Cyanide is found in the kernel of a peach
|
|
|
Nvidia versus ATI at the high end
AT LONG LAST Nvidia's long awaited next generation high end DX11 parts have surfaced. The Fermi-based GeForce GTX480 - this one is still a bit rare, though - and GTX470 are out to compete against the incumbent single-GPU performance leaders, the AMD ATI Radeon HD5870 and HD5850. You've seen the avalanche of benchmarks all around the web, and most of them show these latest Nvidia powered graphics cards taking the lead, although not by much usually.
The wider memory buses - 384 bits wide on the GTX480 and 320 bits on the GTX470, compared to 256 bits on the ATI parts - help in attaining higher memory performance, as well as more bandwidth for the computational tasks in which Nvidia's GPGPU chips are expected to excel.
Now really, how important are these results in this current round of benchmark leapfrogging? Let's see.
First, ATI is expected to have sped up parts on both single and dual GPU fronts. Let's tentatively call them the HD5890 and HD5990, although the final names may differ. Count on anything from a 10 per cent average speedup on the single GPU part, resulting mostly from the 850MHz to 950MHz GPU clock jump, to more than 15 per cent on the dual GPU card, thanks to resolving power and cooling issues there.
Second, many vendors like Asus, Gigabyte, Sapphire and XFX might offer pre-overclocked cards with the existing HD5870 and HD5970 that attain these clock speeds by default anyway. Couple that with Eyefinity-enabled versions supporting 2GB video memory per GPU for that extra oomph in games and benchmarks and the updated cards should even out the battle with Nvidia on the single-GPU front.
On the dual-GPU front, a possible GeForce GTX490 consisting of two GTX470-class GPUs joined together, is still far away, probably awaiting another tuned, lower power GPU stepping. In the meantime, the HD5970 and its expected higher-clocked successor would rule the roost among the dual GPU cards.
Talking about another GPU stepping from Nvidia, there's another reason to look for a useful update there by Computex time two months from now. Wafer yields didn't allow Nvidia to enable all 512 shader cores on the GTX480, forcing it to limit the GTX480 to only 480 shaders. A yield or stepping improvement might allow Nvidia to release an 'updated' card, call it GTX485 for instance, having all 512 cores turned on. However, whether or not that will be possible for Nvidia is still a very big 'if' at the moment. Such a card could also have a 3GB large memory option.
In the meantime, if there are useful yields of chips with all 512 cores, you'll be most likely to see them in the most expensive OpenGL professional 3-D cards for workstations and visualisation clusters. A Quadro FX5900 6GB card comes to mind, expected sometime in April.
So, in summary, the performance match between Nvidia and ATI at the high end will see another round later this spring, with - we can hope - an interesting and competitive speed-wise benchmark battle between the two graphics rivals' updated chippery. If that happens it should help push prices down a bit, too. Therefore the current bunch of comparisons you see will most likely be very short lived. That won't dissuade the hard core enthusiasts from getting their own 'newest and fastest' card first, but at the very least, I'd personally hold off for the updated GPUs, come about May or so.
Another competitive aspect that's often overlooked is that, looking back at the CUDA and OpenCL experience and the associated programming difficulties, Nvidia made a major step forward with these Fermi GPU parts. You can program them directly in C or Fortran, rather than having to use obscure special approaches. I'm not sure yet how efficient all of that will be, but, for now, being able to tap nearly a TeraFLOPS of double-precision IEEE standard - that is, usable for most applications - floating-point power in a much easier way can mean a lot, such that many more PC programs will be able to take advantage of the new performance resource. Now, even Excel spreadsheets can easily be GPU-accelerated.
On the other hand, Nvidia could have kept a 512-bit wide memory bus for this purpose, simply to allow for more onboard GPU memory, up to 8GB on Quadro or Tesla and 4GB on the GeForce cards. The problem with GPU computing is that it can be very very fast, as long as the code and data sit in the local GPU RAM. Once it goes across the PCIe lanes to access the main memory, the performance penalty can be over an order of magnitude, negating the need for GPU acceleration. The previous GeForce and Quadro generation had this advantage due to its wide buses. This problem will remain until the GPUs link directly to CPUs via QPI or HyperTransport rather than over slow PCIe links with high latency.
At that point, assuming a compatible memory mapping and management structure with the system CPU, the GPU can, through direct inline C++ and Fortran coding, become a co-processor to the CPUs, just like the old x87 from Intel was to the x86 before floating-point was integrated into every processor. ATI, as part of AMD, will have an advantage there as it will have AMD's Hypertransport access by default, but the problem then will be that it will limit ATI to a small market share.
Nevertheless, we will be bringing you our own tests of the new graphics cards as they come along, hopefully including the customised higher-performance versions of the GTX480 and GTX470. Gainward and EVGA are two vendors to look at here in the Nvidia space, besides the usual Asus and Gigabyte. At the very least, the graphics card competition has turned interesting again. µ
Share this:
Share
Both Nvidia and AMD produce great cards. When it comes to professional and industrial designing then Nvidia Quadro ftw. When it comes to speed without spending much then AMD wins.
For gaming in 2009/2010 AMD's Radeon 5970 was untouchable. Normally Nvidia has the fastest top card but AMD took the industry crown with their dual GPU 5970.
Nvidia brought their 580 gtx to the market at the end of 2010. It was a little late due to probs with design and chips yields but it put them in line with AMD in the top card slot.
AMD just went thru their largest graphics card architecture change in years. They threw out their old shader architecture and rewrote new and more optimized shaders from scratch. The new cards with the new architecture starts with the 69xx series cards. These card are new in 2011 and are already showing up on benching sites and on gaming sites.
When it comes to raw gaming power these days Nvidia and AMD are closely matched. If you're having difficulty in choosing a card consider what features you're looking for.
*3D*
Both newer cards from Nvidia and AMD now support 3D. Nvidia's 3D Vision is purchased as an extra kit ($199) and sadly is a closed architecture. You don't get to choose how or with what you view your 3D. You need to buy an Nvidia 3D Vision approved monitor or TV that works with their kit.
AMD's 3D solution on the other hand is an open solution so you aren't restricted. You can choose to use your own active or passive shutter glasses, your own 3D tv, blue/red glasses with a normal monitor, or even an Nvidia compatible TV or for the hell of it, 2 monitors and toilet paper rolls if you want to. But if you go with Nvidia's 3D Vision you won't be disappointed with the quality.
*Multiscreen gaming*
For multiscreen gaming AMD has Eyefinity with their cards. The Eyefinity 6 screen card lets you play games using up to 6 screens at the same time for surround-wall-screen gaming without buying any extra cards.
For NVidia they have multiscreen gaming but you need to buy a graphics card for every 2 screens you use, so for 6 screens you need to buy 3 cards. Thats too expensive for the average user because it requires expensive motherboards to accommodate all the graphics cards.
When it gets down to it, if money is no object and you can meet any psu/cooling demands then grab whichever card is at the top, regardless of which brand it is. If money is an issue then buy whichever card will give you the best bang for buck. If you want to do multiscreen gaming then go for an AMD Radeon with Eyefinity (all new Radeon cards have it). If you want to go 3D then go nVidia if you have a compatible TV or Monitor or have the money to buy one. If you want to go 3D later but buy now then go AMD that way you can choose your style later.
Look at the Tom's H/W benches with AA turned on
- the GTX480 wins just about every bench
- usually by a good margin.
- and let's face it, someone who buys one of these cards is going to be playing at 1920x1200 with 4xAA at least...
I recently purchased a 5870 upgrading from an Nvidia Geforce 285. I didn't think I would miss it but having to turn off the hardware accelerated physics ( Nvidias PhysX) in a few key titles is pretty disappointing. It's great being able to run my games with all the eye candy set to max but because I know how the game used to play with the hardware physics enabled I kinda miss my pre-upgrade Nvidia card....
The GTX 480 has had its double precision legs cut off. It's limited to 1/8 of the single precision rate, not the 1/2 that the chip is capable of.
This means that peak theoretical FP throughput is 1345 GFLOPs in single precision, and just 168 GFLOPs in double precision. In comparison, the 5870 has 2720 GFLOPs single precision and 544 GFLOPs double precision. While it has difficulty putting all the single precision power to the tarmac, it's very good at reaching peak double precision rates - I'm easily hitting 400+ GFLOPs with even "first attempt" shaders. That's over double the theoretical maximum that the GTX 480 is capable of.
NVidia really need to un-gimp the double precision on the 400 series. I was seriously considering switching my main GPGPU dev platform back over to NV, but there's no point focussing development on something that's so far behind in performance.
I'm torn between the HD5870 and the GTX480. Currently a user of high-end DX10 cards in SLI, I care less about FPS than features. If any of you had 2 computers in front of him, same high-end setup but system has an HD5870 and the other has a GTX480.. can you tell any difference in any of the benchmarked games just by the naked eye? But what if it was the Unigine Heaven heavy-tessellation demo? what if it's a physx-supported game? And god knows what other features in fermi game developers are going to use. Nvidia are known for their influence on game developers. HD5870 is more of near-perfect card than GTX480, as well GTX480 is like this this unpredictable crazy fella. Soon the DX11 3DMark will be released, If GTX480/GTX470 perform much better in this test then we might as well gain this performance in real games.
The possible driver bug I mentioned is in single player only. Multi-player is fine. so far...
I think they were only comparing Price/heat/power in gaming. I will say Kyle and crew at HardOCP tend to be a lot less biased in their reviews. Also, their testing methodology seems a lot more professional, well thought out and vendor neutral than some other sites.
It is interesting to compare real benchmarks and tests to the marketing hyperbole spouted by either (or any) company. If you had only listened to Nvidia's press releases, the 480/470 was smacking the 5000 series around and taking it's lunch money. ;)
All in all the offering's from both companies are competitive enough to drive down prices and is overall a good thing for us, the consumers.
As a side note, I think I found a driver bug in the 10.3 Catalyst drivers in Battlefield Bad Company 2. When I die and start to respawn, the game crashes to desktop every time. I haven't pinpointed it yet, but if I do, I'll submit it to ATI.
develop a few generations of fermi cards down the line, nab a console contract, further expand fermi and will you have something juicy?
"no reason" is a bit overdoing it by them, nvidia has a different focus and if you want/need that or bet on tessellation or CUDA or C compliance because you are some sort of developer, or the 3D thing, then there is a need.
But thanks for naming a (or another) site as example, and I might add to my previous comment that I noticed in the past that very often the reviews of 0-day sites rate graphicscards much better than all sites do a month or more later when suddenly the difference are much less or figures are reversed, those NDA's and testsamples (with testdrivers) do seem to skew stuff it seems.
Actually, HardOCP's benchmarks had the 480 winning half the tests and the 5870 the other half. In both cases, not by much. Their conclusion was the 480 was not worth the added cost, heat and power draw.
With the 470 it was worse for nvidia. The 470 benched the same speed or lower in all tests vs the 5850. HardOCP said there was "no reason to buy the 470 at all", given it's similar performance/price and higher power/heat.
@Muhammad Imran/mi1400
I have to question how the hell toms manages to be the only site on the planet where the 480 is slower in benchmarks than an hd5870, seems damn fishy to me.
Although it is true most sites signed NDAs and did their testing some time ago, but seeing raw specs it just makes no sense to not have the 480 be a few percent faster than the hd5870 in all benchmarks.
Not that I'd pick a 480, it uses too much power and gets too noisy/hot for my taste, but the ATI stuff is also either too expensive (5870 and higher) or not on sale anywhere (5850).
Plus do you really want to pay all that money to nividia/ATI and then have months of anger at buggy drivers that crash your shit?
If it wasn't for those factors I might have picked up an 5850, but for now they are both out of luck.
Slow PCIE lanes eh, 16 lanes of v2 (same as 32lanes of v1) you call slow?
Seriously, if you use it for excel it will have to be fed back to the program and be processed by the CPU anyway, and the CPU will be many many many times slower, so to call the PCIE the bottleneck in that scenario seems a bit out there, drawing the results will be slower than blitting the damn data between the card and main memory, and you have to wonder what kind of data you plan to put in excel that gets close or even exceeds a gigabyte, tracking the cost of electrons in the universe?
most articles show 5870 (1x GPU) loosing on hands of GTX480 (1x GPU) becuae of i believe using default drivers. Now THG has put an eye opener.
GTX480 Vs 5870 and rest
http://www.tomshardware.co.uk/review_print.php?p1=31846
Astonishing to see ATI 5870 still crush most of 480 benchs. Also to note ATI at default clock. 5870 can do 1+GHz where 480 is already saturated at TDP. ATI 5880 slot is empty wink! wink!, a minor architecture refresh can home here!
Too BAD Article Came Late in Day, AS So many have Opinion. Nvidia Is running about same number of Transistors As BEST CPU At Three times core Frequency, Giving 3 Times heat. Once Nvidia Gets Thru to PURE State of NT6.1 engineering, Route is Down in nm. Nvidia Has Time, for Now.
Ati Is Doing INtel Shuffle, Going Faster, Faster FASter than Speed of Life, market for 6 Display setup is limited. Ltd. Yet, Raw Speed can match simple test scores, Theres NO Beating Good Engineering. Nvidia has Design that Can go Places on Inside. Maybe Start into Chipset Biz Again.
Ati is In Nice Puddle of Way, COST Being Power Factor, Very Powerful Factor, In Fast Changing World. Yet Ati hasn't Crown, Only Numbers. Fast Numbers.
Double Precision IS Best. Making Crossfire System, well single precision Isn't that BAD. 2 5750 Will Take 'er Over TOP. Just arn't Mains to Handle Power Displayed by both Ati & Nvidia. theDOG Needs Grooming.
By hiting 7.9 Consistantly in Experience, Know vista hardware needs have been met. As First Stated by Ultimat Persona of drashek Corp USA, 64 X enviorment IS Needed, well Pci-e 2.0, Great help. Someone Call Pci-e Group & Tell Um, Get pci-e 3,0 Ready by End of Summer For: Chocolate Cutters to Start Cutting, While Its' ALL 40 nm. In Favorite Chocolate Time of ALL Halloween. Cutter Are Cuttin Ham Bone right Now.
Then By Grace of GOD, NT 6, & Server Potential Will Be realized. 2008 Server fact Home. By Extra Grace of Mother Goddess, Then 2 pci-e 32x Slots Can Cover Ground, taking bunch of Ringee' Stuff out of Main & putting Power To Peddlers.
Signed:Baker Space Astronault & Supreme Commanding Monkee.
If Nvidia manages to get a 512 core version of Fermi going, they'll still call it the GTX480. This way if it has a lower power usage or other defining feature, all the makers stuck with the old ones will still be able to sell them to the unknowing users.
See the GTX260, 3 different versions with the same model number. Maybe I'm being a bit harsh on Nvidia here, but I've been burned by their products one too many times.