 | |||
|
|
|
|
|
 |
GeForce GT 240 2-Way Roundup December 03, 2009 Brandon Sandman Bell |
|
 |
|
| |
Summary: The GeForce GT 240 is intended to replace the 9600 GT in NVIDIA's graphics lineup, and with more shaders, GDDR5 memory, and DX10.1 it could have all the right credentials for the job. See how it fares as we look at two factory OC'ed retail cards based on NVIDIA's latest GPU in this article!
 GeForce GT 240 2-Way Roundup | Page:: ( 1 / 15 ) |
[image]
| <% print_image("01"); %> |
Introduction
This couldn’t be further from the truth.
The truth is you have to spend your money on the right components. Focus on the components that have the greatest impact on gaming performance like your graphics card; you don’t need to spend all your money on the CPU and RAM, as the GPU plays a bigger role on your frame rates.
You don’t have to spend a ton of money on your GPU either. If you’ve got a 20” or 21” LCD running at 1600x1200 or 1680x1050 – which is probably the most common configuration amongst most users – then you don’t need to shell out $300 or more for ATI or NVIDIA’s latest and greatest graphics card. In fact, thanks to the astounding number of price cuts we’ve witnessed over the last year, you can purchase a pretty capable graphics card for around $100. At one point, Radeon 4850s could easily be found for $100 or less, but supplies of these cards have dried up tremendously at the retail level, leaving NVIDIA’s GeForce 9800 GT to enjoy all the sales in this price range.
The GeForce 9800 GT, and its predecessor, the 8800 GT, will probably go down as NVIDIA’s best mainstream graphics card since the GeForce4 Ti 4200. In fact, you could make an argument that it’s one of the most significant two or three mainstream cards ever made. By mixing 112 stream processors with high clock speeds and a 256-bit memory interface, NVIDIA concocted an excellent 3D performer and priced it to move.
The 8800 GT delivered such an outrageous price/performance ratio the card was sold out for months after its release. The G92 GPU it’s based on has since been used in a variety of cards ranging from other budget cards like the GeForce 8800 GS to the (once) high-end 9800 GTX and GeForce GTS 250.
Four months later, NVIDIA followed with an even cheaper 8800 GT derivative, the GeForce 9600 GT. The 9600 GT utilized NVIDIA’s smaller, more affordable G94 GPU. G94 had half the number of shaders as G92, just 64, but NVIDIA offset this by cranking up the clock speeds even higher: topping out at 1650MHz, the 9600 GT’s shaders run 125MHz higher than the 8800 GT’s. Just as importantly, its memory subsystem carried over unchanged from the 8800 GT. NVIDIA implemented the same 256-bit memory interface and 900MHz memory.
Because of these tweaks, the 9600 GT was able to perform well. It blew away competing cards in its class and was even capable of giving the 8800 GT a run for its money in some cases.
But those days are now long gone. The 9600 GT is nearly two years old and as anyone who follows the PC hardware industry can tell you, that’s an eternity in this biz. A replacement for the 9600 GT was expected last Spring, but thanks in part to TSMC’s issues with 40-nm, that part, the GeForce GT 240, is just now seeing widespread release.
Replacing a card like the 9600 GT is never easy though, especially in today’s market where consumers expect to get more performance for the same, or less money. GeForce 9600 GT cards typically sell for anywhere from $80-$100 before rebates, so the GeForce GT 240’s $99 MSRP already misses that mark.
[image]
| <% print_image("02"); %> | <% print_image("03"); %> |
The GeForce GT 240 sports more stream processors than the 9600 GT though – 96 versus 64 – and boasts an onboard audio controller with support for uncompressed 7.1 LPCM HD audio. The 7.1 LPCM audio stream can be sent natively via HDMI, negating the need for a dedicated pass through audio cable. In terms of power consumption, the card draws just 9W of juice at idle. Max board power is 69W.
The clocks aren’t quite as high as the 9600 GT though, and the 256-bit GDDR3 memory interface is replaced with a narrower 128-bit interface using GDDR5 or GDDR3.
That’s just scratching the surface of the differences though. Let’s go over the specs in more detail…
| GeForce GT 240 Specifications | Page:: ( 2 / 15 ) |
Like GeForce 210 and GT 220, the GeForce GT 240 is built on TSMC’s 40-nm manufacturing process and supports DirectX 10.1.
Up to this point, the chip has been offered as an OEM-only product, limiting sales to system vendors who may want to use the card in specific PCs (HTPCs for instance). Now that TSMC’s 40-nm process is more mature though, supply has increased to the point where NVIDIA’s opening sales to the retail level.
Officially NVIDIA’s slating the GT 240 between the GT 220 and 9800 GT on their food chain of products. This slot was previously held by the 9600 GT.
On paper, the GT 240 has a couple of advantages over the 9600 GT. While NVIDIA still doesn’t tout DirectX 10.1 as a compelling addition over DX10, the newer shader model support should give GT 240 an edge in titles that can take advantage of it. Thanks to its 40-nm manufacturing process, the GT 240 also boasts lower power consumption than the 9600 GT and doesn’t need an external power connector.
The following spec chart summarizes the rest of the differences between the GeForce GT 240 and the 9600 GT:
| GeForce 9600 GT vs GT 240 Comparison | |||
| GeForce 9600 GT 96W | GeForce 9600 GT 59W | GeForce GT 240 | |
| # of Stream Processors | 64 | 64 | 96 |
| Graphics Core Clock | 650MHz | 600MHz | 550MHz |
| Stream Processor Clock | 1625MHz | 1500MHz | 1340MHz |
| ROPs | 16 | 16 | 8 |
| Pixel Fillrate | 10.4 GPixels/sec | 9.6 GPixels/sec | 4.4 GPixels/sec |
| Texture Units | 32 | 32 | 32 |
| Texture Fillrate | 20.8 GTexels/sec | 19.2 GTexels/sec | 17.6 GTexels/sec |
| Memory Clock | 900MHz | 900MHz | 1700 MHz GDDR5, 1000MHz GDDR3, 900MHz DDR3 |
| Memory Size | 512 MB or 1 GB | 512 MB or 1 GB | 512 MB or 1 GB |
| Memory Interface | 256-bit | 256-bit | 128-bit |
| Memory Bandwidth | 57.6 GB/sec | 57.6 GB/sec | 32.0 GB/s (GDDR3) 54.4 GB/s (GDDR5) |
| DirectX Support | 10.0 | 10.0 | 10.1 |
| SLI Support | Yes | Yes | No |
| GeForce 3D Vision Ready | Yes | Yes | Yes |
| NVIDIA PhysX Ready | Yes | Yes | Yes |
| NVIDIA CUDA Ready | Yes | Yes | Yes |
| OpenGL Support | 2.1 | 2.1 | 3.2 |
| Standard Display Support | Two Dual Link DVI HDTV | Two Dual Link DVI HDTV | Dual Link DVI, VGA HDMI |
| HDCP | Yes | Yes | Yes |
| Max Digital Resolution | 2560x1600 | 2560x1600 | 2560x1600 |
| Max VGA Resolution | 2048x1536 | 2048x1536 | 2048x1536 |
| Audio Input for HDMI | SPDIF | SPDIF | Internal |
| Reference board height | 4.376 inches | 4.376 inches | 4.376 inches |
| Reference board length | 9 inches | 7 inches | 6.6 inches |
| Max Board Power | 96W | 59W | 69W |
| Supplementary Power Connectors | 6-Pin PCIe | No | No |
| Minimum Power Supply Required | 400W | 300W | 300W |
| |||
Notes
As you can see, the GeForce GT 240 features more shaders than the 9600 GT – 96 vs 64 – but is armed with slower clocks. This negates the advantage of the extra shaders somewhat. You’ll also see that the GT 240 features fewer ROP units. With fewer ROPs and slower speeds, peak pixel rate is down to just 4.4Gpixels/sec, less than half that of the 9600 GT. Texture fill rate is also down thanks to the slower clock speeds.
Finally, thanks to the narrower 128-bit memory interface, memory bandwidth is also lower than that of the 9600 GT.
Based on all this, the GeForce GT 240 is going to have its work cut out for it if it intends to keep up with the 9600 GT.
Fortunately DirectX 10.1 can be used by the GT 240 to enhance performance in games that take advantage of it. One effect that’s being used by increasingly more games is high definition ambient occlusion (HDAO). In games that use HDAO like STALKER Clear Sky/Call of Pripyat, and HAWX, the GT 240 could potentially use DX10.1 to gain an advantage over the 9600 GT. We’ll see how this pans out in the benchmarks.
[image]
| <% print_image("04"); %> | <% print_image("05"); %> |
One important figure that’s not mentioned in the table above is price. Street prices for the GeForce GT 240 start at $90 for the slower DDR3 model, and hit $100 and up for cards equipped with 512MB of GDDR5. This is quite a bit more than the 9600 GT, which starts at $80 before mail-in rebate. So the 9600 GT costs less and features better performance specs than the GT 240. The GeForce GT 240 also lacks SLI support.
This is a feature NVIDIA typically omits on their entry-level VGA offerings like the GeForce 8400 GS and 9400 series. Previous mainstream offerings like the 8600 GT, 9500 GT, and of course the GeForce 9600 series have all had SLI support, so this is a bit of a departure for NVIDIA.
NVIDIA says that the typical GeForce GT 240 user doesn’t plan to use SLI, and while that’s certainly true, we still think it’s a glaring oversight considering the card it replaces has always supported the feature.
[image]
| <% print_image("06"); %> | <% print_image("07"); %> |
| Retail GT 240 cards | Page:: ( 3 / 15 ) |
[image]
| <% print_image("08"); %> |
To test the mettle of NVIDIA’s GeForce GT 240 GPU, we received two retail graphics cards based on the chip that are shipping today. From EVGA we have their GeForce GT 240 Superclocked, while MSI sent over their N240GT MD512 OC/D5.
Both cards are factory OC’ed to speeds beyond NVIDIA’s GT 240 baseline specifications, and both are equipped with 512MB of GDDR5 memory, which is the most popular memory size so far.
While 1GB of memory is pretty much considered baseline amongst performance and high-end cards nowadays, 512MB is enough for the cards in this class as they don’t have enough GPU horsepower to run really max graphics settings in most cutting edge games. (Nor do they have the memory bandwidth to run these titles with 4xAA.)
We’ll start by discussing EVGA’s GT240 Superclocked first.
[image]
| <% print_image("09"); %> | <% print_image("10"); %> |
EVGA GeForce GT 240 Superclocked
With its black PCB and single slot cooler, EVGA’s GeForce GT 240 Superclocked looks decidedly better than the GT 240 reference design and cooling.
Surprisingly enough, the board is actually a little longer than the GT 240 reference board, measuring 7.5” in length. That’s nearly an inch longer than the stock board. Why is EVGA’s GT 240 card longer than reference? We’re not exactly sure. It’s possible that EVGA wanted more space for their cooler, which is a single-slot design with an all-aluminum heatsink and small fan.
While the heatsink is pretty large, we should note that it doesn’t directly cool the board’s memory modules or any of the VRM circuitry. The fan runs quietly, quieter than any card we tested today in fact, and generally the entire cooling system does a good job of battling heat. We observed temps as high as 61 degrees Celsius under load.
[image]
| <% print_image("11"); %> | <% print_image("12"); %> |
EVGA clocks the Superclocked board at 550MHz core/1340MHz stream processors/1800MHz memory. These speeds are the same as NVIDIA’s reference specifications for the graphics core and stream processors, while the memory speed is improved by 100MHz.
Finally, we should note that none of EVGA’s GT 240 cards support EVGA’s lifetime warranty program. Instead the warranty period is limited to two years. As always, EVGA reserves the lifetime warranty program for their higher-end GeForce cards, beginning with the GeForce GTX 260 GPU. You can easily determine which EVGA cards support the lifetime warranty by looking up the part number (p/n). Cards with a lifetime warranty have a P/N ending with the letters “AR”.
MSI N240GT MD512 OC/D5
Considering that the GeForce GT 240 isn’t designed to shatter any world records in 3DMark or any other performance benchmark, you’d think MSI would lay off a little on a low-end card like the GeForce GT 240. If you thought that however, you’d be very mistaken.
MSI’s N240GT MD512 OC/D5 isn’t a copy/paste reference design. MSI has come up with their own design that’s built for tweaking.
[image]
| <% print_image("13"); %> | <% print_image("14"); %> |
Like their higher-end GeForce and Radeon boards, MSI integrates their “military-class components” on the N240GT MD512 OC/D5. This includes solid state chokes and highly conductive capacitors. These board-level components supply the GPU with very clean, stable power and can tolerate higher power loads. This lengthens the life of the board, and gives you more headroom for OC’ing. Another added benefit of the solid state chokes in particular is low noise. Sometimes when cards are pressed under load you’ll hear a buzzing noise emanating from the card. Solid state chokes don’t generate this noise. MSI says they use golden SSCs which are designed to handle higher power flows on their GT 240 board, enhancing its capabilities even further.
[image]
| <% print_image("15"); %> | <% print_image("16"); %> |
With this in mind, MSI offers voltage adjustment on the N240GT MD512 OC/D5. The board ships with MSI’s Afterburner software, which can be used to up the GPU voltage, or OC the board’s graphics core, stream processors, and memory.
Even though it’s the newest tool on the block, Afterburner has quickly become our favorite utility for GPU OC’ing that’s provided by a graphics board manufacturer. MSI worked directly with RivaTuner’s Unwinder on the Afterburner interface, and it shows. It’s both easy to use and powerful at the same time.
[image]
| <% print_image("17"); %> |
From one single page Afterburner offers both clock speed and GPU voltage adjustment, as well as temp monitoring and fan speed adjustment. You can also save up to five custom profiles, which is useful in case you want to setup profiles for different games (for example, this is helpful in case you can get the board to boot at really high speeds for Modern Warfare 2, but need to back it down a little for Crysis). The only feature missing for the GT 240 that stands out is memory voltage adjustment. Check out the screenshots to see Afterburner in action.
As the “OC” in the board’s official designation implies, the card we received for review is an OC Edition that is overclocked from the factory for improved performance. More specifically in terms of clock speeds, MSI sticks with the stock frequencies of 550MHz core/1340MHz shaders, with only the board’s memory OC’ed to 1800MHz, an improvement of 100MHz over stock. Apparently instead of OC’ing the board to outrageous levels, MSI instead leaves most of the work up to the end user, who is free to push the board as far as you wish using Afterburner.
Physically MSI’s N240GT MD512 OC/D5 measures a little shorter than the EVGA GT 240 with the board length just a hair under 7”. For enhanced cooling performance, MSI integrates a larger dual-slot heatsink/fan unit, so you will have to leave the expansion slot directly adjacent to your PCI Express graphics slot empty if you plan to use this card. We also noticed that the board’s fan generated a little more noise than many of the other cards we tested (it wasn’t the loudest though) although it’s by no means unbearable, and as you’ll see in the temp testing, MSI’s implementation delivered substantially improved cooling performance as a result. Load temps were orders of magnitude better than the other cards tested.
Software and accessories
You’re looking at a very minimal bundle with both boards. In fact we almost didn’t write this section. Obviously there’s no game bundle offered with either card, and with each board offering DVI, HDMI, and VGA natively on the back plate, no DVI-to-VGA adapters or other hardware accessories are bundled either. Instead each ships with their respective driver CD and manual, that’s it.
| System Setup | Page:: ( 4 / 15 ) |
AMD Phenom II X4 965 Black Edition
Gigabyte MA790FXT-UD5P
6GB (2x3GB) OCZ Reaper HPC 1600 @ DDR-1066 Speeds
NVIDIA GeForce 9600 GT 512MB
NVIDIA GeForce 9800 GT 512MB
EVGA GeForce GT 240 Superclocked 512MB
MSI N240GT MD512 OC/D5
ForceWare 195.55
ATI Radeon HD 4670 512MB
ATI Radeon 4770 512MB
Catalyst 9.11
2TB Seagate Barracuda XT
Microsoft Windows 7 Ultimate 64-bit
Benchmarks
Call of Duty: Modern Warfare 2
Crysis
Left 4 Dead 2
Far Cry 2
STALKER Call of Pripyat Benchmark
Batman: Arkham Asylum
Resident Evil 5
Wolfenstein
Tom Clancy’s HAWX
| STALKER Call of Pripyat | Page:: ( 5 / 15 ) |
STALKER Call of Pripyat DirectX 10/10.1
| Modern Warfare 2 | Page:: ( 6 / 15 ) |
CoD: MW2 – DirectX 9
| Crysis | Page:: ( 7 / 15 ) |
Crysis – DirectX 10
| Far Cry 2 | Page:: ( 8 / 15 ) |
Far Cry 2 – DirectX 10
| Left 4 Dead 2 | Page:: ( 9 / 15 ) |
Left 4 Dead 2 – DirectX9
| Resident Evil 5 | Page:: ( 10 / 15 ) |
Resident Evil – DirectX 10
| Batman: Arkham Asylum | Page:: ( 11 / 15 ) |
Batman:AA – DirectX 9
| HAWX | Page:: ( 12 / 15 ) |
HAWX – DirectX 10
| Wolfenstein | Page:: ( 13 / 15 ) |
Wolfenstein – OpenGL
| Power, Temps and Overclocking | Page:: ( 14 / 15 ) |
Power Consumption
Temperatures
Noise
Overclocked Performance
| Conclusion | Page:: ( 15 / 15 ) |
It certainly looks like NVIDIA’s engineers tried to deliver a decent 9600 GT successor. The shader count looks right, and while the number of ROPs is definitely on the low side (NVIDIA shouldn’t have compromised here in our opinion), we like NVIDIA’s integration of newer technologies like GDDR5 and DirectX 10.1. The problem is that the GeForce GT 240’s stock clock speeds aren’t high enough.
The core clock needs to be at least 100-150MHz higher than where they are now, and the stream processors need to be running at least 1650-1750MHz. Given the low core clock and shader speeds, NVIDIA got the GT 240’s memory clock speed right (this card isn’t starving for more memory bandwidth), but we still think a 9600 GT successor like the GT 240 should at least deliver the same 57GB/sec of peak bandwidth as the original.
And these figures are being conservative.
So why didn’t NVIDIA bump up the clocks more? Based on our OC’ing results, it looks like they definitely couldn’t do it. Even with our dual-slot MSI GT 240 board which was also overvolted as far as the software would allow, we couldn’t hit these kind of clock speeds.
We’d wager that TSMC and their 40-nm manufacturing process is probably partially at fault, NVIDIA wouldn’t have been able to remain within their power budget at higher speeds.
It’s likely that NVIDIA’s design also played a role as well. Hopefully NVIDIA’s learned a lot from this experience at 40-nm, and as a result has been able to integrate some of the lessons learned into Fermi in time. It certainly would explain why this chip and Fermi are both late to market. GT 240 was supposed to debut last Spring, and of course NVIDIA was hoping to have Fermi ready in time for Q4 of this year.
While it may not outperform its predecessor in 3D gaming, thanks to its lower power consumption, compact dimensions, and native support for uncompressed 7.1 LPCM HD audio, the GT 240 is great for HTPC use. In our opinion this is what NVIDIA really should be touting the most with this chip. It can also be used as a dedicated PhysX accelerator, although due to its high price keep in mind that there are cheaper options out there if this is what you’re looking for.
It’s the high price tag that really perplexes us the most about the GeForce GT 240. Officially NVIDIA says that the card sits between the GT 220 and the 9800 GT in their lineup, but it’s priced closer to a 9800 GT than the GT 220. In fact, you can find multiple 9800 GT cards for sale on sites like Newegg priced cheaper than the GT 240, especially when you factor in mail-in rebates. Given its 3D performance, the GT 240 needs to sell for at least $25 less than NVIDIA’s current asking price. That would bring the GT 240 GDDR5 down to about $75 -- $5 less than the 9600 GT.
Here we can only guess that the yields are so low that NVIDIA doesn’t want to do this just yet.
Of the cards represented here in this article, it’s MSI’s N240GT MD512 OC/D5 that will probably appeal the most to the crowd who craves performance. With its dual-slot cooler and MSI’s Afterburner software, the board runs cooler and has the potential to be OC’ed further thanks to Afterburner’s support for GPU voltage adjustment.
If you want something a little quieter and single-slot, the EVGA GeForce GT 240 is a fine solution. The board’s smaller profile and low-noise fan make it perfect for that HTPC enthusiast looking to build or upgrade an existing computer. For these same reasons it would also be the more ideal dedicated card for PhysX.
Hopefully NVIDIA can address our concerns about the GT 240’s price in the very near future though. As everyone knows, ATI’s equivalent part, codenamed Redwood, is just around the corner. Redwood will potentially offer Radeon 4770-class performance with the addition of DirectX 11. Considering that the GT 240 can’t keep up with the 4770 today, NVIDIA could be in trouble once those cards arrive in Q1 of next year. Current rumors suggest Redwood could arrive sometime in January, so that doesn’t leave NVIDIA a whole lot of time to get this issue fixed…