Review: BFG NVIDIA GeForce 9800 GTX. Got £200 for a graphics card? Read this

Final thoughts

NVIDIA's GeForce 9800 GTX is a product borne from the knowledge that the company has no real competition in the discrete graphics-card market at around the £200 mark.

Evangelised as the successor to the excellent GeForce 8800 GTX, the new SKU is, in fact, a backward step in some respects. It features significantly lower memory bandwidth and a smaller frame-buffer - both of which are cost-cutting measures to keep production values down.

The GeForce 9800 GTX can be thought of as an incremental upgrade to the 8800 GTS 512; the architecture is practically identical, save for three-way SLI support, slightly faster clock-speeds and a few features borrowed from other 9-series cards, such as HybridPower and a software-upgraded PureVideo HD engine.

Now we can see why NVIDIA was keen to stop its partners releasing hugely-overclocked GeForce 8800 GTS 512 cards a few months' back, because such a move would have rendered the new GTX redundant. Why buy this card when a partner-overclocked 'GTS 512 is shipped with higher core and shader speeds but with a lower street price?

Priced at around £220 for a default-clocked model, the 9800 GTX comes in at around the same financial outlay as the erstwhile GTX, yet, on balance, it's no faster than a design released 18 months ago.

NVIDIA's GeForce 9800 GTX undercuts ATI Radeon HD 3870 X2's pricing but generally matches its performance. In that respect, it's a reasonable card, but we really had hoped for more, much more, from the all-new GTX.

Looking to spend around £200 on a card? We'd recommend readers look for a factory-overclocked GeForce 8800 GTS 512 for around £180 and put the ~£40 saving towards some other part of your system.

We could make a case for users that absolutely need provision for three-way SLI, but they're few and far between, frankly.

What really needs to happen for the GeForce 9800 GTX to become an appealing proposition is for partners' pricing to be slashed to sub-£200, with all other high-end 8-series cards falling further down, right in step.

BFG's card is based on the reference design and, as such, all previous commentary applies to it, too.

Bottom line: the GeForce 9800 GTX should have been called the 8900 GTX 512. The card's performance is no better than the SKU it effectively replaces and too close to one that's significantly cheaper and readily available in factory-overclocked SKUs.

HEXUS Where2Buy

The card is currently on pre-order for £218.50, including VAT.

HEXUS Right2Reply

HEXUS invites manufacturers to comment on our review's findings. If any of BFG or NVIDIA's representatives wish to do so, their HEXUS Right2Reply will be printed here, verbatim.

HEXUS related reading

Login with Forum Account

Don't have an account? Register today!

the bang4buck page on the graph at the bottom shows that it performs worse than an 8800GTX..

it OCs very well though.. nearly 2.5GHz memory speeds :O

I hate the 1st April, you sit there with a reasonably convincing article but, at the back of your mind you're thinking “there's no reason for this to be true”

Would be a good buy if the 8800GT and GTS 512mb didnt exist.

Ok, something has been puzzling me for the last few reviews…

… why dont you use the same system to test all of the components on? You have the ATI cards running on one system and Nvidia on another. And they dont use the same memory.

Also the drivers you are using are in some cases old (october 2007), as are the game patches and because of this I am not even sure if the results give a true reflection on what the current comparisons are.

Whats going on? Some clarification on the impact of the above things would be appreciated.

Oh, and on the system setup page the square advert overhangs the table in Firefox 2 unless you have the window really wide.

I have other questions but i'll leave it at that for now.

Edited. Didnt mean for my original post to sound as harsh :) Just need some clarification so i can understand what the review is or isnt telling me. thx

choc
… why dont you use the same system to test all of the components on? You have the ATI cards running on one system and Nvidia on another. And they dont use the same memory.

In all of our testing, we do our best to ensure that the results are as comparable as can be.
However we need to balance this against the need to test components in a fair and appropriate environment.
Part of doing this involves using the various chip manufacturers recommended platforms for testing their products.
In this case; up until just very recently, the recommended platform to test high-end NVIDIA GPU based cards on, was to use an NVIDIA nForce 680i SLI chipset motherboard.

Likewise we use the MSI X48 based for similar reasons as it's an appropriate Intel CPU based, AMD CrossFire certified platform.
We'd preferred to have used a DDR2 based motherboard, however at the time (at the beginning of January), the X48 boards that were available were few and far between and most were DDR3 based.
In real-world game play you'll be hard pushed to tell the difference between identically equipped machines, with equivalent speed DDR3 vs. DDR2 modules anyway, let alone be able to reliably measure the difference in frame-rates, hence this small compromise was deemed acceptable.

In my experience (and indeed some of my colleagues), we've found it unwise in the past to mix video drivers from NVIDIA and AMD on the same software build.
In some cases it can result in additional crashing whilst benchmarking and sometimes even problems during driver installation.

nForce 790i Ultra SLI boards recently became available at retail, and thus we'll begin evaluating these boards soon with a view to upgrading the NVIDIA graphics test platform.

choc
Also the drivers you are using are in some cases old (october 2007), as are the game patches and because of this I am not even sure if the results give a true reflection on what the current comparisons are.

Due to the sheer volume of test permutations required; to test each and every card across multiple games and resolutions, it's not feasible to always run with the latest drivers.
However, we do check the release notes of all newly released video drivers before running any new benchmarks results, and if we find that there has been a performance improvement, or fix related to any of the games we use, then we update to that driver and mention the reasons in the system setup and notes page, or the relevant results page of the review.

We also keep an eye on the game patch level and if there are bug fixes which may affect performance, or GPU vendor related fixes e.g. fixes specific to NVIDIA or AMD GPU's then we will patch the game to the latest version.