Home » Opinions » ATI HyperMemory vs nVidia TurboCache (Updated)

ATI HyperMemory vs nVidia TurboCache

by | Go to comments

Share:

Update:
This is a revised version of a review that originally appeared on 28 Apr 2005. Due to a mistake by nVidia we were supplied with a 32-bit TurboCache card to test instead of a retail 64-bit card. As the scores in the original review are therefore no longer valid we have retested and reworded the review accordingly. nVidia and TrustedReviews apologise for the error.

_ _ _

As a technology journalist and as an enthusiast, when it comes to graphics, the high end market is where it’s at. Like the latest cars, the faster and more heavily featured the better. However, as you’ve probably noticed, there are more Vauxhall Astras, than there are Ferraris on the road. The meat and potatoes of the huge graphics market is all about cheap and affordable solutions. The fact that Intel, with its integrated solutions is the number one vendor in the market tells you all you need to know.

However, as users become more educated, integrated graphics on the desktop has become less appealing. This has created a space in the market for low end but branded graphics cards and this is where ATI’s HyperMemory and nVidia’s TurboCache technology come in. The idea behind the technologies is that the graphics processor shares system memory with a small amount of local on-board memory, dramatically reducing costs. The ATI card comes in two flavours – one with 32MB of local memory that's able to scale up to 128MB with system memory and another with 128MB of local memory that can scale to 256MB. nVidia’s TurboCache meanwhile has three versions – a 16 and 32MB version that scales to 128MB with system memory, and a 64MB that goes to 256MB.

Interestingly, we discovered (the hard way) that there are also 32-bit versions of TurboCache. These are nominally aimed at the OEM market, but as our results prove they have also found there way into retail boxes, such as one we have from AOpen. However, this has no mention on that box that it is a 32-bit card, so the only thing we can say is 'Caveat Emptor' (Buyer Beware!).
/94/ee5570/6f12/1334-1.jpg
ATI first announced its technology in September 2004, with nVidia following in December. However, nVidia’s TurboCache cards seem to be readily available now while ATI’s cards don’t seem to have hit the market. Whether this is down to lack of interest from card manufactures or technical delays is uncertain. Either way, we’ve now got two equivalent 128Mb cards in our labs and a 32-bit 256MB supporting TurboCache card, so we threw them into a pit, or at least a testbed system, and let them battle it out.

Before we get down and describe the technology though, a word about the names. Where do they get them from? The HyperMemory brings to mind Lando Calrissian in the Millennium Falcon shouting, “they told me they fixed it! They were supposed to have fixed it! It's not my fault!” As for TurboCache - wasn't it Michael Knight’s favourite button on KITT’s dashboard?
/94/c25e8b/b095/1334-2.jpg
The idea of dynamically using system memory isn’t a new one and the feature was actually built into the specification for AGP. However, the bandwidth wasn’t really up to making the memory swapping transparent, and the technique was effectively ignored by the industry. So if the technology snob inside of you groans at the thought of cheap low-end graphics chips then don’t blame us. Blame PCI Express, as it is this technology that has made these cards possible.

On both chips the memory management has had to be heavily tweaked to enable the memory to be dynamically addressed across the PCI Express bus. This is no mean feat and requires changes to the hardware and good driver support. So you’re just going to have to face up to it, low-end graphics card are not only big, but now they’re clever too.

Go to comments
comments powered by Disqus