And the PC itself:
(That's the radiator [left] and the pump)
Test setup
I will be running an overclocked Athlon 64 3500+ with 1 GB memory. The specs are as follows:
AMD Athlon 64 3500+ (CPU clock 2400 MHz)
ASUS A8R32-MVP Deluxe (Chipset clock 240 MHz, HTT 3x)
1 GB OCz Platinum Dual-channel (RAM clock 240 MHz, 2.5 CAS)
1x Western Digital 160 GB HDD (PATA) for system
Onboard sound, LAN, etc.
Windows XP Professional x64 edition
And obviously, the graphics card is an ATi Radeon x1950 XTX PCIe.
The card out of the box
The first thing that caught my eye was the new cooler. It looked sweet in pictures, but it’s even cooler in real life. I also noticed that the card is quite heavy, much more so than my old x850, and longer too, but this is to be expected.
The card came with almost everything I needed for it to run (cables, drivers, etc). except it does not come with a Molex to PCIe power converter! Make sure you purchase one of these separate; though most sites tell you this part is included, it wasn’t in my BBATI box.
Upon installing the card, it was just the perfect length, not hitting my hard drives (which are installed backwards, cables facing the front and hidden) and leaving plenty of room for airflow. Turning the computer on, the cards stock fan speed is extremely quiet, but it offers little in the way of cooling at this speed (13%): upon entering windows and laoding ATiTool, the temperature with 2D clocks was over 50 degrees Celsius! Turning the fan speed up to 100%, it is a noisy beast, but still quieter than an x1900 XTX at 100%. However, the card is louder at 50% than the x1900 XTX at 50%. This just seems to be the trade-off.
This card, like newer ATi cards and nVidia cards of the last generation, changes clocks and voltages based on running applications: on the desktop when little graphical power is needed, the card switches to a much lower clock speed, fan speed, and core voltage, and when a 3D intensive program is loaded it immediately switches up to the higher clock speed, fan speed and voltage. Which this is good for the average user, it is a PAIN for any serious overclocker. Luckily, it can be disabled in one of two ways: If you’re using an old (pre-October 2006) version of ATiTool, you can disable the ATI Hotkey Poller service to prevent the clocks from switching. REMEMBER YOUR 3D CLOCKS AND VOLTAGES HOWEVER. Then, using ATiTool, create a profile with the 3D clocks and voltages and use that to play games. If you’re using the latest beta release of ATiTool (0.25 Beta 16), there are additional “x1000 Overclocking” features. One of these automatically loads 3D mode then turns off the service when ATiTool starts and prevents it from restarting until the computer restarts. This itself is very useful, (a) because it loads the 3D profile, not the 2D one, and (b) it saves you having to mess with services if you don’t want to. The downside is turning them off manually bugs ATiTool and it may not work properly.
Overclocking
My blog IS called “Diary of an Overclocker”, so not writing about the cards overclocking abilities would complete defeat its purpose!
Stock out of the box with no voltages changes, my card overclocked to about 680/1050. Pretty respectable, but it seems a bit limited. Pumping up the voltages to very high levels however (maxed out the vCore voltage and set vRAM at ¾ slider), I was able to get 735/1125 before getting major hardware artifacting. I consider these a damn good overclock, and the temperatures never rose above 60-65 degrees Celsius at 80% fan.
For my reviews, I will be using an overclocked card exclusively, with slightly lower speeds than my max:
GPU clock: 715 MHz
GDDR clock: 1098 MHz
VGPU: 1.525 volts
MVDDC: 2.285 volts
MVDDQ: 2.272 volts
VDDCI: 1.430 volts
Fan speed: <> 70 degrees, 100%
Test 0: Synthetic benchmarks
I’m not a big fan of synthetic benchmarks for reviews, mainly because their nature makes them CPU/RAM limited. However, for the sake of conformity, I’ve included two here: 3DMark 05 and 3DMark 06.
3DMark 05
Pretty standard stuff; a nice score, though it is a bit lower than anticipated.
Note: No Batch Size Tests were run.
3Dmark 06
MUCH lower than anticipated, but the low CPU score is most likely to blame.
Test 1: The Elder Scrolls IV – Oblivion
For this test, I will be running the game at 1280x1024, “Very High” quality, Tree and Actor fades maxed, Item and Object fades at ½, Grass off, exterior and interior shadows at ⅛, HDR on, Specular Dist at ½, no AA, all other settings to default. Relevant mods include the Natural Environments mod (HDR choices where available), all official Bethesda mods, and the Unofficial Oblivion Patch (latest version).
My test took me from the Chestnut Handy Stables outside the Imperial City straight through to the road branch on the other side of the lake. This crossing of the lake allowed me to witness water effects. I looked left and right throughout the test (to get the distant trees and mountains), and fired both arrows and spells to capture the effects. Finally, 2 humans and a horse at the end of the path, as well as many trees, rounded out the test. 60 seconds.
Total Frames Time (ms) Min. Max. Average
2774 60000 27 62 46.233
Test 2: Need For Speed – Most Wanted
Though the game is fairly old (2005 release), its graphics are still high-quality and it is still a good benchmark.
Resolution 1280x1024, all settings (including AA) maxed.
My test took me from the safehouse in Central Rosewood around the immediate area. NOS and speedbreaker were used for the effects. Several intentional collisions occurred as well to test particle emission. 60 seconds.
Total Frames Time (ms) Min. Max. Average
2550 60000 31 62 42.500
Test 3: Battlefield 2
Testing at 1280x1024, all settings maxed, 4x AA. Random gameplay around Daqing Oilfields with 16 CPU players. 60 seconds.
Note: The drop to low/mid-20s FPS was the result of textures and AI loading and was a system, not graphics, issue.
Total Frames Time (ms) Min. Max. Average
3317 60000 21 78 55.283
Conclusion
All in all, I feel this was an excellent purchase. Though it was a little pricey (a month ago, $550 CAD), this card performs beautifully, running all the above games (really the only ones I play) at my monitors max resolution without any problems.
A good purchase, even with a CPU/RAM-limited system.
Additional
