Toward Faster Memory
For a long time Nvidia was the precursor when it came to using the latest memory technologies. After the DDR memory used for its GeForce in 2000, the Santa Clara firm was the first to use GDDR2 with the GeForce FX, then GDDR3 with the GeForce 5700. But for awhile now, ATI has been acting as the pioneer — GDDR4 first appeared on its Radeon X1950XT, and now, two years later, it has begun to offer the first card to use GDDR5: the Radeon HD 4870.
There’s no secret when it comes to increasing memory bandwidth. There are two ways to do it: The first consists of widening the data bus, and the second is to make the memory operate faster. The first method comes up against numerous obstacles. A wider bus makes routing on the PCB more complex, and also requires a larger number of pins on the package. Then all those pins have to be connected to the chip, which requires a large number of pads (the interconnections placed around the periphery of the chip). So, a wider bus requires that the die be of a certain size – which is one reason why entry-level GPUs were limited to 128-bit buses for a long time, while their high-end equivalents used a 256-bit or 384-bit bus. Another disadvantage is that it takes more power, which increases the chip’s consumption.
So, it’s no wonder that the option was used parsimoniously. In fact, 128-bit buses were used for a long time on high-end GPUs, from the Riva 128 to the Matrox Parhelia, and the ATI Radeon 9700 four years later. In the same way, the 256-bit bus didn’t get wider until the arrival of Nvidia’s GeForce 8800 late in 2006. And yet, the bandwidth demands of GPUs are constantly increasing, despite the technologies for saving bandwidth that have been perfected with each generation.
So, the solution lies in running memory faster. That’s easier said than done, however, because as with any circuit, there’s a limit to the clock frequency at which memory chips can operate. To get around the limitations, manufacturers have used various tricks. DDR memory enabled transfer of data on both the rising and the falling edges of the clock cycle, doubling the data rate for a given memory frequency. To do that, DDR memory uses what is called a two-bit prefetch — at each memory access, instead of transferring one bit from the prefetch buffers, the DDR memory transfers two. Successive developments in DDR technology have consisted of moving more and more data at a given memory frequency by increasing the width of the prefetch. DDR2 used a 4-bit prefetch, like GDDR3. GDDR4 introduced an 8-bit prefetch.
Seems like this is going to be my new graphics card
Im currently running 2x 3870s in crossfire - would it be better performance (and power)wise to buy a 4870 and run that instead?
I got two of these in crossfire and even with a Q6600 cpu unclocked it makes Crysis fly. Fans are suprisingly quite but Temps are Hot. The back of my PC is almost too hot to touch! Had no problems running any games so far drivers seem quite stable.
an single 4870 is about 2 times more powerfull then your 3870 in CF mode you be able to now use AA as well as thay fixed the AA (thay put Hardware back in there so it can do aa now and very well)
the card seems to beat the 260 alot of the time and thats only one card get 2x 4870 in CF and when CF is working you out do an gtx 280 for the same as 2x4870 in CF (basicly same price)
the 2x and 3x ATI cards was not so good
This is an amazing card. As an avid Nvidia fan, I never thought I'd say that. And yet, here I sit with a smile on my face, content in the knowledge that I am now about to upgrade my ancient 6600GT to the ATI HD4870 which kicks the GTX260's ass at a much better price. In fact I couldn't care if Nvidia made the 260 cheaper, I'd still go out and support ATI for having the balls to make such a good card at such a good price. At the end of the day, a price war is fantastic for the consumer, and in order to keep that up, we gotta support ATI for this round...
Guess what chaps... Nvidia has decided to drop their prices for the GTX 280 by $90 and the 260 by $30, due to the sheer dominance of the HD4870! Check out the link: http://www.tomshardware.com/news/a [...] ,5818.html
Basically the GTX 280 is set end up costing about $559 and the GTX 260 about $359 after the price cuts...
And even after hearing that news, which clearly demonstrates the power of competition, I'm still getting myself a 4870 (perhaps even the 1GB version)... Suck on that Nvidia.
Isn't unreal 3 cuda optimised to run the physics through the GPU, hence leaving the CPU free to help boost the framerates.
I think it is unfair to use a game as a benchmark that has been clearly optimised in this way.
On the other hand if that is the road developers are going to go down then AMD should pull their finger out and get their physics acceleration stuff out.
I'll be getting the 1GB version as soon as I come back from my holiday!
Isn't unreal 3 cuda optimised to run the physics through the GPU, hence leaving the CPU free to help boost the framerates. I think it is unfair to use a game as a benchmark that has been clearly optimised in this way. On the other hand if that is the road developers are going to go down then AMD should pull their finger out and get their physics acceleration stuff out.
The CPU won't give much of a boost to framerates as such, but having the physics somewhere quicker than the CPU will generally help the system as a whole.
Note about memory transfer frequency vs bus width.
Don't forget that power consumption increase with frequency is quadratic, which is not the case with bus widening. The main problems really are
1) technical: routing complexity not only on the PCB (as said in the article) but also on the die,
2) financial: pad buffers take a lot of silicon area, which increases the cost and reduces the yield.
Hi guys, need a little advice...
I'm keen to get the ASUS EAH4870/G/HDMI/512MB but was wondering about the 1GB version that's supposed to be coming out in August. Do you think there will be a major difference in speed if I'm using a 19" wide or 22" wide screen, between the 512MB version and the 1GB version?
If not, I think I'll just get the 512...
Thanks!
Can't wait for the 4870X2
Sooo depressed now... bought a HD3850 six months ago to replace MY 6600GT, and now jackbnimble is making me very jealous... should have waited!
Which is part of the reason why UE3 games (UT3, Mass Effect etc.) have such an abnormally (not anomalously) low score in comparison. BUT there is actually supposed to be a third party working on translating nVidia's Ageia stuff to AMD cards very successfully (albeit illegally) so AMD does have the hardware for it... just not the software license! >
And AFAIK there's a lot of politics and licensing behind the CUDA/physics angle... Intel has Havok and nVidia has Ageia, AMD has the hardware but no code for it
for gaming the 38x0 cards was not worth it as you could not use AA as it killed the card due to not having any texture back enders it was all been done in the GPU core {SP} an 8800gt would of been an better buy as it would of been able the same price, starting from the 4850 well worth it (it can do 4xAA as well)
the physX think well some one has made an ATI>CUDA>physX so that it can be used on ATI hardware Nvidia are actively supporting this,ATI is not supporting it but are not stopping at as well (as its the physX driver and the CUDA thats been modified to make it work)
CUDA is Free and open as why above Nvida is actively supporting it, this should make physics move along faster in game support beening on Nvidia or ATI cards
The card is a great deal.......I own a 3870x2 whic is outstanding in price and quality.Hope nvidia catch up with the current dominator in graphics..... AMD for life and ATI RADEON till death.....
The card is a great deal.......I own a 3870x2 whic is outstanding in price and quality.Hope nvidia catch up with the current dominator in graphics..... AMD for life and ATI RADEON till death.....
I hope AMD didn't put these price hoping that they will sell enough to get theyr money back, what I mean is I hope that 4800 will not be a los of money for them, becouse I hope to see even better cards at better prices! Also I feel sorry for nVidia in a way, they were hard kicked in the ass!
HD4870 can compeate with GeForce 6600GT and 8600M.
9800 gx2 is 87% faster than HD4870.
HD4870 is same fast than HD3870.
Some tests from Apricot Optimizing GLSL;
http://www.yofrankie.org/?p=261#comments
Sorry wrong link in 30/07/2008 08:59
http://www.yofrankie.org/?p=258#comments