Review: The tiny giants—Nvidia GT430 vs. AMD Radeon HD5550
Posted by By Robert Dow and Jon Peddie on October 13th 2010 | Discuss
Categories:
Hardware Review
Tags:
Or should we call them mighty midgets?
In late April AMD introduced the DirectX 11 Radeon HD 5550 entry-level Value Segment graphics AIB. It was well received and praised for its value, and performance. Introduced at $70 with 1GB of DDR3 and fanless, it represented a lot AIB for not too much money. Nvidia had their DirectX 10 GT220 to go up against it, and despite the hoopla of DirectX11 and a GB of memory on the AMD AIB, the older Nvidia AIB still managed to sell well.
Well better late than never, and now Nvidia has released their fanless DirectX 11 GB AIB, the GT430, at the MSRP of $70 and it’s serious but not overwhelming competition for AMD.
These AIBs are really terrific, and offer an amazing amount of features and capability.
Nvidia’s GTS has three outputs: VGA, HDMI and DVI, and they all work together. Nick Stam of Nvidia proved this by hooking up an Alienware 2310, which is a 19x10 120Hz monitor with HDMI, along with the Dell 30” DVI.
At first it didn’t come up in the Nvidia control panel because you have to activate the HDMI port on the monitor with its on-screen controls. Then it came up fine as a second monitor with both DVI and HDMI active. Stam tested both “extend the desktop” and “clone mode” and both worked.
He also made the HDMI the “primary monitor” and that worked just fine. Then he tried HDMI with VGA, and it worked fine, also switching which one was primary, etc.
While you can physically connect all three monitors to the board, and see all three show up in the Nvidia control panel, only two are active, and if you try to activate a third monitor, you get a warning that says “This GPU supports up to 2 displays.
The boards are similar in size, with the Nvidia GTS 430 a bit shorter. In both cases the VGA connector could be removed and the back plate replaced for a half-height configuration.
We tested the AIBs on a 3.2 GHz Phenom II, with 4 GB RAM and Windows 7, 64-bit. Performance-wise the two boards were close on the three tests we ran, as illustrated in the following charts.
Unigine Heaven
Unigine can be run in DirectX 10 or DirectX 11. For our testing we ran it in DirectX 11. The Unigine engine demo turns out to be a really wonderful benchmarking tool. It runs the most advanced rendering features supported by the graphics hardware. All modern features are supported by the Unigine renderer:
- Per-pixel dynamic lighting
- Normal & parallax occlusion mapping.
- 64-bit HDR rendering.
- Volumetric fog and light.
- Powerful particle systems: fire, smoke, explosions.
- Extensible set of shaders (GLSL / HLSL).
- Post-processing: depth of field, refraction, glow, blurring, color correction and much more.
Besides that, it uses and exercises:
- Shader Model 5.0 support.
- Hardware tessellation.
- SSAO (screen-space ambient occlusion).
- Unique materials system with parameter hierarchy support.
- Vegetation system.
- Live water with a surf zone and caustics.
It also offers native support of stereo 3D devices (although we didn’t do any benchmarking in stereo). If Futuremark doesn’t get their act together and deliver 3Dmark DX 11 soon, they are going to be forgotten since Unigine is offering such a great test.
“Resident Evil 5”
“Resident Evil 5” only uses DirectX 10. We chose that game because it has pretty good graphics, a built-in benchmark, and most games right now are still DirectX 10. As a survival horror game, you can see the cheesy arcade game play designed for a console in it. As such it’s an OK benchmark for a PC, but it’s not a very nice or fun game, and its racial sexist overtones are pretty disgusting.
“Lost Planet”
“Lost Planet” is a three-year old DirectX 11 game that really stresses the graphics AIB. The game has a built in benchmark with a choice of A or B, A is DirectX 10, and B is DirectX 11.The primary purpose of Test B is to push the PC to its limits and to benchmark the maximum performance of the PC. It utilizes many functions of Direct X11 resulting in a very performance-orientated, very demanding benchmark mode. As the chart shows the game is not even playable with an entry level AIB in DirectX 11.
We summed all the scores to come up with an overall performance rating.
We then calculated the Pmark using $70 as the ASP for the AIBs.
Who buys this stuff?
As you know, the desktop PC market is segmented into two major categories: Consumer, and Commercial or Enterprise. Each one of those categories is segmented into two more categories: Integrated graphics, and discrete AIB graphics. And, as you’ve no doubt have heard, integrated graphics accounts for 70% of the desktop sales. But, more of the integrated graphics based PCs go to the commercial segment than the consumer segment. So that means it’s the consumer who is buying these low cost AIBs.
And, as mentioned, these $70 entry-level boards are half height and designed to lose the VGA connector if anyone wants to use them in a media center box. That’s a further indication that OEM and ODM AIB builders have targeted these low cost but powerful AIBs right at the consumer. Remember, the Fermi and Evergreen chips have pretty good video engines built into them. So if an OEM builds a media box with one of these in it, the experience is going to be fine. Also, these board’s GPUs have world-class scalers in them so they can accommodate almost any screen size., and they’ve got enough frame buffer memory and PCIe bandwidth that the video will be smooth, bright, full color, and never stutter.
There’s also the issue of geographic customers. For example, it turns out in the U.S. consumers who buy low cost PCs don’t think about adding graphics board or anything else for that matter so a low cost AIB is strictly an OEM deal AMD or Nvidia or their partners. In Europe, however, consumers are not only willing, but actually anxious to make upgrades. And in Asia, particularly in China, a “PC” is thought of as having a graphics board, and it’s even a cachet for a user to have such a thing. In Japan where houses are a bit more constrained, the PC has to serve several needs, and so a graphics board is a requirement there too. The net result is that although we are seeing some fantastic developments in EPGs and HPUs, there will still be a good market for entry-level AIBs in the consumer market.
What do we think?
For this test sequence AMD is the winner. Obviously other tests could be run (or some omitted) that might give different results, but we doubt there would be very much difference in the performance scores, and so when the Pmark is calculated, AMD would win because of its lower power usage.
From the consumer’s point for view, no one is a loser here, these AIBs are great value for the money.
Overall impression
These boards are great value for money. Their scores, physical size, price, power consumption are almost identical.
We summed all the scores (even the lousy ones from “Lost Planet”) to arrive at the performance numbers which are shown in Table 3 to use in the Pmark calculation shown in Table 4.
AMD gets a better Pmark score and that is because they beat Nvidia not only on power consumption, but also on performance.
But a consumer wouldn’t do badly with either board. So we give them both a thumbs up, just that AMD’s thumb is a bit higher.
| GTS 430 | HD 5550 | |
|---|---|---|
| Core Clock | 1.4 GHz | 650 GHz |
| Memory Clock | 900 MHz | 900 MHz |
| Memory | 1024 MB DD3 | 1024 MB DD3 |
| Driver | 260.6 | 10.9 |
| Power | 49 W | 39W |
| core | 96 CUDA cores | 320 stream processors |
| texture units | 16 | 16 |
| arch | 40nm | 40nm |
| Transistors | 585M | 627M |
| Video outputs | DVI-DL | DVI-DL |
| HDMI 1.4 | HDMI 1.3 | |
| VGA | VGA | |
| S3D | S3D |
| GTS 430 | HD 5550 | |
|---|---|---|
| Sum of scores | 232.4 | 242.3 |
| GTS 430 | HD 5550 | |
|---|---|---|
| Pmark | 0.067755102 | 0.088754579 |
> > Testing AMD’s Radeon HD6870, HD6850, and Nvidia’s GTX 460 review
< < The Tests - Quadro 4000, Quadro 5000, FirePro V9800 and FirePro V7800 review
