ASUS RADEON EAH5850 Print E-mail
Written by Michael Schuette   
Oct 21, 2009 at 11:04 AM

Do you like bold? Bold is beautiful, especially in the case of bold styling. A red and black Dodge Challenger is beautiful because its styling is bold and because it has a lot of power under the hood. There is an interesting parallel between the styling of graphics cards and the styling of automobiles. Three years ago, everybody displayed their heat pipes and the more surfaces and convolutions there were, the better the product had to be. The tables have turned, though, it is the confidence of a winner that dictates a rather sober styling of graphics cards now. In the case of today’s contender, that is, ASUS’ RADEON EAH 5850, it is about as simple as it is understatedly yet decidedly bold.

To give credit where credit is due, the reference design was done by ATI and all current cards are following it. Notwithstanding, there are differences between manufacturers, primarily with respect to features, warranty and some of the bundled software. ASUS advertises the EAH5850 to be equipped with their “voltage tweak” technology, enabled in software through the “smart doctor” utility, which also allows independent overclocking of core and memory. To ensure that everybody feels safe about these features, ASUS backs its products with a three year warranty. Similar to XFX, Sapphire and HIS, ASUS also includes a coupon for a complimentary copy of Colin McRae Dirt2, to be released on December 11, 2009 through Steam.

RADEON HD 5850 Technology At One Glance

In short, the GPU used in the 5850 series is the same as that used in the 5870 series, albeit running at a lower clock speed and with 2 of the 20 SIMD arrays disabled. Each of the SIMD arrays features 80 stream processors, hence the total number of ALUs comes down from 1600 to 1440. Since each of the array is associated with a texture unit, the number of the latter also drops to 18 units capable of generating 72 texels per clock (four texels per unit and clock). In classic ATI fashion, the naming convention calls for the "Pro" suffix, so that we have a Cypress Pro GPU powering the 5850 series as opposed to the full-fledged Cypress GPU on the 5870 models.

The other difference between the 5850 and its big brother is the lower clock speed of the GPU and memory. Whereas the Cypress GPU is typically clocked at 850 MHz, the Cypress Pro is chucking along at a slightly more relaxed pace of 725 MHz for the Core, likewise, the memory clock is reduced from 1200 MHz to 1000 MHz or 4Gbps data rate using GDDR5.

For a quick comparison of the raw technical specs of the current offerings in the high-end graphics sector, we put everything into an "easy-viewing" table:

GPURADEON HD4870 RADEON HD5850 RADEON HD5870 GeForce GTX280 GeForce GTX285
Manufacturing Process55 nm 40 nm 40 nm 65 nm55 nm
Graphics Core Clock Speed750MHz 725MHz 850MHz 602MHz648 MHz
Stream Processor Clock Speed750MHz 725MHz 850MHz 1296MHz1476 MHz
Number of Stream Processors800 1440 1600 240240
Memory Clock/Data Rate0.9/3.6
Memory Bus Width256-bit256-bit256-bit 512-bit512-bit
Memory Bandwidth115.2 GB/s128 GB/s153.6 GB/s141.7 GB/s159 GB/s
On-Board Memory0.5/1GB 1GB 1GB 1GB1/2 GB
ROPs16 32 32 3232
Texture Units40 72 80 8080
Texture Rate30.0 GTs 52.2 GTs 68 GTs 48.2 GTs51.8 GTs
Fill Rate Rate12 GPs 23.2 GPs 27.2 GPs 19.3 GPs21.4 GPs
Power Connectors2x6-pin 2x6-pin 2x6-pin 1x6/1x8-pin2x6-pin
Max Power160W 151 W 188 W 236 W183W

Aside from other architectural idiosyncrasies, one of the most obvious differences between the current nVidia and AMD strategies is the memory interface and choice of memory components. AMD has very early on embraced GDDR5 with four transactions per clock whereas nVidia's current offerings are still using GDDR3 with 2 transactions per clock. At the same time, AMD's memory interface is only 256 bit wide whereas nVidia uses twice the bus width so that by the end of the day, the net result is the same. At least in the same ball park.

Last Updated ( Feb 14, 2010 at 02:15 AM )
<Previous Article   Next Article>