Graphic Cards Hypermemory In GPU Cards

A

Amarbir

Skilled
Guys ,

Many Times Our suppliers Have Cards With Hyper Memory Meaning is The Card Has Real RAm 256MB They Will Have a Card Saying XYZ Card With 512 MB Ram With a Small Star Saying 256 True And 256 Hypermemory .I Am sure That If We Take a NEw PCI Express Card With 256 Mb Ram And Use it In a PCI Express sLot Or a AGP Slot It Can Take Main Memory Also When Needed ,I Mean Ram From Main Memory .Today To My surprice i Get a Mailer FRom Supplier XYZ Card With 2.8GB Hyper Memory :huh: DAmn
 
HyperMemory is ATI Technologies' method of using the motherboard's main system RAM as part of or all of the video card's framebuffer memory on their line of Radeon video cards and motherboard chipsets. It relies on new fast data transfer mechanisms within PCI Express.

However, to make up for the inevitably slower system RAM with a video card, a small high-bandwidth local framebuffer is usually added to the video card itself. This can be noted by the one or two small RAM chips on these cards, which usually have a 32-bit or 64-bit bus to the GPU. This small local memory caches the most often needed data for quicker access, somewhat remedying the inherently high-latency connection to system RAM.

The local and system memory areas are not noticeably separate to the user and often HyperMemory solutions are advertised as having as much as 512 MB RAM when this is actually referring to the potential use of system RAM.
Click to expand...

Source :HyperMemory - Wikipedia, the free encyclopedia

So u are referring to something from ATI :)

P.S. Wiki also said TurboCache is what it is for Nvidia.....same technology basically....rivals using different terminology :lol:
 
^true as far as the system has that much ram :)

for example a person taking the card with the 2.8gb hypermemory and having only 2gb of overall system memory.

so its more like the ISP's they should advertise it as *upto 'whatever' mb/gb to give a clear picture to a end user.
 
Usually such a feature is on low end cards to keep the cost of the card down i guess. Versions of the card with larger on board memory will be faster.
 
Zloyd said:
Usually such a feature is on low end cards to keep the cost of the card down i guess. Versions of the card with larger on board memory will be faster.
Click to expand...

Well ,

My Question is That If For Example a 4350 Card Advertises 2.8 GB Hyper Memory And The Client in Question Does Have this Much Memory its Not BAd Right .
 
Well its like this if there are 2 cards say 4350 one is 256 mb and the other is 512 but the 256 has the hyper mem feature and the 512 does not.

The 512 will be the faster card.

(I know this from the nvidia turbocache days , im not sure its all the same as the ati version)
 
^now that depends , people have a VERY WRONG perception about this RAM thing on graphic cards, give some a 8600GT with 1GB RAM and they go dancing away, give them a faster card with 512MB DDR4/5 whatever they will say , areeee only 512mb :mad: woh 1 gb walla do....

so if a person takes the card all happy thinking 2.8gb of hyper mem and stuff and still aint able to run some stuff on it then the persons gonna go bonkers thinking so much memory still wont run.....

this will really be usefull for graphic rendering or 3d work i guess. and really doubt if any game would need onboard and 2.8gb of hyper mem , for such cards the core speed and the onboard ram would be so fast that hyper mem would never get full :lol: so basically , its all abt the marketing. pls correct me if im wrong.
 
Hypermemory is useless. A card with 256MB main memory will outrun a card with 2GB Hypermemory/Turbocache.

Since it shares memory with the system, the overall system response slows down. Performance is similar to an integrated graphics chipset with sideport memory - whereas Sideport is a step up from regular integrated graphics, shared memory technology is a step backward from discrete cards.

The seemingly large memory amount is rendered useless due to the slower bus speed. Memory in a GPU is only used to render on-screen textures, not for actual storage. For a screen up to 168x, 256-512MB is more than sufficient for most applications. After that, it's only the memory speed/latency and bus width that make an actual difference to application performance.

A card with hypermemory is around the performance of integrated graphics. Not advisable to spend money on.
 
Back
Top