For a long time I have been skeptical about the performance of the 5870. No doubt it is the fastest card, but is it as fast as listed in many benchmarks online? I always thought i may be faster. I used to own an ATI RADEON 4890, the fastest card of its generation. The number raster operators, fragment pipelines & texture units in 5870 is double than the 4890, hence it should be almost double the speed of 4890. But somehow in benchmarks, it was not that fast. For long i have been waiting to buy a DX11 video card & i was waiting for NVIDIA to do some magic. they sure tried, but didnt impress. the performance was amazing, but the temperature & power usage was abysmal. In those benchmarks too, the 5870 appeared dwarfed. I decided to buy a 5970, but could not arrange the funds. So I bought myself a 5870 from XFX, the XXX edition. Now I needed to check if the benchmarks posted online were worth something or not. so spent a good 5-6 hours on my computer & ran some benchmarks myself. I have a few games & the tools to do it.
& My Computer
i7 920 @ 4.2 GHz, 6GB tri channel memory @ 1866 MHz, OS hard drive 2X500GB WD Raid 0, Gigabyte UD7 mobo & 5870 with 900Mhz GPU & 5200MHz memory.
The first issue I wanted to check was the heat of the card. I have seen online that this card went as high as 95c & as low as 33c. I live in India where conditions are extreme. the temp while benchmarking was around 35c-38c already. My 4890 went as high as 95c. I control the fan of the GPU using the MSI AFTERBURNER. Setting the fan speed at "user defined" makes the fan stay at 50% if the GPU temp is below 50c & increases as the temp increases & goes to 80%-85%. The minimum temperature using the 4890 at stock fan speed was 68c & went to 95-96c at auto. Using the us"user defined" mode the minimum temp in 4890 was 58c & highest was 88-90c. Increasing the fan speed at 100% in 4890, the minimum was 56c & max 84-86c. I was expecting the similar result in 5870. But I was surprised. with fan speed auto, the minimum temp was 50c & max was 78c. with "user defined" mode in afterburner, the minimum was 44c & max was 77c. at 100% the minimum was 42c & max 75%. Which is amazing considering the room temps at 35-38c. there was only one fan at the side of the cabinet blowing air to the 5870 though, there were two for the 4890, one at the side & one at the back. I have ordered 3 more 140mm Lian-Li fans to cool the card, but ATI have done amazing job to cool this card. the power usage was also exceptional. now i dont have any particular devices to check power, but I use a APC BACK-UPS 1500VA. This UPS comes with a software which provides total amount of power provided to the devices attached. The computer with 600wat 80 plus tagan u33 PSU was connected to the UPS along with a Dell SP2409WFP monitor. With 4890 the idle power was 316watt & load was 424watt. if we reduce 40watt for Screen & then consider 80% efficiency then the num comes approximately 220w & 300w. with 5870, these numbers were 224w & 386w. After calculation these become 140 watt & 280watt. These numbers may not be extremely correct, but the difference is exceptional, considering the amout of heat is generated & power is consumed by NVIDIA.
I used almost all the games for benchmarking, some games had internal benchmarks & for some which didnt, I used FRAPS benchmarks. With the games using internal benchmarks FRAPS was disabled & was enabled when screenshot was supposed to be taken. & where I have used FRAPS, i have enabled a graph too. the results were considerably more than the results I saw online. Also I have anti-aliasing disabled in games where enabling or disabling it does not make a big difference in visual quality. screen-shots are of the settings used & the benchmark results.
Here are teh results in alphabetical order.
Avatar is a 3rd person action-adventure game. Benchmarks were performed using FRAPS with all setting maximum.
BATMAN: ARKHAM ASYLUM
Batman: Arkham Asylum is a 3rd person action game. In-Game benchmark were used with all setting maximum.
BATTLEFIELD: BAD COMPANY 2
BATTLEFIELD: BAD COMPANY 2 is a first person shooter game.Benchmarks were performed using FRAPS with all setting maximum & DX11.
CRYSIS is a first person shooter game.Benchmarks were performed using FRAPS with all setting maximum.
DEVIL MAY CRY 4
DEVIL MAY CRY 4 beat'em up game. In-Game benchmarks were run with all settings maximum.
DIRT 2 off-road racing game, In-game benchmarks were used with all setting maximum & 8XAA & DX11.
H.A.W.X. is a Flight Action game.In-game benchmarks were used with all setting maximum & DX10.1.
JUST CAUSE 2
JUST CAUSE 2 is a modern action adventure game.In-game benchmarks were used with all setting maximum. There are three benchmarks in the game, all benchmarks were run.
METRO 2033 is afirst person shooter game.Benchmarks were performed using FRAPS with all settings maximum & DX11.
STALKER: CALL OF PRIPYAT
STALKER: CALL OF PRIPYAT is afirst person shooter game. In-game benchmarks were used with all settings maximum & DX11.
RESIDENT EVIL 5
RESIDENT EVIL 5 is a horror action adventure game. In-game benchmarks were used with all settings maximum. There are two benchmarks in this game, both were used.
NEED FOR SPEED: SHIFT
NEED FOR SPEED: SHIFT is a racing game.Benchmarks were performed using FRAPS with all settings maximum. The track used was the most graphic intensive track of the game & with 15 super cars, which means that the minimum frame-rate in this benchmark is the minimum frame-rate you will ever see in any race.
STREET FIGHTER 4
STREET FIGHTER 4 is a 3D fighting game.In-game benchmarks were used with all settings maximum.
CRYSIS WARHEAD is a first person shooter.Benchmarks were performed using FRAPS with all settings maximum.
ALIENS vs PREDATOR
ALIENS vs PREDATOR is a first person shooter game. Benchmarks were performed using FRAPS with all settings maximun & DX11 where tessellation was also enabled.
All these tests took almost 4-5 hours & the GPU temp did not go above 77. I'll post more pictures when I have received the fans.
But speaking about the card, the performance i received was simply more than my former card & was definitely a strong competitor to the GTX480. The temperature & power usage is also exceptional. Though apart from Heaven Benchmarks, no game gave any noticeable visual improvements with tessenation enabled.
No doubt it was the fastest card ...... back in October ..... that is if ya don't count twin GPU cards like the 295 GTX and 4870x2. Today it sits in the 3rd tier behind the 1st tier 5970, and 2nd tier GTX 295, GTX 480 and HD 4870X2.
well i was talking about the single GPU cards in each of the two cases, 5870 comes neck to neck with GTX 480 in many games & beats GTX480 in some. & the second case was specifically for ATI, again single GPU.