Review: 32MB vs. 64MB - MX400 Shootout!

32MB vs. 64MB - MX400 Shootout!

NVIDIA GeForce2 MX400 Shootout

Introduction

NVIDIA’s GeForce2 MX400 chipset is a funny one for me. I can’t quite see the need for it, unless they are targeting a very specific price point with the cards.

NVIDIA’s own GeForce2 GTS chipset is clearly quite a bit faster than the MX400, due to it’s increased memory bandwidth provided by DDR memory with a 128bit path to the GPU. It’s also not that much more expensive.

Due to the release of the GeForce3, cards based on the GeForce2 core have tumbled in price with the GTS and MX400 around the same price point at the moment.

However, it does fill the niche for quick, budget orientated cards quite nicely. It’s a good chipset, with a decent speed increase over its sibling, the standard MX, for not much more money.

To highlight the differences in spec between the GTS, MX400 and standard MX, here’s a quick feature list for comparison.

Here we notice that the MX400 offers nothing more over the MX than a 25 MHz increase in core speed. All 3 chipsets utilise 166 MHz memory, however the GTS uses DDR memory for an effective SDR clock speed of 333 MHz, offering twice the memory bandwidth of the SDR impaired MX’s.

We all know from previous experience that NVIDIA cards love memory bandwidth. Increasing the memory clock speed on an NVIDIA card is the single biggest cause of a performance increase. Core clock speed is less important. The GeForce2 core, in whatever flavour is fast. Increasing its core clock offers a performance improvement, but the memory bandwidth is where it’s at.

Like the MX before it, MX400 cards are offered in both 32Mb and 64Mb flavours. NVIDIA recommends 64Mb for MX cards in its specification. What we are going to take a look at today is whether or not the doubling of memory is worth the extra money and does it improve performance significantly.

MX400 has everything we have come to know and love on NVIDIA’s GeForce2 cards. 2^nd Generation Transform and Lighting Engine. NVIDIA Shading Rasterizer, 128 bit memory interface running at 166MHz. 200MHz core clock, 400 million pixels/sec, 800 million texel/sec fill rate, 2.7GB/sec memory bandwidth etc. All the good stuff.

32Mb or 64Mb? What’s better?

“Ryszard, Got a couple of MX400’s for you here to review!” said Dave, Hexus’ Chief. “Sure thing, send them up!” I replied. I was looking forward to seeing whether the MX400 offered any improvement over my current stock MX. I was of the opinion they were another NVIDIA money maker with little difference and improvement over the current budget leader.

Imagine my surprise when upon opening the box, I found OCZ’s TITAN II 32Mb MX400 and the 64Mb Abit Siluro T400 MX400 inside.

A couple of ICQ’s later and the focus of my review is the difference between memory sizes on MX400 cards.

So, what’s better? Should you shell out some extra green on a 64Mb card? Here are some hard facts, the kind we know and love best, 3DMark 2001 benchmarks and some Quake3.

Before we start though, I’d like to go through details on the 2 cards.

Abit Siluro T400 64Mb

You can find a great summary of the Abit T400 on Abit’s website here:

http://www.abit.com.tw/eng/product/mm/mx400.htm

Its standard MX400 fare topped off with a lovely black PCB. Everything is present and correct including NVIDIA’s 2^nd generation T&L engine, AGP4x, TV-Out, 128bit SDR memory interface. Do check out the Abit webpage for full details.

The Siluro, while served well by the 12.90 Detonators I used for the review, was not picked up properly by the Overclocking part of the drivers. I couldn’t choose any memory or core speed at all. Trusty Powerstrip 3 was used for overclocking the cards and worked flawlessly. The OCZ card was recognised however by the Overclocking section of the drivers.

OCZ Titan II MX400 32Mb – 5ns memory

I couldn’t find much technical info about the OCZ card. Apart from being an MX400 card with 32Mb of memory, all I can find to differentiate it from the reference design is 5ns memory. It ships with a Blue Orb cooler for the GPU. They ship versions of the Titan II MX400 with DVI, TV-Out and ram sinks as seen on the MAX version. We are testing the plain 5ns version with no DVI or TV-Out.

The OCZ ships with 183MHz as its default memory clock, compared to the 166 MHz of the Abit. It wasn’t apparent to me that this was the case until I had finished benching the Abit and has popped the OCZ in. Therefore you’ll see a lone benchmark of the Abit at 166 MHz memory speed. Both cards were run at 183 and 200MHz however. I didn’t feel the need for underclocking the OCZ card. The Abit was perfectly happy at 183MHz and as you’ll see, the numbers are so close, that you can guess what the 166 MHz OCZ scores would be anyway! The 5ns memory is the reason for the 183 MHz default memory clock.

Benchmark Setup

• AMD Thunderbird @ 1428 MHz

• 256Mb RAM

• 40Gb IDE RAID0 array using IBM disks

• IWill KK266-R motherboard

• Windows 2000 with Service Pack 2

• 12.90 NVIDIA Detonator Drivers

• Vertical Sync Off in all tests

Benchmarking Results – MadOnion’s 3DMark2001

The results are fairly interesting. Note that I cut off the bottom half of the Result Browser results. They don’t count when calculating final 3DMark2001 score so I didn’t run the tests. Also notice the lone MX400 result from the Abit clocked at 200/166MHz and result from my own GeForce2 MX at 1.5GHz. Notice how the MX400’s at 1.43 handily beat the MX at a higher CPU clock.

As for the MX400 results, clock for clock, the 32Mb and 64Mb cards perform eerily similar. 3DMark2001 uses some high resolution textures and the high CPU speed should max out the MX400’s leaving a pure indication of the cards performance.

However, in 3DMark2001 at least, the extra memory makes no difference in final score. The 64Mb has a slight advantage in the final scores, but nothing to write home about and in all cases, less than 1fps difference.

Benchmarking Results – iD Software Quake3: Team Arena

Just a quick table of results from the MX400 cards at comparable clocks, with the lone Abit 200/166 score for reference. Team Arena was set to 1024x768 in 32bit colour, all options turned on, texture quality to the max, compressed textures, highest geometric details, 32bit textures etc. Anything that would enhance the visual quality was enabled.

I wanted to push the MX400’s because while being budget cards, they are not budget performers. Users shouldn’t always have to take a compromise in visual quality to gain some speed. Budget cards these days are very capable.

As always, scores are taken from an average of 5 runs, with any abnormal results discarded until 5 similar results are measured.

On the type of platform MX400 will likely be used on, I refrained from running at higher resolutions. Not out of laziness, but due to 1024x768 being a sweet spot at the moment as far as gaming is concerned. You can tell where the most popular resolution sits by looking at the common midrange system and its target resolution. I think the most common system sold today is likely to be around a 1 GHz Athlon and an MX/MX400. Performance is good at 1024x768 on a box like that.

Conclusion

So what are we looking at when we consider the MX400? It’s too close in price to the normal MX to disregard on grounds of cost. We can see the performance is better than the MX and they overclock well enough on the memory side of things (well, the two tested cards) to give them that extra memory bandwidth increase that the GeForce2 core loves so much.

They are excellent performers for the price. In the same league are the ATi Radeon VE and the Kyro II boards from VideoLogic and Hercules. The MX400 has the advantage of extremely solid drivers on all platforms. Windows 2000 and XP performance of any GeForce, GeForce2 or GeForce3 card is very close to it’s Windows 98 performance. I can’t say the same for ATi, although its Windows 2000 driver support is much improved from the early days when I had a Radeon 64Mb DDR. The drivers caused me to switch to a GeForce based card (god I miss that blue Hercules Ultra).

With the MX400, in either of the two flavours, 32Mb or 64Mb, you have a great card with potential to be pushed that little bit further. A

couple of notes on the individual card themselves. The OCZ is a miracle of downsizing. It really does have a small PCB compared to the Siluro. I can also honestly say it has the best image quality of any card I’ve used with my current monitor. I don’t know if OCZ use a better RAMDAC than everyone else, but the MX400 gives me a pin sharp, rock solid display, all the way to 1600x1200 in 32bit colour and 85Hz. I’d love to keep it but sadly, I think it’s going back. Kudos goes out to OCZ for making such a great card. As mentioned before, it comes outfitted with a Blue Orb chipset cooler that requires hooking up to a 4 pin Molex or 3 pin motherboard header. Not as convenient as an onboard connector, but nothing too bothering.

The Siluro comes equipped with a lovely black coloured PCB. There is a lot of call for ‘fashionable’ looking cards with the advent of case windows and the like. The case mod scene is alive and well and windows are the norm. Personally, I love it. I have a case window and something like a black PCB really adds something to the looks. For others however, a computer is a tool. It doesn’t have to be like that folks! The Siluro is also a great card. Maybe a little hard to find compared to most manufacturers cards. The same goes with OCZ. I’ve never seen their cards widely available at all. It’s a shame because both are great.

Sure, they are not GeForce3’s, but we don’t all have the resources to land NVIDIA’s finest. For the rest of us to whom the GTS, Pro, Ultra and GF3’s of the world are just out of reach, the MX400 seems to be a solid buy. I stated at the top of the review that I can’t see much need for the MX400 in the marketplace, but it certainly does the job and Abit and OCZ deserve credit for producing great cards based on the chipset. I can easily recommend them as a perfect budget choice for the cost conscious gamer. In that respect, I can find no fault with either card. 10/10 for both cards in the confines of their price point.