NVIDIA’s GeForce GT 220: 40nm and DX10.1 for the Low-End
by Ryan Smith on October 12, 2009 6:00 AM EST- Posted in
- GPUs
Conclusion
There’s really no way to sugar-coat this, so we won’t: the performance of the GT 220 is abysmal. Or rather, the pricing is.
The GT 220 is a value card, and that’s something we can appreciate. But for a value card to be a good value, it needs to be at the right price. NVIDIA wants to see these cards sell for $69-$79, with the best cards (those with GDDR3) selling at that $79 price point. However depending on your taste for rebates, we can get a Radeon HD 4670 for between $59 and $69, or on the NVIDIA side a 9600GT for between $69 and $85.
Either one of these cards is simply going to beat the GT 220 silly; it was never meant to compete with a 64SP NVIDIA card, or a 320SP AMD card. And don’t even get us started on the 4850 that Newegg is selling for $85 right now…
Certainly the GT 220 has some positive points. We’re glad to see that NVIDIA has finally ditched the S/PDIF cable and gone internal to enable additional HDMI audio formats, and the ability to finally offload MPEG-4 ASP decoding to the GPU is intriguing. Similarly we’re happy to see DirectX 10.1 support arrive on an NVIDIA part, and the 7W idle power usage on this card is amazing.
But so many of these things are just catching up – AMD had a card that could do DX10.1 and additional HDMI audio formats a year ago. The only thing NVIDIA has going right now is that they’re benefitting from this being a 40nm product, thanks to the lower power usage and lower production costs.
Ultimately we think this has the makings of a very good HTPC card. It’s the quietest actively cooled card we have, it runs cool, and it’s the only thing that can offload MPEG-4 ASP (or at least, will be once support is enabled). But we just can’t justify paying this much more for less performance, especially when there are passively cooled 4670s that can meet/beat the GT 220’s acoustic performance. It’s frustrating to see what’s going to be a very good HTPC card price itself right out of the market.
At the end of the day the GDDR3 GT 220s need to be priced at under $60 to be performance competitive with existing AMD and NVIDIA cards. And the cards with slower memory should be priced even lower (then again, when did memory configurations ever make sense at the low-end?). But at this point such a thing is basically a pipe dream.
As for the Palit GT 220 Sonic Edition that we’re looking at today, Palit really can’t do anything to escape the GT 220’s larger problems. With its slight factory overclock it’s going to be among the fastest GT 220 cards, but it’s a bit like being the king of a desert island. It will get you respect, but it doesn’t mean that very many people are going to want what you have.
This is turning out to be a rough fall for NVIDIA. AMD has them undoubtedly beat in price-to-performance on the high-end of the market. And with today’s launch of the GT 220, it looks like AMD has them beat on the low-end too. There are very few NVIDIA cards that offer the right level of price and performance right now; the GT 220 is not one of those cards.
We’d like to once again thank Palit for providing us with their GT 220 Sonic Edition for today’s launch
Facebook
Twitter
RSS
80 Comments
View All Comments
Joe90 - Monday, October 12, 2009 - link
Last week I bought a Medion P6620 laptop. It's fitted with an Nvidia 220M graphics card with 512MB of GDDR3 ram. The blurb of the box said it was a DirectX 10.1 card. Have NV released any details about the laptop 220M as well as the desktop 220? All I could find out from Wikipedia & Notebookcheck.net was that the 220M was a 40 micron version of the 9600M GT but the Anandtech review suggests that things are more complicated than this.My only contribution to the debate is that I picked up this laptop for a paltry £399. Given that my machine has a T6500, 320GB HD, all I can think is that NV has be giving away the 220M to OEMs for pennies!
Finally, has anyone seen any signs of Windows 7 drivers for the 220M? I got a freebie upgrade but I'm a bit scared of using it just in case there were no drivers for a graphics card that NV don't recognise on their own website.
JarredWalton - Monday, October 12, 2009 - link
I'm assuming http://www.nvidia.com/object/notebook_winvista_win...">these won't work? I know they don't list 220M, but they're the latest official drivers from NVIDIA; otherwise you're stuck with whatever the notebook manufacturer has delivered. I'd guess NVIDIA will have updated mobile notebook reference drivers relatively soon, though, to coincide with the official Win7 launch.Joe90 - Tuesday, October 13, 2009 - link
Many thanks for the link. I think you're right. Although this NV driver page refers to the '200 series', it doesn't explicitly mention the 220M card. Until it does, I'll hang back from doing the Windows 7 upgrade.As it happens, I've only just made the leap from XP to Vista. Despite all the bad stuff I've read, Vista seems ok so I might just stick with it for a while. I'm finding I can play games ok with the Vista/220M combination. I'm currently playing FEAR 2 at 1366 x 768 with all the detail settings on maximum and the game plays fine. Given the low price of the laptop, I'm more than happy with the graphics performance.
JarredWalton - Tuesday, October 13, 2009 - link
I've never had a true hate for Vista, but there are areas I dislike (several clicks just to get to resolution adjustment, for example). Windows 7 is better in pretty much every way as far as I can tell, although it's not like Vista is horrible and Win7 is awesome. It's more like Vista reached a state of being "good enough" and 7 addresses a few remaining flaws.Now, if someone at MS would fix the glitch where my laptop power settings keep resetting.... (Every few tests, my battery saving options will reset to "default" and turn screen saver, system sleep, etc. on after I explicitly disabled it for testing. Annoying!)
plonk420 - Monday, October 12, 2009 - link
it's AMAZING for this gen's "lowest end" (at retail)... beats the pants off my 9400GT currently in my HTPC. probably even equals or betters my once only 2 generation old nearly high end HD3850.this is going in my HTPC in a heartbeat when it hits $30-35 (unless power consumption is stupid high, which i don't think it is. i THINK i read about this somewhere else, and someone was moaning about it using 4 watts more than some ATI part it was being compared to).
ltcommanderdata - Monday, October 12, 2009 - link
Given the more competitive desktop landscape, these new 40nm DX10.1 chips are not impressive at all. However, their mobile derivatives were announced months ago, and in the mobile space where mainstream GPUs seem to be made up of many, many combinations of nVidia's 32SP GPUs, a 40nm 48SP DX10.1 GPU would actually bring something to the table. It'd be great if we can get a review of a notebook with the GT 240M for instance.JarredWalton - Monday, October 12, 2009 - link
It looks like there are laptops with GT 240M starting at around $1100, and the jump from there to a 9800M card (96SPs, 256-bit RAM) is very small. Unless you can get GT 240M laptops for around $800, I don't see them being a big deal. Other factors could change my mind, though - battery life perhaps, or size/form factor considerations.apple3feet - Monday, October 12, 2009 - link
For some people, the most important detail of any new NVIDIA card is the CUDA Compute Capability. Will it run my scientific simulation code? Only if the CUDA Compute Capability is 1.3 - so that it supports double precision floating point arithmetic. Could we have just one line somewhere to tell us this vital info?jasperjones - Monday, October 12, 2009 - link
LOL,for dgemm (double-precision matrix multiplication) the GTX 285 is only about twice as fast as a Core i7 920 (using CUBLAS and Intel MKL, respectively).This suggests you shouldn't run your double precision code on a GT 220. Hell, even running CUDA code on your CPU using emulation might be faster than running on the GT 220..
apple3feet - Wednesday, October 14, 2009 - link
You're wrong. Given the right problem, tackled using the right algorithm, and well-written code, TESLAs can do 20x the throughput of a Nehalem - even in double precision.