X1950 R580+ reviews

probably preorder remember hearing overclockers uk taking preorders.

 

So how much can it gain with driver opts?
Right now it's looking like it has more bandwidth then it knows what to do with.
10% gain on average with a 30% increase in bandwidth.
Ati should have tried to clock the R580 up in the 700 mhz range...

 

I'm thinking another 5 to 10 % if the timings are different on the ram the drivers might have to be tweaked for ring bus to fully utilize it.

 

I hope so, it looks kinda weak for having all that bandwidth at it's command.

 

Well, they need the extra bandwidth and GDDR4 is a good selling point and it doesnt come any slower than this (I believe ATi is actually even running it slower than its rated for) as far as GDDR4 goes.

I would have liked to see a nice bump on the core, which probably would have netted pretty nice performance improvements, but if it meant a louder cooling system then I'm okay with not doing it.

 

There is some sort of problem with the new Intel chipsets that Core Dou runs on and GX2. Probably not enabling SLI at all.

Any review that uses that chipset with GX2 is out the window.

I know Anand and FS used Nvidia chipset with GX2. Both reviews show as youd expect, GX2 ahead of X1950XT by a large margin in most cases.

To claim otherwise would be to basically claim X1950 is faster than two 7900GTX's in SLI (albeit slightly downclocked), which clearly doesn't make sense.

I started to look through all the reviews for evidence of the GX2/Core Dou chipset anomoly issue, but it seemed like a big pain, as many of the review are organized so poorly it's difficult to find info fast (for example, I believe the Guru 3D article first up on the B3D list doesn't even state what chipset were used on the test platforms at all, nice).

So how about this, if anybody (Skyring?)claims X1950 is as fast as the GX2, then you show me that review/reviews you have in mind that proves it, and I'll show you an example of the chipset bug.

Here is one Anand bench:

http://images.anandtech.com/graphs/ati radeon x1950 xtx launch_08230650831/12908.png

This is the sort of spread you should expect from a 79GX2 v x1950 comparison.
________
Vaporizer

 

You mean two underclocked 7900gt's, hence the low profile, low-noise cooling solution. Kinda makes it more likely then to see an x1950 besting them in situations.

 

What situations?

The GT is stock clocked to 450/1320, the GX2 at 500/1200, btw.
________
Naugtymamaxxx cam

 

Or a better one like this:

http://images.anandtech.com/graphs/ati radeon x1950 xtx launch_08230650831/12909.png

or

http://images.anandtech.com/graphs/ati radeon x1950 xtx launch_08230650831/12918.png

or

http://images.anandtech.com/graphs/ati radeon x1950 xtx launch_08230650831/12911.png

 

The xbitlabs review was a lot more interesting to me than some of the initial ones.

I think I rather read a more comprehensive review days later than something too short.

 

Thats just not the case. The GX2 works fine on 975 boards provided the bios for the board supports the GX2 and the driver you choose to use works properly with a single GX2 board.

In my review i used the Badaxe board with a GX2 supporting bios (cant get the exact revison just now because the board isnt installed atm) and 91.31. With this board/bios/driver config the dual GPU mode works perfectly providing you use a game which is profiled in the driver.

Infact, it works so well that with a single GX2 and the X6800 i reached 20th place in the PCmark05 ORB yesterday afternoon.

EDIT: What you could say is that if a board doesnt appear on this list: http://www.nvidia.com/content/geforce_gx2_sbios/us.asp and is used in a review which uses the GX2 then there will more often than not be weird or incorrect results.

 

Aaaargh!! Xbit's pissing-in-the-wind conclusions (after each graph) irk me so damn much !!!

 

Actually that game if AF is on would stress the bandwidth a great deal, and technically, the gx2 has less bandwidth then the 7900 gtx, (x2 the bandwidth really doesn't exsist with dual cards as most reviewers have stated), and in games that stress bandwidth the x1950xtx should be closer to the gx2 if not even marginally faster.

 

Well its possible to tune them out.

 

X-Bit labs has posted their review.

 

The issue goes beyond that. I think Nvidias last few batches of drivers have been plagued with bugs. There are lots of guys reporting 50% drops in performance across the board on GX2 systems with all different setups--not just Core Duo. A lot of it has to do with settings in the new control panel not adjusting properly. Ive witnessed this many times in the last month on the Nvidia forums.

Its really an absurd issue and something Nvidia had better fix quick. The reputation of their cards is going downhill quick between this and the 7900 series issues. I may never get another Nvidia card because of this mess. ATi has never given me this kind of trouble.

 

You have to understand that a brand new control panel wouldn't be much of an issue, had it not been made on top of Microsoft's .NET framework.
Just like ATI's CCC, the new NV CP is really meant for Windows Vista -which uses .NET 2.0 extensively-, although it continues to be backwards compatible with WinXP.

So, take them both as test versions of what's to come in Windows Vista, and use the old CP for greater reliability on XP.

Anybody knows that the speed of ACCC and NNVCP of both IHV's are no good in MSWXP.

 

Aiieeee! My eyes! Can we neg rep for egregious use of acronyms?

 

I dont have any problem with the new NV CP, it's quite fast (surprisingly given ati's is slow), but I do think it isn't designed very well so I use the classic CP.

 

I like that review... has more games where the 1950 has a nice lead over the 1900.