The Elsa Gladiac Ultra
It sure is fast...but you'll pay for it!

 
By Marco "BigWop" Chiappetta - November 14, 2000
 
H.H. Test System
Get the napkins ready...

InWin Full Tower ATX Case w/ 300W PS, Pentium III 933EB, Asus CUSL2 (i815) Motherboard, Elsa Gladiac Ultra 64MB AGP Card, 256MB of PC133 True CAS2 SDRAM from Mushkin, IBM 30.7Gig 7200 RPM ATA/100 Hard Drive, Plextor UltraPlex 40max CDROM, WinME, DirectX 7.1, nVidia reference drivers (Detonator 3 6.34)

A hot topic when talking video cards is FSAA (Full Scene Anti-Aliasing).  Looking at the options available in the drivers on the previous page, you'll see that FSAA is available in both Direct3D and OpenGL.  Here is an example of the quality you can expect when enabling FSAA on the Gladiac Ultra...


Modem users beware...the full size pic is over 200K.

This shot was taken in Star Trek: Elite Forces at 1024x768 resolution at a 32Bit color depth.  We used the highest FSAA setting (2X2) available.  This is not the most colorful screenshot but it is full of lines that would normally be marred by "jaggies".  Notice how smooth the lines become with FSAA enabled.  Look especially close at the red laser light coming from the head of the BORG...there is not a single pixel missing.  Performance isn't so hot at higher resolutions with this level of FSAA enabled.  Here are some Quake 3 numbers using the "High Quality" setting with Geometric and Texture detail set to maximum.

Quake 3 Arena (OpenGL)

We'll go more into our overclocking experience with the Gladiac Ultra a little further down the page but for now he's a taste of what some tweaking will do with FSAA enabled.

 

Benchmarks With The Elsa Gladiac Ultra
Drop the Checkered Flag!

 
We put the Elsa Gladiac Ultra through our usual barrage of testing.  We ran MadOnion's 3D Mark 2000 v1.1 to test the Direct 3D performance of the card and ran MadOnion's Video 2000 to test it's prowess at video playback performance and quality.  To test the OpenGL performance of the Gladiac Ultra we used MDK2's built in timedemo and also the defacto standard OpenGL benchmark, Quake 3 Arena.

MadOnion Video 2000

We didn't use our normal Video 2000 bar graph in this review because it does not clearly show the total video marks.  Surprisingly, the Gladiac Ultra did not perform as well as it's non-Ultra counterpart.  This could be due to the differences in the test machine but nonetheless something worth mentioning.  Next we ran 3D Mark 2000...

MadOnion 3DMark 2000 v1.1 (Direct 3D)


(V-Sync and FSAA Disabled)

These are some excellent 3D Mark scores. The fastest we have seen to date from a board available at retail.  If you want to see how how an Ultra will really go...check out this page.  What site is at the top of the list?  Oh Yeah, that's us!

We ran MDK2's built in timedemo at both 16 and 32-Bit color depths with hardware T&L enabled.  MDK2 tests a video card's OpenGL performance.

MDK2 Timedemo (OpenGL)

We were CPU limited at virtually all resolutions at a 16-Bit color depth.  Let's see how we fare at 32-Bit color...

Again, these are the highest numbers we've seen from a "retail" board.  Being the lunatics that we are we overclocked our Gladiac Ultra and ran this same benchmark.

We had fairly good luck overclocking our board.  We hit a maximum core clockspeed of 300MHz and an effective clockspeed of 500MHz with the RAM.  We suspect that we could have clocked the RAM higher because we experience no visual artifacts or "snow-flaking" at all.  However, at anything higher than 500MHz, our display would distort and the test system would lockup.  Other hardware review sites have seen the same thing happen, so we suspect this could be a deliberate limitation.  Because we were CPU limited the scores were virtually unchanged at the overclocked speed.  We did have one anomaly at 1280x1024 though.  No matter how many times were ran the timedemo, upon completion it would not display the dialog box with the score.

More Over-clocking, Quake 3 and The Rating