gain.

Might be an idea, but I don't have any external thermal probe stuff, so that would ave to come from the control panel and not all hardware uses temp probes on chip/board. We don't necessarily run in a normal way either, since its an open rig.

Thanks for the cool responses. In particular I'm interested in understanding a little more from the fillrate graphs as I had assumed and no doubt incorrectly that in order to see how well the card scales a number of CPUs or replicated CPU speeds would need to be included. The testing of so many cards these days utilises the highest CPU available in order to determine how well the GPU performs and although essential I believe this results in the majority of every day users not seeing how the card will perform in a realistic environment. As I said I was literally shocked to find out that the cards are bottlenecked in most situations by a P4 2.8 at 10x7. with max eye candy :shock:

Finally, "That said, I may go back and look at this in the future." This would be a great treat Dave that I'm sure many an X800 would love to read.

[Bambers]

The same dip at 1024x768 in the fillrate graph for TR:AOD also shows up in CS:S, something to do with pixel shading then?




As Alphawolf mentioned, I did discuss this in a thread beforehand, and a the moment it doesn't appear to be just a software trick, but a BIOS update - I wasn't keen on replicating something that may nor be able to return the board in its initial state, from both a reviewing perspective and an end user perspective. We've debated for a long time whether or nor to even include overclocking because the results are so variable board by board, and that's even more so with things like unlocking pipelines - thats why I put the warning in the conclusion; I don't want to give the expectation that people will actually get the extra pipelines.

agreed. Although I've not heard of a VIVO failing the mod yet (including mine :twisted: after the unforunate death of my 9800pro) quite a few don't make 520/560 stable (like mine, 513/574 tops) which means either a further bios flash back to a pro or using atitool to 'underclock' it on startup and it's a rather more blatant voiding of a warrenty than a simple overclock or using a modded driver as well.

[Broken Hope]

If Dave is able to get hold of a Gigabyte X800 Pro ViVo then he will be able to compare the 12 pipes to a 16 pipe card since the Gigabyte card comes with all 16 pipes enabled as standard.

[Dave B(TotalVR)]

Dude, take another look at your Doom 3 bench's. You've got some extra -'s in there :roll: