It's that time again (turbo cache)

I looked this time to nvidia to explain what it was, but they don't on their website. I don't blame them I guess since it sucks but anyway.

Turbo cache basically gives the graphics card a direct link to the memory through the PCI express bus which circumvents the CPU.
Let's say your card has 128mb of onboard memory, if your chip is fast enough that it runs out of memory turbo cache makes sure there is a slice of your system memory for it to access instead, it's like a VIP access all areas pass, it does not have to wait and ask to get in, it can use it anytime it needs to which helps take load off the CPU and decrease latencies (waiting time) for the graphics chip.

However that does not take away the main penalties of sharing ram:

1. The PCI express access still happens over the main bus of the memory (the only way to access this). This is usually fully used by the CPU and rest of the system. Taking away their bandwidth hurts performance.
2. There is only so much system memory, taking 128 or 256mb of system memory away from everything else can hurt performance if you are already running low.
3. System memory is designed for a different purpose to graphics memory so even if the graphics card had exclusive access to the memory bus it would still be worse. It is designed to be a fast response lower bandwidth design while gfx memory is less fast to respond but has a MUCH higher bandwidth.
4. The system memory is further away (both physically and if you laid out what sub-systems it had to go through) which increases laitencies, true turbocache has reduced this a bit, but it still has to go through the system's memory controller etc.

 

So turbochache is not the reason 8600M card run faster on 3GB, right?

 

Saying that TC sucks is complete opinion. I'd much rather have a card with TC memory then going through normal shared memory way. And TC only happens when you have the memory to use it.

So if you have 3GB of memory, and you card already have 256mb of memory it grabbing another 256mb for something isn't going to hurt anything. And if the entire system and CPU is using the entire bandwidth it has, well something is wrong with your system or you just shouldn't be playing that game. The performance loss do you have is outdone by the performance gain.

 

With cards that can use more than 256mb, having to resort to turbo cache means its game over anyway. You are not going to notice the difference. Sure it helps the buget cards limp along but dont see turbocache and get excited.

 

I don' think turbocache is bad, between ATI and NVIDIA, nvidia's solution for sharing system memory is alot better.

 

What are you talking about????? ATi, and Nvidea do the exact same thing, they just have different names, ATi is Hypermemory, and Nvidea is Turbocache.

 

Well, actually...they are kinda different..

TurboCache uses hardware techinuques to allocate system RAM
HyperMemory uses software techinuques to allocate system RAM

and generally TurboCache is better, though both are useless inmyopinion

 

Yes, as previously stated, Nvidia uses hardware, so it is generally better. I didn't mean to say alot better.

 

basically, nvidia's system is direct memory access (bypasses the cpu)

ati's hypermemory goes through the cpu. any ati card can use hypermemory. it is software based.

 

The requests for access to the mem are still fast tracked though