Monday, March 26th 2007

Qimonda Skips GDDR4, Goes GDDR5

In order to outmaneuver its large competitor Samsung, memory chip vendor Qimonda plans to skip the GDDR4 graphics memory technology generation in favor of GDDR5. With the move, the company seeks to address the high end of the market. "While GDDR3 presently holds a share of about 90 percent of the high end PC graphics market, in 2011 the mainstream memory technology in this segment will be GDDR5 - not GDDR4...Most customers will move from GDDR3 directly to GDDR5." said Feurle. While GDDR4 offers several improvements in terms of performance and feature set over GDDR3, GDDR5 is intended to offer the missing low-power capability, along with tripling the performance of today's 800 MHz GDDR3 chips. In addition, it will offer features that enable more robust system designs, Feurle noted. The company presently is pressing ahead with the JEDEC standardization process and expects the standard to be finalized by summer 2007, with mass production scheduled for the first quarter of 2008.Source: EETimes
Add your own comment

7 Comments on Qimonda Skips GDDR4, Goes GDDR5

they never stop;i'm happy with my "old"ddr 433 is running like ddr2 almost;when i build a sys i have in mind to use it at least 3 years;if they make this ddr5 will be another chain reaction;don't understand me wrong i'm not against development i'm angry because we're forced to open the wallets sooner or later.

btw timings 10-10-10-30 :laugh:
Posted on Reply
Remember this is aimed at the graphics industry, and not DIMMs.
Posted on Reply
Yes, but his timings ARE funny :roll:
Posted on Reply
Honestly, I don't really see the benefit of GDDR4 over GDDR3 at the moment. In particular the x1900xtx vs 1950xtx only showed a 1-4 FPS increase in frame rates. IMO, if this translation is correct on other video cards GDDR4 is not worth the premium for new tech and should be scrapped. Also note, in some cases when using 512Megs of ram GDDR4/GDDR3 on the video card can actually slow down frame rates vs. 256 Megs. This suggests that the memory performance itself is simply to slow and a bottleneck for the GPU.
Posted on Reply
Bird of Prey
I think they are moving way too fast on the memory front. I dont see a need, or a bottleneck really, on graphics cards. I havent experienced DDR2 memory yet to see if they have a bottleneck with any current cpus. Does anyone else fascilitate a need to move ahead this quickly?
Posted on Reply
Wow, if the best processors today are a bottleneck for the 8800 series graphics cards, how much worse will it be for this?
Posted on Reply
Completely Bonkers
The GDDR5 memory spec is fantastic news for GPU. Currently, high end GPU's are getting memory bandwidth constrained (at 256bit), and mid end cards ARE ALREADY bandwidth constrained (at 64bit). By trippling the bandwidth, you can obtain much higher performance on a narrower bus. A narrow bus makes a GPU smaller and cheaper. By alot.

Or, for extreme high end GPU, you get a lot more throughput. Which means you can throw much larger textures around (nice candy) without a performance penalty.

I like it.

The lower power consumption is greatly needed. If we double the memory capacity... then we can keep power levels the same with ULV GDDR5. Or, same memory capacity, with lower power, means less heat, equals greater overclock potential.

I like it alot.
Posted on Reply