[staryoshi]

The GTX460 has taken the graphics world by storm. Its combination of performance, efficiency, and value has sent it flying off store shelves, virtual or otherwise, since its July 12 release. I have taken the time to benchmark the MSI Cyclone GTX460 1GB, stock and overclocked, and compare it to both the HD5850 (Stock and overclocked) and more importantly, in SLI (Also stock and overclocked). I sincerely hope that you find my benchmarks and analysis useful, as always feel free to ask questions or request further benchmarks.

Thanks for reading!

Testbed Configuration:
Intel Core i7 860 @ 3.36Ghz - EIST, Hyper-Threading, Turbo Enabled
4GB G.Skill ECO DDR3-1333 CL7 @ DDR3-1600 8-8-8-24
Asus Maximus III Gene
Samsung Spinpoint F3 1TB
Windows 7 Home Premium 64
Drivers
ATI Catalyst 10.7
nVidia Forceware 258.96

Test System:

MSI Cyclone GTX460 1GB:

GPU Specifications:

Overclocking:
The GTX460 is a capable overclocker, much like the HD5850. The reference model comes in at core, shader, and memory clocks of 675Mhz/1350Mhz/900Mhz(3600Mhz Effective) respectively. MSI's Cyclone model bumps the core clock speed up a respectable 50Mhz to 725Mhz. For overclocked testing, I went with clocks of 900/1800/1050(4200Mhz Effective). This amounts to a 24.1% and 16.7% increase in core and memory clocks respectively. However, in SLI I was forced to pull the clocks back to 880/1760/1050 due to my motherboard configuration and the resulting thermal limitation. For the HD5850, I overclocked it from its reference 725Mhz/1000Mhz to 900Mhz/1250mhz. The memory clock is faster than that of the overclocked GTX460, but both will be operating at the same core speed. I do not have power consumption figures, but consider the additional power requirements of overclocked GPUs when pushing them.

Synthetic Benchmarks

-3dMark Vantage
--Much more relevant than 3dmark06, Vantage is a great gauge for relative GPU strength. Testing was conducted at performance and extreme presets.

-Stone Giant
--A DirectX 11 benchmark featuring varying grades of tessellation. Tested at 1680x1050 and 1920x1200 resolutions with tessellation set to high.

-Unigine Heaven
--The most prevalent tessellation benchmark, like Stone Giant it's a good measure of a graphics card's tessellating capabilities.

Game Benchmarks

-Batman Arkham Asylum
--I ran this game for comparison amongst the GTX460 only do to its additional Physx workload.

-Dirt 2
--Run at 1680x1050 and 1920x1200 resolutions using the High preset and 4xAA. Dirt 2 is a good measure of DX11 GPU performance, utilizing new technologies to deliver a graphically-breathtaking experience.

-Far Cry 2
--A bit outdated, but still a good measure of GPU performance. Tends to favor nVidia cards.

-Final Fantasy XIV
--I'm still up in the air on this one. The benchmarks were run at high (1080p) and low (720p) setting as a Hyur male. It's no-fuss but takes a little while to run. Not sure how mature the engine is, but the results are what they are

-Just Cause 2
--One of the most graphically demanding games out there featuring one of the largest virtual landscapes to date with incredible detail, Just Cause 2 is my favorite sandbox game in years. I disabled the nVidia-specific features for the purpose of benchmarking. Testing was run at 1680x1050 and 1920x1200 resolutions at the specified settings.

-Resident Evil 5
--Not a game that I care for, but the benchmarking tool is useful. Testing was run at 1680x1050 and 1920x1200 resolution at max settings using the fixed benchmark.

-Stalker: Call of Pripyat
--Testing was run at 1680x1050 and 1920x1200 resolution at the pictured settings. Only the SunShafts benchmark scores are included as it is the most demanding of the benchmarks.

-Street Fighter IV
--While not the most resource demanding of games, SFIV scales very well and is useful for measuring relative performance. Testing was run at 1920x1200 resolution at max setting using the "Ink" effect.

 

[staryoshi]

Benchmark Results
-Here they are! Enjoy!

Synthetic Benchmarks
-3dMark Vantage, Stone Giant, Unigine Heaven

3dMark Vantage

At stock setting, the GTX460 trails the HD5850 by 11% at the performance setting and the gap narrows to 4.5% at extreme. A very respectable showing. The GTX460 scales well when overclocked and when run in SLI. The 24.1% core (and 16.7% memory) clock boost nets it 22% and 23.6% gains in Performance and Extreme presets respectively, a very nice, linear improvement. When thrust into SLI the performance gain is equally impressive, netting 81.4% and 91.3% improvements respectively. The HD5850 is no slouch here either, netting similar improvements given the same core clock increase (however the memory clocks were increased by 25% so it has a bit more oomph). The CPU score at my modest overclock was around 21-22000, so we start to see a CPU-bottleneck when running SLI at Performance settings. this is understandable since the resolution is set to 1280x1024. The story changes once the test is ran at Extreme setting, shifting back to a strong GPU bottleneck.

Stone Giant

This benchmark allows the GTX460 to flex its tessellation muscles, revealing why the HD6000 refresh is necessary. The GTX460 is faster than the HD5850 by nearly 40% at stock as well as overclocked settings. This test also displays a result I have yet to see: greater than 100% scaling. When run in SLI, the performance improvements at 1680x1050 and 1920x1200 are 113.5% and 109.1% respectively. GTX460 SLI scaling is a force to be reckoned with.

Unigine Heaven

In Heaven we see the HD5850 fall behind the GTX460 by nearly 20% at 1680x1050 and 16% at 1920x1200. Again, overclocking yields consistent results. Both cards experience about a 22% performance boost when overclocked. SLI scaling yields promising results again, although not as impressive as that seen in Stone Giant. At 1680x1050 and 1920x1200 the performance increase is 79.5% and 84.6% respectively. Both cards perform well given the intensity of the benchmark, but the GTX460 shines again. It is foreshadowing a consistent theme: in applications utilizing DirectX 11 and tessellation, the GTX460 will generally conquer the HD5850. In other applications the HD5850 comes out on top (excluding games that favor one brand over the other).

Game Benchmarks
-Batman Arkham Asylum, Dirt 2, Far Cry 2, Final Fantasy XIV, Just Cause 2, Resident Evil 5, Stalker: Call of Pripyat, Street Fighter IV

Batman Arkham Asylum

I only ran Batman: AA on the GTX460 configurations so I could test at the same settings with all the bells and whistles. Performance improvements are seen but nothing amazing. We see a 24.2%/20% jump when overclocked and a 39.4%/45.5% increase with SLI at 1680x1050 and 1920x1200 respectively. When running a SLI configuration without a dedicated Physx card, it is imperative to run Physx processing on GPU2. With Batman, I saw about a 6 FPS improvement in average frame rates and a more significant improvement in minimum frame rates by running on GPU2 instead of Auto (GPU1).

Dirt 2

The GTX460 bests the HD5850 by nearly 14% at 1680x1050 and over 7% at 1920x1200. SLI offers a 75.4%/82.3% respective boost over stock performance. Both cards net about 20% gains when overclocked.

Far Cry 2

The GTX460 bests the HD5850 by 20.3% and 18.4% at 1680x1050 and 1920x1200 resolutions respectively. Overclocking boosts GTX460 performance by 21% at both resolutions and helps the HD5850 by 17.6%/18.6% respectively. SLI scaling is not as effective as other applications, but it is still satisfactory. The gain is 62.4% at 1680x1050 and we see improvement to 76.3% at 1920x1200. Games that utilize DirectX 11 or favor nVidia GPUs also favor the GTX460 in this comparison.

Final Fantasy XIV

The Final Fantasy XIV benchmark goes against the trend seen so far and strongly favors the HD5850. At stock clocks, the HD5850 beats the GTX460 by 32.4% at 720p and a whopping 49.2% at 1080p. SLI scaling is minimal, only exhibiting performance increases of 16.4%/26.8%. I'm not sure if this is due to the immaturity of the engine or the technologies they utilize. We will have to see if these performance figures hold true when the final product is released.

Just Cause 2

Just Cause 2 is a very demanding game, but unlike Dirt 2 it does not favor the GTX460. The HD5850 beats the GTX460 by 14.3%/18.2% at stock settings. The trend so far is that the HD5850 widens or lessens (if it's faster or slower than the GTX460 respectively) the performance gap at 1920x1200, showing that it has a bit more muscle from a technical standpoint. SLI scales VERY well, with over a 90% improvement seen at both resolutions.

Resident Evil 5

The HD5850 edges the GTX460 by 3% at 1680x1050 and pulls away a bit at 1920x1200 by offering a 9.2% higher frame rate. The trend previously evidenced holds. SLI scaling is respectable, achieving 85.9% at 1680x1050 and 79.6% at 1920x1200. When overclocking the SLI configuration, it seems like a wall has been reached as the performance increase is less than 3%.

Stalker: Call of Pripyat

Only the SunShafts benchmark scores were included, as it is the most demanding of the four individual benchmarks. The two cards trade blows at stock and overclocked settings, with the GTX460 being slightly faster. Both cards see a gain of over 20% when overclocked. SLI scaling is solid again, improving by 85.9%/79.6% at 1920x1200 and 1680x1050 respectively.

Street Fighter IV

The HD5850 edges out the GTX460 at stock settings and scales nearly identically when overclocked. SLI scaling amounts to 69.7%, pulling nearly 200FPS. The actual frame rates seen here are mostly irrelevant, as they are all over 100FPS, but the relative performance figures are useful.

Those are all of my game benchmarks, see my next post for my conclusion. Feel free to recommend additional testing though!

 

[staryoshi]

Conclusion

The GTX460 1GB presents an amazing value. For $230, you receive a graphics card capable of trading blows with the HD5850, often besting it depending on the game or benchmark. Beyond its performance, the GTX460 offers solid thermal performance and utilizes a PCB shorter than most motherboards, a nice trait to see considering the GTX260 was several inches longer.

Featuring a 160w TDP, the GTX460 is much more competitive in terms of power consumption than other Fermi models. It still exceeds the HD5850's 151w, but it is great to see a less hungry Fermi. This particular GTX460 model (MSI Cyclone) has very respectable thermal performance. Given an ambient of 23.5C its idle temperature is 32C and its Furmark load with auto fan speed is 54C after 10 minutes. It manages this temperature at hardly audible sound levels as well. However, if you plan on running these cards in SLI I would recommend at least one additional slot of space between the two cards (aka PCIE / slot / slot / PCIE) to give them ample room to breathe.

Consumers now have an ample stable of choices in the $180-240 price range thanks to the introduction of the GTX460 (768MB and 1GB). If buying new, I would definitely recommend the 1GB GTX460 before any other card near its price point. However, if the ancillary benefits of the card do not impress you and you're concerned for frame rates alone, the GTX465 has now become an excellent value due to aggressive price slashing. The GTX470 can also be had for much less on average. One may also consider a second-hand HD5850 as well, which have fallen down into the low $200 range.

You can't go wrong with the GTX460! It's Yoshi-Approved!

I hope you enjoyed the benchmarks!

Here are a few bonus shots! (The PCB is from the EVGA GTX460 1GB SC EE)

 

[SkillzKillz]

Will you be comparing 1GB SLI 460's vs CF 5850's?

 

[saint19]

Nice, u also can find in other site a benchmark of the 460 SLI against a 5870 and 480 when the 460 SLI is the winner in all the test.

 

[staryoshi]

Quote:

Originally Posted by SkillzKillz Will you be comparing 1GB SLI 460's vs CF 5850's?

No can do, I no longer have two HD5850s. The HD5850s would outperform the GTX460 in applications where the HD5850 already defeats it generally, but I would think that the GTX460s will scale better, especially as drivers mature.

BENCHMARKS ARE FINALLY FINISHED! Enjoy!

Please feast upon the fruits of my labor

Here's a picture of my temporary config: Waiting for my black zip-ties to get here so I can make it look good.

 

[Derp]

Great job on the review, I can tell you spent a good amount of time on it.

 

[Kny]

Very nice. I was hoping to see exactly what I can hope to expect out of SLI'd 460s as opposed to running a single one. It's a shame I'll have to wait a while to throw my second card in, seeing as how the SLI hack isn't up to date with the 258.XX drivers. I don't know how long I'll be able to keep my second one boxed up ; ;

 

[Tator Tot]

Great work as always Yoshi

 

[Graimito90]

Did you have two 5850 in Crossfire before?
If so, Would you say the change was good? I have two 5850 Crossfired and I'm getting tired of the failures of ATI's drivers in multiGPU. Great work.

Thanks

 

[metalfox7]

Very informative! Good work.

 

[ZealotKi11er]

Thats a low OC for the HD 5850. Most cards can do 950Mhz+

 

[HeliXpc]

Quote:

Originally Posted by ZealotKi11er Thats a low OC for the HD 5850. Most cards can do 950Mhz+

Ive had several 5850's, most cannot do 950. 850 most likely.

 

[staryoshi]

Quote:

Originally Posted by Graimito90 Did you have two 5850 in Crossfire before? If so, Would you say the change was good? I have two 5850 Crossfired and I'm getting tired of the failures of ATI's drivers in multiGPU. Great work. Thanks

I used to have crossfire 5850, yes. The results should reflect the benchmarks I've listed in the single card comparison. I would imagine that they would trade blows depending on the app. I've had better experiences with SLI overall than crossfire. (I have used SLI GTX 285, GTX 460 and crossfire 5750, 5770, 5850, and 5850+5870). I would expect better scaling overall with the GTX460 and fewer issues. I used to get "rippling" in games that would only be resolved by using v-sync with the xfire 5xxx series. I have yet to run into that issue with SLI.

Quote:

Originally Posted by ZealotKi11er Thats a low OC for the HD 5850. Most cards can do 950Mhz+

I was limited by cooling. I also wanted to keep the core clock the same. With the thermalright spitfire I could hit 1075/1300Mhz. Most cards no longer have voltage control, but reference models should get between 850Mhz core and 950Mhz core on average without additional cooling. I could have pushed both cards further if I had so chosen.

 

[Graimito90]

Quote:

Originally Posted by staryoshi I used to have crossfire 5850, yes. The results should reflect the benchmarks I've listed in the single card comparison. I would imagine that they would trade blows depending on the app. I've had better experiences with SLI overall than crossfire. (I have used SLI GTX 285, GTX 460 and crossfire 5750, 5770, 5850, and 5850+5870). I would expect better scaling overall with the GTX460 and fewer issues. I used to get "rippling" in games that would only be resolved by using v-sync with the xfire 5xxx series. I have yet to run into that issue with SLI.

Thank's for your answer dude, I think I'm gonna have the change to that SLI.

 

[Aluc13]

+rep thanks for doing this. It looks like they perform pretty decently at a great price ^^

 

[captain_clayman]

Quote:

Originally Posted by HeliXpc Ive had several 5850's, most cannot do 950. 850 most likely.

i saw one guy on youtube get like 1ghz on air with CF 5850's, idk which cards he had though.

 

[staryoshi]

Quote:

Originally Posted by captain_clayman i saw one guy on youtube get like 1ghz on air with CF 5850's, idk which cards he had though.

The best 5850s can hit 1k core on stock cooling, but those are not common. Most can do 950-970 with voltage control. The highest I ran on my stock directcu cooler was 970/1300. Using the spitfire got me to 1070/1300

 

[gunner05]

fantastic review staryoshi. so, looks like the 460 cant be beat at that price point.

 

[xBISHOPx]

Awesome write up +1

.