Author: Kevin Spiess
Editor: Howard Ha
Publish Date: Wednesday, February 24th, 2010
Originally Published on Neoseeker (http://www.neoseeker.com)
Article Link: http://www.neoseeker.com/Articles/Hardware/Reviews/hd5830launch/
Copyright Neo Era Media, Inc. - please do not redistribute or use for commercial purposes.
The HD 5000 series just continues to roll out unabated, following mostly in the footsteps made by the HD 4000 series: just this time, with more regularility, and less competition. The intervening time since the launch of the HD 5870 six months ago, until now, Nvidia has pushed back the release of their next-generation Fermi GPU further along the calendar, while ATI has been releasing new models monthly; like clockwork. With today's HD 5830 we have a fully fleshed out line, from the entry-level HD 5450, to the dual-GPU beast HD 5970, and just about everywhere in between. Possibly the only absent brother in the HD 5000 family, after today, may be the expected HD 5890, which certainly isn't late, as it (will probably) be a refresh product.
So, where abouts does the HD 5830 fit in? The simpliest answer would be right between the HD 5770 and the HD 5830. Last generation, the HD 5830 offered great value, and in fact, the MSI R4830 was one of this reviewer's personal favorites looked at last year. The HD 4830 could generally be overclocked to the point where it exceed stock HD 4850 performance, thus, offered great value for an inexpensive price tag. This generation's HD 5830 is a bit different, relatively speaking, so we'll have to see if it offers the same whiz-bang today.
The HD 5830 will be selling for a recommended price of $239 USD. This seems like it will actually be the price you'll first see this card at -- unlike, sometimes, with other suggested retail prices that don't match up with expected targets. To put this in perspective, currently HD 5770's are going for about $160-$180, while the HD 5850 is rarely less than $300 USD. As you can see, the HD 5830 will be selling for a price closer to the HD 5770 -- so the performance turns out to be instead closer to the HD 5850, then we might have a great deal on our hands. But conclusions will have to await of course, for our benchmarks.
In a bit of a change of pace for ATI, the HD 5850 will be launching with custom designs out of the gates, from board partners today. Usually, when a new model comes out, initially all the board partners sell basically the same thing: a card physically identical to the reference board design, made by ATI. This isn't the case here -- board partners have had HD 5830s for a while it seems, so have had time to leave their own marks on the product. Below is a gallery of these diffferent variations. For our review, we will be looking at engineering sample HD 5830 from ATI that has the familiar HD 5770 and HD 5xxx series cooler on it.
Below we have offerings from our friends Gigabyte, XFX (who switched from Nvidia to ATI), PowerColor and Sapphire. As you can see these aren't just rebadged cards, they are significantly different constructions of the HD 5830.
On the next page we'll take a look at our HD 5830 from ATI.
Once again: please keep in mind that this is an engineering sample. It does not look like this actual design of the HD 5830 will be sold in stores. However, the same Cypress LE GPU is running the show, so the video card here should offer the same performance as ones available to purchase come tomorrow.
As for our HD 5830, it bears the striking family resemblance common to most of the high-end HD 5000 cards of this generation. We have the customary black and red cooler, with the air ducts at the end jokingly compared to something on Batman's old Batmobile. Actually, from looks alone, this HD 5830 is virtually indistinguishable from a reference design HD 5870 - same length, same black back-plate; same all over.
However, looks don't tell the whole story of course. Underneath that cover we have shorter heat fins, making for a lighter video card than the HD 5870. Which either means the cooling requirements are lower for the HD 5830 (which is questionable, as the clock speeds are fairly high) or that we will see a higher operating temperature -- at least on this engineering sample -- compared to the rest of the HD 5800 series, which has already has a fairly high operating temperature at load, for the reference board designs.
For display output options, again we have the same selection found on other HD 5000 cards: with two DVIs, one DisplayPort, and one HDMI. This varied selection gives you a lot of leeway with what displays you will be able to use for now, and in the future. With the standard Eyefinity support as well, you can run up to three simulatenous displays with the HD 5830, which is a plus. Three displays might seem like overkill to many people, but if you had a chance to game on three displays, or have three displays just for general applications, you might be surprised to see how quickly you may change your mind on the matter.
Powering the HD 5830, we have the Cypress LE.
When making cutting-edge processors, you might be surprised to hear that not all of them are created equally. Vaguely comparable to a bell-curve marking different human traits, you'll get some chips that just, from the luck of the draw and their position on the silicon wafer, turn out to be more capable than others. The top-chips, the grade 'A' materials, they go to the highest-end cards. Meanwhile some Cypress chips may be pretty much good to go, but have a flaw or two -- instead of tossing these chips, a company will shut down the defective bits, in order to salvage the GPU. These get used in parts such as the HD 5830.
The specs of the Cypress LE are interesting. Take a look-see here:
As you can see, the HD 5830 has just about as much in common with the HD 5770 as it does with the HD 5870. For the stream processors, we have 1140 on the HD 5830. This is 320 SPs beyond the HD 5770, and 320 SPs short of the HD 5850. The core clock speed is 50MHz less than the HD 5770; 75MHz more than the HD 5850. The 1GB of GDDR5 is clocked the same as the HD 5850 (lower than the HD 5770). As for the power, the HD 5830 saps more than does the HD 5850 or HD 5770...We don't know about you, but we are interested in seeing how this plays out in the benchmarks: closer to a HD 5770's performance, or a HD 5850s?
Beyond the GPU, you have the other HD 5000 series features, such as the aforementioned Eyefinity, and DirectX 11 support. Nvidia's oft-delayed Fermi is coming out at the end of next month, but we expect only the high-end chips, retailing for at least $500 to be offered first, which means, for many folks, that ATI is your only option to go for if you want DirectX 11 support at the moment. With the HD 5000 series you also get the UVD engine for HD video playback, CrossFireX support (allowing you to run the HD 5830 in tandem with any HD 3000 series or above video card), and PCI Express 2.0.
One area where Nvidia does have an edge on ATI is with their CUDA technology. CUDA allows for programs to be run utilizing the massive parallel nature of the GPU. In this respect, ATI is trying to play catch up of sorts: they have ATI Stream, which is a similar tool, enabling code to be run from the GPU, instead of your CPU. To encourage use of ATI Stream, there will be a contest announced over the next few weeks, called the ATI Stream Developer Contest. Prizes will be offered to developers who submit winning proposals outlining potential projects to be undertaken utilizing ATI Stream, DirectCompute (a GPU programming option from DirectX 11), and OpenCL. The five categories of competitions will be: games, video content creation, image processing, consumer graphics, and porting CUDA applications. Our statement that ATI is playing catch-up to CUDA in this respect, can probably be justified alone by that last contest category mentioned.
Here is how the HD 5830 compares to variety of recent generation video cards on the market today:
|GTX 285||GTS 250||9800 GT||GT 240||HD 4850||HD 5750||HD 5770||HD 5870||HD 5850||HD 5830|
Memory Clock (effective)
|512 bit||256 bit||256 bit||128 bit||256 bit||128 bit||128 bit||256 bit||256 bit||256 bit|
|512MB GDDR3||512MB GDDR5||512MB GDDR3||512MB GDDR5||512MB GDDR4||1024MB GDDR5||1024MB GDDR5||1024MB GDDR5|
A quick note here: we did not have much luck OC'ing the HD 5830. The HD 5830 relentlessly crashed, with only slim overclocks. However, this might have been due to the early BIOS we were using. Additionally this particular card will not be available for sale, so, for felt that for our real overclock adventures we'll have to wait to get a HD 5830 from a board partner in order to see what can be seen.
We expect that with the high power requirements (compared to the HD 5770 and HD 5850), the HD 5830 is already running fairly close to top speed. We do not expect the HD 5830 will OC as well as the HD 4830. We guess the OC limit will be around 8% for the core clock. Certainly it will not overclock as well as the HD 5850 often does.
Alright -- lets get to some benchmarking.
Video cards used in the benchmarks include a XFX HD 5870, a BFG GTX 285, an Asus ENGTX 275, a MSI R4890 Cyclone, LeadTek GTX 260 Extreme+, a Powercolor HD 5770 PCS+, a VisionTek HD 4850, and a Gigabyte HD 5670.
For the drivers, all the ATI cards used the Catalyst 10.1 drivers, except for the HD 5830, which used beta drivers, while the Nvidia cards used Forceware 195.62 drivers.
We have updated our benchmarks. Here they are:
Bioshock: For this benchmark, all of the Detail settings were set to 'High'. All of the graphic option switches were set to 'On', with the exception of the following three settings: Vsync, Windowed mode, and Force Global Lighting. We used FRAPS to measure frame rate performance. The FRAPS run was 138 seconds, triggered from pulling the switch in the sub at game's beginning. The sub's dive involves many big models moving around, which should strain the GPUs and be a good measure of the game's engine.
Batman: Arkham Asylum: Gotham's Greatest Detective makes for a good benchmark. We used the in-game bench, running at 2560x1600, with the highest quality settings possible. We chose to only test without AA as there has been some controversy that AA is unnecessarily handicapped in this game for some video cards.
FTL_Blunderbuss: This is a demoscene demo by the group Fairlight, which came in second in competition in October 2009. It makes very heavy use of particles, and is a good GPU workout. We used FRAPS to measure the average framerates of a run through the program, running at 1680x1050, with 4xAA, and 'high' detail.
Crysis: Warhead: Games don't get much more demanding than Crysis. We used the 'Gamer' pre-set level of details, which is the middle level setting out of 5 options. We ran the benchmark on the 'avalanche' map, using the FrameBuffer Crysis benchmarking tool, version 0.29, in DX10 mode.
Enemy Territory: Quake Wars: We use this id FPS benchmark to test out higher resolutions. We used the highest possible detail settings. We tested the resolutions at 4x AA as well as at 8x AA. 16x AF was also used.
Far Cry 2: This open-world FPS is great looking game that really puts the strain on a gaming rig. We used the built-in benchmarking tool, and the overall 'Very High' quality setting was used.
Furmark: This intensive, synthetic benchmark models a ring of fur. We benched at 1680x1050.
Street Fighter IV: You have probably heard of this famous fighting game. It has 3D graphics, but generally does not require much GPU horsepower to run well. We used Capcom's stand-alone PC benchmarking tool for our tests, and ran everything at its highest possible settings, using 4xAA, and the 'Watercolor' setting.
Unreal Tournament 3: We tested the game using a fly-through of the vehicle capture-the-flag map 'Suspense.'ShangriLa (map) running for 90 seconds. Details were set to 'High', and a AF setting of 16x was used.
World In Conflict: We used the built-in benchmark of the demo version of this game. We ran the benchmark in DX9 rendering mode, with a 'High' level of quality. For the AA testing, we used a setting of 4x, and a setting of 16x for AF.
If you would like any further information about our benchmark settings, feel free to ask us in the forums.
The HD 5830 starts stuff off with a huge bang here -- Vantage scores the HD 5830 within close reach of the top single-GPU NVIDIA card currently offered, the GTX 285. If it turns out that the game performance is something like this, well, what can you say: things don't look great for NVIDIA's chances of swaying over gamers.
In the feature tests, the HD 5830 puts in a good numbers, coming in third often. Interesting stuff.
In this big ring of fur simulation, the HD 5830 does good get numbers over the HD 5770, though with AA on, is a decent step back from the HD 5850 Toxic (which is a moderately factory overclocked model). The HD 4890 -- which you can still find on sale if you look around, selling for about $200 or so -- also beats out the HD 5830 with AA on. Nonetheless these are fine looking numbers.
The NVIDIA cards just can't compete in this benchmark, with AA on. Let's see what else we see.
Here the HD 5830 shines: in this particle heavy test, it comes well ahead of the HD 5770, and beats out the GTX 285 fairly well too. Is the HD 5830 going to really makes things difficult for non-Fermi DirectX 10 GPUs? Let's see what happens in the games.
This is too early to say for now, but we'll say it anyways: if this performance keeps up, the HD 5830 is a straight-up GTX 285 killer. This -- at least for this reviewer -- is quite unexpected. But so far, when you consider the performance between the $235 HD 5830, and the $350 - $400 GTX 285, it just plain gets murdered. No arguments will be entertained on this point!
So far, the HD 5830 is getting enough headway over the HD 5770 to make the price worthwhile. The HD 4890's still offer very good value, but with Eyefinity and DirectX 11 support, no doubt most will want to stick with the new tech: the HD 5000 series.
Well, H.A.W.X is one game anyways were the GTX 285 rains supreme. The 1140 stream processors don't seem to be enough here to get breakthrough performance.
Maybe it was too early to call the HD 5830 a GTX 285 killer after all? Here NVIDIA puts in a better show, with their GTX 275, which goes for around the same price as the HD 5830, puts in a better performance in the match-up.
In this particular game, the HD 5830 scores much closer to the HD 5770 than the HD 5850.
So far these benchmarks have been fascinating.
Again the GTX 275 pulls head -- not to mention the GTX 285 which is off to the races -- though the HD 5830 certainly isn't showing is poor performance by a long shot. It seems it really depends on the game whether the HD 5830 is fighting against the HD 5770 or HD 5850.
With this well used engine, the HD 5830 perhaps puts in its weakest performance yet, relative to the rest of the line-up. Though we use 'weak' loosely -- considering it is still delivering over 100 fps at 2560x1600, there isn't all that much that is weak about it.
Here the HD 5830 performs on about par with the GTX 260.
It seems like for the very intense games, with AA on, those shader processors get a touch overwhelmed. When they are not overwhelmed, the high speed memory is able to work with the high clock core speed to deliver exceptional framerates. This might not bode well for the undoubtedly demanding DirectX 11 titles due in 2010 and beyond.
Here the HD 5830 sits comfortably in about the middle of the line-up: on par with NVIDIA's similarly priced GTX 275.
It is a bit of a surprise to see the HD 5830 beat out the HD 4890 at top res' here. Unfortunately for ATI though, the GTX 260 gets a hair more performance here, in Bioshock.
With the sequel to this game out now, we suppose this will be one of our final Bioshock benchmarks. So, see ya later Bioshock, thanks for all the giant drills and stuff.
This is one of the better of the HD 5830's performances, coming in at a very solid fourth here in this very demanding game.
For these results, we used a Kill-o-Watt measuring device in conjunction with the program OCCT to measure 'load' temperatures once 200 seconds elapsed.
These numbers are all that significant here, to be honest, because the cooler on our HD 5830 is assured to be different than one you will have, if you choose to get a HD 5830. However, we can say though, that judging from the numbers, and the heatsink employed on our engineering sample, the HD 5830 has just about average temps' -- it should behave similarly to the HD 5850. The extra board wattage and high clocks do not seem to lead the HD 5830 to overly excessive load operating temperatures. In the 80's is toasty, but nothing to be too concerned about.
However the XFX HD 5830, as shown on page one of this review, from appearances, looks like it might get a bit hot (the cooler seems less effective than what is on our model). But of course we can't say this for certain without having tested the video card ourselves.
For these results, we used a Kill-o-Watt measuring device in conjunction with the program OCCT to measure 'load' temperatures once 200 seconds elapsed.
The HD 5770 is the real king of the performance-per-Watt, but the HD 5830 is no slouch. It saps more power relative to its performance than does the HD 5850, but still, at 40nm the power demands are fine, coming in well below the NVIDIA option of the GTX 275.
In the first half of these benchmarks, the HD 5830 was really knocking off our socks -- seeing it beat out the GTX 285 a few times was most impressive. By the end of the benchmarks though, things balanced out: like much of the HD 5000 series, the HD 5830 performs admirably for its price point, but does not deliver any new, explosive surprises.
Surprsingly, the GTX 275 still matches up well to the HD 5830, on performance and price metrics. However, with DirectX 11 support, and Eyefinity, and the flexibility of CrossFireX over SLI (with CrossFireX you can mix-and-match cards, with SLI it has to be the same card, limiting your options) we feel that the HD 5830 is the clear winner in this contest; unless, perhaps, you particularly appreciate NVIDIA's CUDA.
The HD 5830 fits in well at about $235. It neither kills the HD 5770, nor destroys the HD 5850, when you consider the price points of all three cards. The HD 5830 just fits in nicely with the HD 5000 family, like the middle-aged kid in a typical sitcom. ATI has just about every possible big base covered now, from $50 - $700. The company should be commended for delivering such a complete line-up, from bottom to top. NVIDIA really has their work ahead of them to challenge this dominance with their workhorse GPUs, namely the G92. The G92 can still trot, but hey, it came out in October of 2007; the glue factory beckons.
That being said, we have a feeling the HD 5830 card just might be the one many gamers have been waiting for. If $200 is your upper limit for what you are willing to pay for a video card, then you might just want to go up to $235 -- for this price you are getting a very capable video card, with significant horsepower. The HD 5830 should suit many folks just right.
Please do not redistribute or use this article in whole, or in part, for commercial purposes.