GTX275 SuperOC versus GTX285 OC
AS THE ATI HD5870 series takes the graphics performance crown while we wait for the Nvidia GT300 to finally come out, the previous throne holders, the Nvidia GTX285 and its junior siblings from the GT200 series, are taking their final ride into the sunset. However, their performance, coupled with the number of Nvidia-optimised and endorsed games - yes, some including the famed PhysX on the GPU - still attracts quite a few gamers.
Here I take a look at two fast and memory-rich cards in this class, both from Gigabyte. One is the previous reference card used in several of my earlier reviews - the GTX285 OC 2GB, running the GPU at a default factory overclocked 660MHz and providing a whopping 2GB of local video memory. The other is the latest offering, the GTX275 SuperOC with 1,792MB of local video memory and running the GPU at a very high 715 MHz as default, with corresponding shader and memory speed increases as well to 1,550MHz shader clock and GDDR3-2520 memory.
Gigabyte redesigned the board by including its famed 2oz double copper as well as better Japanese solid capacitors, on top of revamping other componentry. As a result, the faster factory overclocked speed seems to be as stable as the normal default operation speed.
Knowing that the GTX275 has a reduced 448-bit memory bus and the resulting 14 per cent memory bandwidth and capacity drop versus the 512-bit memory bus on the GTX285, yet keeping the same shader number as the bigger brother, what would be the result in a typical benchmark like the 3Dmark Vantage, at the resolution and effect settings expected to be used by high end gamers getting such cards?
Here is the comparison in the Extreme settings of 3Dmark Vantage of both cards. Take a look at the difference:
GTX285 Extreme:
GTX275 Extreme:
As you can see, the heavily overclocked GTX275 SuperOC wins this time, although the differential is quite small.
Personally, I still prefer the more 'common sense' capacity and bus width numbers on the GTX285 OC. After all, the simple numbers like 2GB capacity and 512-bit bus sound more comfortable than odd numbers like 1,792MB and 448-bit bus, among others, don't they?
Another point to note is that, since PhysX on the GPU is enabled here, the higher shader clock on the GTX275 SuperOC skews the CPU score in 3Dmark Vantage as well. Check the scores again, there is a nearly 6 per cent difference. Whether this would have any actual impact in games is hard to say, as physics simulation still doesn't impact the gaming experience as much as it really should.
In summary, if you're an Nvidia fan, either of these cards will do a good job. The speed is there, and the large memory can enable the most effects, greatest detail or longest view distance settings in nearly any game. If using computation on the GPU, well the extra memory on these cards can host more data locally without going to and from the main system memory over a 10-times slower PCIe link. Although, in that case, I believe the GTX285 will still be a better deal due to more real memory bandwidth and fewer potential overclocking-related stability issues in computation. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
Didn't the GT200 series die already?
Can we please chop it's head off and move on?
I would much rather see an article on hardware-disk encryption, electrical noise from fan motor ratings, or even if they could take methyl mercury from the vaccine and use it as a cooling liquid for CPUs. But no more GT200 yesterdays sausage turned into meatloaf please.
Actually covering all the IT products that contain mercury would be a real news article if you guys are up to it.
As a PC gamer who has been putting together my own rigs for years I have used both sides with good performance for the time. I currently have a stock XFX GTX 285 on a Gigabyte MOBO supported by a Q9650. The GTX 285 is sweet, quiet and fast as hell.
I do not see any games bringing this rig down for a long time, especially since they have moved to creating games on the kiddie consoles first then port them to the PC, sometimes with horrible results.
Dragon Age, from Bioware looks promising and look forward to it. Developed on the PC first.
over video cards, Since its all about the consoles these days. Why even bother anymore. i have a 9800gtx+ and have zero urge to upgrade any part of this rig. like regulus said they build them first for the kiddie consoles then port them over mostly in a half assed state. So big whoop. pretty graphics that work 25% of the time. I dont even bother wasting my money on the cool new game, these days i wait til everyone else blows 60 bucks on it first this way i dont waste my money on a half assed port that doesnt work and might be fixed in the next 20 or so patches.
An they wonder why people pirate?
well gee lets see.
Only in it for the buck. Half ass programing, 50/50 shot it might actually work coming out of the package.
Piss poor support. hmm. i have no idea why people would do such a thing.
@ Regulas and James.
You guys are just both idiots. First of Regulas you card is EOL and will be not be running any of the NEW titles in 2010 that support DX 11 so basically either UPGRADE your Video Card or just keep enjoying YESTERDAYS titles.
As for James well he is simply brain dead and I'm not sure there is any hope for him since this idiot apprently knows more about GAME programming then the people that make the games.
Listen bub, your 9800 is grabage and should be burried in the back yard , better yet donate it to a charity. Now go out and get a NEW Card or Buy Regulas's card off him and at least get with the program. No wonder your crying about not enjoying gaming. As for Consoles....they SUCK A## .
Now Lets MOVE ONWARD to 2010 shall we ?
jesus, I hate people that always complain but yet do nothing to move forward.
Advances in 3D technology over the last few years have been increasing compute-demands on the GPU rather than increasing memory bandwidth needs. nVidia cards are over-engineered solutions that provide more memory bandwidth than is beneficial, and they do this using slow GDDR3 memory by making a wide, expensive memory bus. In contrast, AMD uses a cheaper 256-bit memory bus, but with faster GDDR5 memory, while optimizing their GPU for high compute throughput.
This test here basically shows that a card with a faster GPU and less memory bandwidth beats a card with a slower GPU and more memory bandwidth, which is pretty obvious since memory bandwidth isn't really a performance limitation on high-end cards.
The majority of PC games are still done in straight DX 9 and how long has DX 10 been out? If you think the inclusion of DX 11 in Windows 7 will make some miracle happen and all new games will be in DX 11 keep smoking on your pipe. If any do have DX 11 support they will still support DX 9 and 10 because unlike you most of us don't run out and buy a new video card every time MS makes a DX move then call people idiots on blogs.
I hate kiddie console gaming too, the kiddie machines are ruining good PC gaming along with the Games for Windows Live (Steam wannabe) abomination. The kiddies console ports (80 % to 90% of new PC games) will not overkill my rig until maybe the next generation of kiddie consols. The makers of Crysis already said their new game engine is for the consols first.
Like it or not the games industry is now driven by the consoles and as such PC hardware might move on but the general technology level employed by most developers is based on console engines and framework.
These days PC gaming is resigned to taking console games and running them at a higher resolution, with more fltering on top. This does indeed imbue them with a nice edge over the consoles versions, but really does not provide a test on modern PC hardware like so so many PC specific games of yore.
In short there are actually such a small percentage of PC games that require a monstrous machine to maximise the settings fewer and fewer PC gamers see an incentive to upgrade every time a new generation comes along. Nvidia's policy much mocked of renaming cards is a direct result of the reduced demand for next generation hardware.
If you look at the big names games out for PC next year, stuff like splinter cell, mass effect 2, RAGE etc etc you know for a fact you are not going to require uber hardware to get impressively high settings. All because they are designed for console first, and then ported onto PC with a mere lick of paint.
There must be only a handful of titles out for PC in the next 12 months that will make a GTX275 sweat @ 1920 x 1200. is it really worth another 300 pounds for a new video card to play maybe half a dozen new games at better settings? think about it. This is partly why PC gaming is dying.
Until co-processed enhanced physics are supported the way graphics is, I can imagine that only a small number of games will truly bank the game on just one GPU.
Nathen: "... you card is EOL and will be not be running any of the NEW titles in 2010 that support DX 11..."
If a computer can run D3D11 software depends just about only on whether the DirX11 runtime is installed or not.
Provided it is; which requires Win7, Win Vista SP2 or Win Server 2008 SP2; just about any graphics card will function properly and support the software.
The only difference is that if the hardware isn't fully D3D compliant it won't look quite as pretty nor will it run quite as fast.