Sign in with
Sign up | Sign in

Nvidia GeForce GTS 450: Hello GF106, Farewell G92

Nvidia GeForce GTS 450: Hello GF106, Farewell G92
By

After dutifully serving the mainstream gaming community for three years, Nvidia's G92 is finally being played out. Meet GF106, the little engine behind GeForce GTS 450. Is this 192-core part still potent, or did Nvidia cut too much from G92's replacement?

There’s something about a game of catch-up that inspires urgency. After struggling to bring its Fermi architecture to market six months after AMD launched its own DirectX 11-class Radeon HD 5000-series, Nvidia’s derivative GPUs are now hitting one after the other, in rapid succession.

I’ll admit—I was worried when the company launched its GeForce GTX 465, based on the massive GF100 with nearly half of the chip disabled. Was it having problems with derivative designs?

Then Nvidia introduced the GeForce GTX 460, based on the GF104 graphics processor. That card impressed me. It was, as our respective review mentioned, what GeForce GTX 465 should have been. Then we tested two GeForce GTX 460s in SLI and gave that configuration another Recommended Buy award—not necessarily for what those two boards did against the competition from AMD, but for the fact that they trounced a GeForce GTX 480 at a comparable price. Suddenly, we were seeing solid value from Nvidia.

Now, two months later, the company has another derivative part ready to rock. Its GF106 is, in essence, one half of a complete GF104. But remember that the GeForce GTX 460 is really seven-eighths of a GF104. So, the GeForce GTS 450 is in all actuality a bit more than one-half of a GTX 460.

Nvidia’s First Unaltered Fermi GPU

Remember back to the GeForce GTX 480 launch. The company’s flagship was (and still is) based on a cut-back version of the GF100 GPU. This was a consequence of not getting the yields needed to launch with a 512-shader part. The GeForce GTX 460 was (and still is) based on a cut-back version of the GF104 GPU. This was to avoid showing up the GeForce GTX 470 with too much performance.

Today, the GeForce GTS 450 represents Nvidia’s first Fermi-based card based on an unaltered graphics processor, GF106.

GF106 is composed of one Graphics Processing Cluster (GPC), with four Streaming Multiprocessors (SMs). Each SM wields 48 shader cores. These are the same “modified” SMs first seen on GF104, equipped with four dispatch units and eight texture units (instead of two and four, respectively). In total, that gives us 192 shader cores, four Polymorph engines (one per SM), and 32 texture units.

The chip’s back-end is similarly cut down to a pair of ROP partitions, each capable of eight 32-bit integer pixels per clock. Nvidia refers to that arrangement as 16 ROPs. Naturally, each partition is associated with a 64-bit memory path, yielding 128-bits aggregate this time around. That’s half of what the GeForce GTX 460 1 GB has available to it, so we expect to see a substantial performance hit with anti-aliasing enabled.

Display all 25 comments.
This thread is closed for comments
  • 1 Hide
    Nintendork , 13 September 2010 14:47
    The card is an epic failue. Even consume way more then HD5770.

    The way it's meant to be botched.
  • 1 Hide
    Nintendork , 13 September 2010 14:51
    Seem really bottlenecked by the slow as hell GDDR5 3600 (4600/4800 on HD5700)
  • 2 Hide
    darksai , 13 September 2010 18:30
    aaaaand.... NI is almost here.
    We know it wont be the usual type of generational performance increase, but the fact is amd are about to start another round of top to bottom, and nvidia have only just got their lowest card that can be considered for gaming. If amd have focused on nvidias strong points (multi gpu scaling, aa, tesselation, drivers) then they could wipe the floor even harder than this year.
  • 3 Hide
    Silmarunya , 13 September 2010 22:08
    darksaiaaaaand.... NI is almost here.We know it wont be the usual type of generational performance increase, but the fact is amd are about to start another round of top to bottom, and nvidia have only just got their lowest card that can be considered for gaming. If amd have focused on nvidias strong points (multi gpu scaling, aa, tesselation, drivers) then they could wipe the floor even harder than this year.


    I doubt that. If AMD would focus on adding tessalation they'd have to increase die size and for extra AA performance extra memory would be welcome. If they'd do that, AMD would have to deal with Fermi's issues - low yields, high power consumption, heat,...

    They could drastically improve drivers, but if they were able to do that, they could have already released such drivers by now. Multi-GPU scaling requires drastic design and driver redesign, not something I'm expecting from a refresh.

    AMD should stick to what a refresh is all about: make minor improvements, improve power optimization, up clocks a bit and so on. You can't give Evergreen all the strengths of Fermi without adding at least some of its weaknesses. AMD should play to its strengths, while Nvidia does what they do best.

    Of course, who knows what a truly new generation like Southern Islands or the GTX 500-series will bring... Nvidia's impressive multi-gpu scaling, CPU/GPU offloading, aa and tessalation performance combined with AMD's simple dies with high yields, excellent power/performance ratio and very modest thermal footprints, wouldn't that be heaven?
  • -2 Hide
    jamie_macdonald , 13 September 2010 22:48
    Wouldn't get one of these budget fermi's but they seem like they could do good for someone with tight pockets ..
    ...As for the Power hungry comments fro mthe ATI fanbois above ...i think they need to realise that more power = more power (i don't want to save power, but utilise it) ;) 
  • 4 Hide
    Silmarunya , 13 September 2010 22:58
    jamie_macdonaldWouldn't get one of these budget fermi's but they seem like they could do good for someone with tight pockets .....As for the Power hungry comments fro mthe ATI fanbois above ...i think they need to realise that more power = more power (i don't want to save power, but utilise it)


    1) Power costs a lot of money. In America it's dirt cheap, but in Europe and especially in developing countries it isn't (and low prices aren't a good thing btw, as they don't encourage people to reduce their power consumption)

    2) Performance per watt is a good indicator of how well a chip is designed.

    3) Fossil fuels are too precious to waste on unused transistors, better spend them on things you actually make use of.

    4) Nvidia doesn't manage to translate that additional power consumed into performance. For every 1% increase in power consumption compared to a competitively priced ATI card, you don't get a 1% increase in performance.
  • -3 Hide
    jamie_macdonald , 13 September 2010 23:02
    I can tell you do not own one ;)  ...them stats are complete baloney ...And i am in europe ...The 480 i have uses less wattage than my old 8800GT SLi setup and performs about 4x better ... if that's not an efficiency improvment what is?!

    I have one of the 5870's in a PC here at work ... you know the card thatys supposedly on par with the 480 .. and well ..it sucks as soon as you ramp up the detail and AA. nuff said.
  • 0 Hide
    irish_adam , 14 September 2010 00:12
    jamie_macdonaldI can tell you do not own one ...them stats are complete baloney ...And i am in europe ...The 480 i have uses less wattage than my old 8800GT SLi setup and performs about 4x better ... if that's not an efficiency improvment what is?!I have one of the 5870's in a PC here at work ... you know the card thatys supposedly on par with the 480 .. and well ..it sucks as soon as you ramp up the detail and AA. nuff said.


    oh your so right, i always thought every online benchmark was a lie too!

    and obviously as they shrink the die with each generation the power consumption will go down you idiot! We are talking about the fact that fermi uses more power than ATI's 5000's but that extra power being used is not translated into more frame rates
  • 4 Hide
    Silmarunya , 14 September 2010 02:38
    jamie_macdonaldI can tell you do not own one ...them stats are complete baloney ...And i am in europe ...The 480 i have uses less wattage than my old 8800GT SLi setup and performs about 4x better ... if that's not an efficiency improvment what is?!I have one of the 5870's in a PC here at work ... you know the card thatys supposedly on par with the 480 .. and well ..it sucks as soon as you ramp up the detail and AA. nuff said.


    Aha, we found an Nvidia fanboi!

    1) I'd seriously hope your 480's use less wattage than your 8800's. Die shrinkages tend to have that effect. It's the Evergreen line you need to compare it with.

    2) Yup, all online benchmarks are complete myths. In other news, the average technophile reports finding your findings far more credible than those of just about any reliable online source.

    3) I'm curious to hear what game doesn't run smoothly at highest settings with a 5870? Crysis maybe, but that's about it. Unless you're one of these people claiming to be able to do something physiologically impossible, like seeing the difference between 85 and 90 fps...
  • 0 Hide
    darksai , 14 September 2010 14:35
    ..on a 60hz lcd too :p  also, 4x or 8x aa is more than enough at 1080p imo. You only have so many pixels to soften up with.
  • -3 Hide
    jamie_macdonald , 14 September 2010 16:33
    LOL!! ...get called an idiot for seeing improvments, and get called an idiot seeing difference between laggy choppy play @ 16x AA 1920x1080 on an ATI 5870 and smooth play at 32x AA on the gtx480 ...is real life experience that hard to understand ..

    ...and yes your'e correct i couldnt give hoot about power consumption ... you not get that? :p 

    PS: Not a Fanboi .. but recognise when something is plainly better ... and by that i mean to play on ... not by the stats others feed me on benchmarks! ...you lot make me laugh :) 
  • 0 Hide
    m_ragab712 , 14 September 2010 20:28
    that's mean that all benchmarks not true??
  • -2 Hide
    jamie_macdonald , 14 September 2010 20:36
    The benchmarks need redoing or something .. i am sure they wouldnt lie eh ...i guess maybe driver versions or cases can make a difference ...as with everything there are some small elements that can effect ..nothings written in stone and all manner of hardware bases can stirr things up a little

    ...not here to call anyone a liar, just sayign what my real life experience is like with these devices as i own both to compare. I get no major heat and noise from the nvidia like all the reviews say! its quiter than my old cards and cooler too. something is wrong somewhere eh?

    My case is a massive dell xps one htough .. i guess mine has more air than some .. i suppose that could effect results but seriously anyone considering either brand of modern cards without decent ventilation has to be nuts right?
  • 0 Hide
    irish_adam , 14 September 2010 20:54
    O.o

    your trying to say your case affects frame rates? out of all the components you could have picked to explain the bottleneck with the ATI card you picked the case? WTF?!?!
  • -2 Hide
    jamie_macdonald , 14 September 2010 23:03
    irish_adamO.oyour trying to say your case affects frame rates? out of all the components you could have picked to explain the bottleneck with the ATI card you picked the case? WTF?!?!



    Only due to a lack of anything else .. you would think the same system as a testbed would be fair for both cards ... but according to the anti nvidia crowd its unfair on the ATI ... also WTF? .. you lot are on drugs or sumin >.>
  • 0 Hide
    darksai , 14 September 2010 23:13
    @jamie
    1. even a 5770 can handle most (by which i mean not just the 10 most demanding games which benchmarks use) games at 1080p with 4x aa, so the conditions at which a 5870 is "choppy" (which is generally accepted to be
  • -1 Hide
    jamie_macdonald , 14 September 2010 23:36
    darksai@jamie1. even a 5770 can handle most (by which i mean not just the 10 most demanding games which benchmarks use) games at 1080p with 4x aa, so the conditions at which a 5870 is "choppy" (which is generally accepted to be


    Guessing their must be a problem with mine then ...

    The only reason i got one is cause after their recent success's i might test the water again with ATI ...i had two of their older cards some years ago and both blew up on me :/ 

    I guess their not ready then... or im extremely unlucky with ATI branded cards? ...either way .. arghh! lol (they're quite expensive things .. wouldnt expect too much fail)

    Also ty for not attacking me like a bulldozer darksai ...seems the fashion on tech sites .. its gone fanboi mad ... i'm jsut stating experiences yknow .. i wish them both (brands) well as its good for all of us. :) 
  • 0 Hide
    darksai , 14 September 2010 23:59
    ..and you say that about maybe they have different case and whatever, but you yourself are comparing a 480 at home with a 5870 at work..
    Does your work pc have as good a cpu? Is trimmed of excess software/bloatware? How about overclocked?
    With review sites, we know for sure that everything besides the gpu's are exactly the same when they compare, so i'm pretty sure their evidence, and by extension, conclusions, is far closer to the truth than yours is.
  • 0 Hide
    williehmmm , 15 September 2010 00:54
    I'm running a GTX 470 sli system (for exceptional FPS under 3D vision), which with the Fermi cards is something like 70% of non 3D vision performance, unlike my old 8800GT sli which gave the usually expected 50% of non 3D vision performance.

    Of all the reviews I read, AA turned on took a much bigger performance hit on ATI 5870/5850 compared to GTX480/470. And minimum frame rates appeared to always be much lower on ATI too.

    The high end single GPU ATI cards are generally cheaper than Nvidia counterparts and are what I recommend for people buidling a single card system, but if you are considering multi GPU, I'd always go for Nvidia. The added value of 3D vision is what swung me, but the 90%+ peformance increase with Sli, compared to the 50% - 60% increase with crossfire makes the hands down winner, in my opinion.
  • 0 Hide
    gamesmachine , 15 September 2010 04:47
    I love my 8800GT SLI setup I do, when the 8800GT came out it was amazing...huge performance, great price and single slot design.

    I really like the 460 but I do wish they (ATI, nVidia) would make cards that actually can render games above 10fps at decent settings but only utilise a single slot of my PC. I have a TV Card, Sound Card and a USB Card...that leaves 2 PCIe for my two 8800GT's. Come on nVidia a good card that wont chew up my electricity bill (UK) and perform well!!

    Make an 8800GT for 2010!! (The 9800GT...2xx version just aint worth the upgrade from my SLI set-up).

    Nice to see progress tho! bring on the GF108?
Display more comments