Test Setup And Benchmarks
We decided to test the Radeon HD 5450 against its primary competition in the $40 to $50 price segment. This includes its predecessors, the Radeon HD 4550 and 4650, the GeForce 210 and DDR2-based GeForce 9500 GT.
First we must note that the Radeon HD 5450 sample we received was gifted with a 900 MHz memory clock, 100 MHz faster than the reference specification. Since the beta driver necessary for use with this card didn't support Overdrive, and our overclocking tools weren't able to properly work with the card, we were stuck with the overclock. It's unfortunate that this happened and an AMD representative even sent us an apology and an offer to retest with a new card, but they didn't send the offer quickly enough to reasonably give us enough time to make our deadline. As a result we're testing the overclocked card. On the bright side, 100 MHz over a pathetic 64-bit memory bus probably isn't going to do squat, it's roughly equivalent to a 25 MHz overclock on a 256-bit bus. But we thought you should know, regardless.
We tested the Radeons 4550 and 4650 with AMD's newest Catalyst 10.1 driver, and both of these cards came equipped with the reference clock speeds. No need to tweak anything here.
Getting the GeForce cards to do what we wanted proved to be a little more complicated as both were factory overclocked models, and the newest 196.21 driver broke all of our overclocking tools -- Rivatuner, MSI afterburner, and even Gigabyte's Gamer HUD lite were rendered useless on our test system. In addition, the GeForce 9500 GT test sample we had was a DDR3 model that is priced too high to be appropriate for the comparison.
Our solution for the 9500 GT, a factory overclocked Gigabyte GV-N95TD3-512I, was to use the older 195.62 driver to enable overclocking tools and then underclock the core to 550 MHz and the memory to 450 MHz. This should give us results comparable to an off-the-shelf DDR2 model of GeForce 9500 GT. We allowed the 50 MHz memory speed advantage vs. the reference DDR2 model in an attempt to negate the notable latency disadvantage that the DDR3 memory would suffer from.
As for the GeForce G 210, a Gigabyte GV-N210OC-512I, we left the clock speeds at the factory setting and used the newest 196.21 driver. Why did we do this? Mostly because we couldn't underclock it and didn't realize until later that the driver was to blame. We didn't have time to re-test, but the results remain valid for two reasons: firstly, the overclocked Gigabyte card can be purchased for $42 online -- notably below the Radeon HD 5450's $50 price tag. Far more importantly however is the fact that the GeForce G 210 is so incredibly slow as a gaming card that the overclocked result is rendered irrelevant.
While the Radeon HD 5450 is admittedly intended for low-impact gaming scenarios, we will put it through a gamut of game tests at lower settings to see if the card is suitable for low-detail use in more demanding titles.
| Graphic Test System | ||||
|---|---|---|---|---|
| CPU | Intel Core i7-920 (Nehalem), 2.67 GHz, QPI-4200, 8MB Shared L3 Cache | |||
| Motherboard | ASRock X58 Supercomputer | |||
| Networking | Onboard Realtek Gigabit LAN controller | |||
| Memory | Kingston PC3-10700 | |||
| Graphics | ATI Radeon HD 5450 Sapphire Radeon HD 4650 Diamond Radeon HD 4550 Gigabyte GeForce 9500 GT 720 MHz Core, 1,566 MHz Shaders, 800 MHz Memory, 1GB DDR3 | |||
| Hard Drive | Western Digital Caviar WD50 00AAJS-00YFA | |||
| Power | Thermaltake Toughpower 1,200W | |||
| Software and Drivers | ||||
| Operating System | Microsoft Windows Vista Ultimate 64-bit 6.0.6001, SP1 | |||
| DirectX version | DirectX 10 | |||
| Graphics Drivers | AMD Catalyst 10.1, Nvidia GeForce 195.62 (GeForce 9500 GT), 196.21 (GeForce G210) | |||
| Benchmark Configuration | |
|---|---|
3D Games | |
| Crysis | Patch 1.2.1, DirectX 9, 64-bit executable, benchmark tool |
| Far Cry 2 | Patch 1.02, in-game benchmark |
| DiRT 2 | Version 1.0.0, Custom THG Benchmark Run 1: Medium Settings, No AA, DirectX 9 Run 2: Medium Settings, No AA, DirectX 11 |
| World In Conflict | Patch 1009, DirectX 9, timedemo |
| Tom Clancy's H.A.W.X. | Patch 1.02, DirectX 10 & 10.1, in-game benchmark |
| Left 4 Dead | Version 1.0.1.5., Custom THG Benchmark |
Synthetic Benchmarks and Settings | |
| 3DMark Vantage | Version: 1.02, PhysX Off, 3DMark scores |
What a terrible opening paragraph! You slated the 5000 series because it wasn't as big an advancement as the 400 series!
1280x1024 is 5x4
1024x768 is 4x3
1280x960 is 4x3
Sexy heat sink albeit at the cost of losing an expansion slot.
As much as you don't like it, I think this is a fantastic card. Perfect for HTPC users right down to the fact it's half height (I believe I mentioned half height cards when these were first announced). Very attractive price point especially given savings on a sound card. Full marks to AMD for this one.
Hmm..how long before this card is effectively replaced by the next IGP? Got mixed feelings about it, to be honest..it's neither good nor bad, it just exists. Tough call buying this one.
The lower end of the 5000 series is a bit disappointing. I'm not convinced the 5450 is worth it for an HTPC given it's high price (>£50) in the UK - especially since a 785G motherboards is around the same price. I was holding out for the 5570, but it must sit between the 5450 and 5670 in price (~£60?) with performance somwhere between the two. The 4650 at
Ouch, Ouch, poor lil' blighter didn't like those benches much. I'll forgive it if the UK price is loooooooow. I do love the passive cooling!
However, aspect ratio of the screen isn't the pixel ratio, it's physical width and height. For example, if I run 1280x1024 on my monitor then it is 16:9 (and horribly stretched).
Not what the article implied though...
So what we've got here is a DirectX 11 card which doesn't have enough power to play DirectX games... If this was a Nvidia card you would have ripped into it no end, I really think Toms is getting ATI Bias; maybe theres some form of agreement here we don't know about.