Week In Tech: Nvidia’s ‘New’ Graphics Cards

Back in Feb we had a little chin wag about the mad dash of annual graphics hardware launches slowing to a saunter. We can add a little more flesh to the bones of that story this week, with some pretty plausible looking details of Nvidia’s upcoming plans – and further confirmation of nothing new from AMD. It’s worth a quick dip into the mucky waters of rumour for anyone pondering a GPU upgrade or a generally a new rig as some new kit – of sorts – is imminent.

So, the latest scuttlebutt goes something like this. Nvidia is tooling up to unleash its shiny new 700 series video boards. The bad news is that we’re largely looking at a bunch of rebadges, not properly new GPUs.

The good is that you’ll be getting more for your moolah. If the rumours are right, Nvidia is planning to pluck the mighty GK110 chip from the mental-money GeForce Titan card, turn a few bits off and stick a GeForce GTX 780 badge on it.

2,496 shaders instead of 2,688, 208 texture units instead of 224, 40 render outputs instead of 48. That sort of thing. Overall, we’re talking 80-plus per cent of a Titan. And thus still a monster of a GPU. But it will be £400 odd instead of £800. Yay.

Cheaper chips

OK, still not exactly a democratic pricing. But there’s more. The existing GeForce GTX 680 and its GK104 chip is also said to be getting a rebadge. It’ll take the GTX 770 slot and purportedly higher clocks. If so, you’ll be getting better-than-680 pixel pumping for roughly 670 pricing. Huzzah.

We’re still talking a likely £300 or so for the 770, so perhaps sir would be interested in the GTX 760 Ti, allegedly little more than a rebaked GTX 670? Now we’re down nearer £200 for a board that currently costs £300.

All these ‘new’ boards are due to begin appearing from the end of the month, or thereabouts. If you’re in the market, therefore, hold fire for a few weeks. If you’re lucky enough to already own a 600 Series board, then cue the rejoicing. Because your hardware effectively remains current when the 700 Series rocks up.

Nvidia’s Kepler-gen cards, like current fave GTX 670, are being rebagded

Apart from the happy side effect of not suddenly making your precious 600 board look about as cutting edge as an Atari 2600, that’s handy because it means you’ll continue to reap the benefit of Nvidia’s best efforts in terms of driver updates. Not that Nvidia instadumps old GPU designs when it comes to tweaking drivers. But, inevitably, the quest to flog kit means the latest chips must be a priority.

One other thing worth remembering about the Titan chip in the upcoming GTX 780 is that it’s a very different architecture from Nvidia’s GPU family proper. It packs twice the transistor count of GK104 but not twice the performance. In simple terms, there’s stuff in there for general purpose computing that doesn’t do a great deal for games. Makes it a beast for things like pro image rendering, if that’s your bag.

I reckon the GTX 780 will fly off the shelves and into the homes of freelance graphic designers.

And AMD?

As for AMD, as per my previous missive, it’s all change, everything stays the same. From what I can tell, AMD is plotting some Radeon HD 8000 series boards for later this year. But according to PDF files freely available from AMD’s website, these boards are straight rebadges – 7970 becomes 8970 etc – with no spec changes. Oh, and they’re only for system builders, not for the likes of you and I to buy. Odd, but there you go.

That said, if Nvidia does indeed roll out the cards detailed above, we should at least get a price drop from AMD to compensate. Generally, there will be much wailing and gnashing of teeth about all these rebadges, but personally I’m cool with it all.

That’s because at worst it’s a bit disappointing for chaps and chapesses who had an eye on a new flagship board. But how many fall into that group? For everyone else, we’ll have access to faster cards for less money. And given how modest the specs are on the next-gen consoles, the current crop of highish-end PC boards look good to go for a while.


  1. Brun says:

    Mmm, my 480 may need an upgrade soon, getting tired of it running so hot. That new 770 looks interesting since it will basically be a cheaper version of an ’80 card.

    We’re getting a new chip architecture with the 800 series though, correct? Might be better to just wait for that.

    • Carbonated Dan says:

      the current 70 is basically an 80 for 30% less money…

    • frenchy2k1 says:

      Most probably.
      The GTX8XX should be based off Maxwell architecture, but it is definitely a way off.
      Remember that in computer hardware, like in games, except for special occasions (like what happened in RAM last fall), there is always something better coming or the current thing can be bought for half the price in a bit.

      There is no end to the wait. Buy when you need something, be it a new graphic card or a new game. Buying from the previous generation may also be a good idea, as they will be cheaper (games get discounted after only a few months, graphic cards are now on a 18 months refresh, so discounts start around 3-4 months and peak around 18 months).

    • mtomto says:

      My “old” 590 card could also use an upgrade, but as you say it would probably only be to cut down on the heat if these “new” cards are only 600 series in disguise. Guess I’ll save for the 800 series then :)

    • Snakejuice says:

      It’s funny because I fucked your neighbor’s mom yesterday and she only demanded $5, she really should get her priorities straight.

  2. HexagonalBolts says:

    I’ve known of the existence of the practice for ages, but could someone please explain how it can make sense to ‘turn off’ parts of a chip and sell it for less when it still presumably costs the same amount to produce?

    I can understand that consumers want things at different price levels and that companies don’t want to do the research twice, but it still seems mad that, if they can afford to sell the chip at a lower price for the same cost of production, that they don’t just don’t this with the chip at full power.

    • Brun says:

      You save money by producing fewer hardware-level chip configurations, while still being able to penetrate lower-end markets.

    • FriendlyFire says:

      It’s the binning process. At the scales we’re talking about, you aren’t guaranteed to be able to fab the card you’ve designed. It’ll get done, but some pieces might underperform (for instance, heat up too much, or be unstable at normal operating values). Instead of scrapping those entirely at great costs, you’ll sell them as cheaper cards with the “bad” bits turned off. You get the benefit of only fabbing a few configurations, you get to recycle your duds and everybody’s happy.

      • PopeRatzo says:

        I’ve heard that explanation before and I’ve never totally bought it. You’re saying that they test every single chip to see if it performs to spec, and if not they just brand it as a lower-level model? That sounds like more work than these companies are going to put into chip manufacture. It’s not like oranges, where if it falls through one hole it’s labeled “small”, another and it’s labeled “medium” and a third and it’s labeled “large”.

        If that really is the way AMD and nvidia fabrication plants work, I’m disappointed.

        • Brun says:

          That’s exactly what it’s like, actually. Just to clear it up – I doubt they benchmark each and every chip in a batch with something like 3DMark. They don’t need to do that as benchmarks aren’t the specifications with which they market their chips. Instead they’re running functional tests on the chips at the hardware level to check for defective transistors and other nano-scale parts. They very likely have a way to automate this testing process. If such defects are detected they disable the bad silicon at the firmware level and sell it as a lower-spec card. This carries the triple benefit of being able to have a monolithic production process, being able to sell the chips that are “bad” rather than throw them out, and being able to sell those “bad” chips cheaply enough to penetrate the low- to mid-tier markets.

          The fruit analogy is quite apt, actually. When you’re working at nano-scales the tiniest, stupidest little thing can cause a batch of chips to be defective. Clean rooms aren’t perfect, so maybe some dust happened to find its way into your silicon, or the temperature wasn’t quite right. Obviously modern IC Fab reduces a lot of this but so far it’s proven impossible to reduce the number of defective chips to a negligible percentage. I imagine the fruit industry is similar – depending on the particular crop you’re still largely at the mercy of the weather, and if an early frost screws up 10% of your oranges you’re better off selling those oranges for juice or animal feed rather than just throwing them out.

    • jalf says:

      There are two angles to it.

      – on big, complicated chips (and GPUs are both big and complicated), the chance that a transistor here and there will be defective goes up, and the cost when that happens rises as well (because you have to discard a big complicated chip, rather than a small simple one). And if that happens, it is very nice to be able to just turn off the bits affected by it, and sell the chip cheaper, instead of discarding the entire chip. (Sometimes, a similar issue may be that the chip *works*, but gets a bit unstable at normal clock speed. But reducing the speed a bit might make the problem go away, so they do that, and sell the chip cheaper.)

      – the other issue actually has nothing to do with the hardware at all. It’s just about making money. Not everyone are willing to pay $600 for a GPU. So if you sell your one true GPU design at $600, you won’t sell much volume. On the other hand, if you sell it for $200, you’ll sell lots of them, but your profit will shrink dramatically because even the few customers who were *willing* to pay $600 now only end up paying $200. On the *other* other hand, you sure don’t want to design two completely separate GPUs just to cover those two price points either. So you design the one *big* GPU, and sell that at full price to those willing to pay, while disabling bits of it and on the cheaper models that you sell to stingy/poor customers. That way, you still get the full profit margin of the super high end models, while *also* selling something, at a reduced profit margin, to ordinary mortals.

      That’s exactly the same thing airlines do. The seat you get is exactly the same, but you can get it at 14 different prices depending on when you buy your ticket, how many bought tickets before you, the day of the week you’re flying, the phases of the moon, and a dozen other factors. The end result is that they get to sell *some* seats at a high price, while still being able to fill up the plane with cheaper customers who aren’t quite as profitable, but are a lot better than flying with a half empty plane.

      • frenchy2k1 says:

        correct and you can actually reach both of the goals you described with one technic: binning.

        Current chips are very complex beasts (both CPU and GPU) and have many duplicated parts.
        For example, current quad core CPUs have 4x the same hardware. Sometimes, some parts will be defective due to fabrication errors. This was the reason AMD created their tri-core CPUs: it allowed them to sell their quad-core chips that had one defective core.

        Nvidia works the same. They test their chips and based on the result, decide how to use them. For example, the same chip (GF104) is currently used for GTX680 (full chip), GTX670 (one SM disabled), GTX 660Ti (more hardware disabled) and the professional cards Quadro K5000 and tesla K10.

        This allows nvidia to maximize usage of a single chip and sell as many as possible (instead of discarding defective ones).

        Of course, the goal is to create products that reflect the number of chips you will get in each binning, otherwise, you will need to bin down (a good chip binned for GTX680 may have to be used in a GTX670 just to meet demand).

        Remember that you manufacture chips by wafers (a silicon disc of 300mm at the moment) and the cost for a wafer is fixed, so (cost per chip) = ((wafer cost) / (number of good chips on a wafer) ).

        The more you can increase that number, the better for the company.

        • PopeRatzo says:

          Nvidia works the same. They test their chips and based on the result, decide how to use them. For example, the same chip (GF104) is currently used for GTX680 (full chip), GTX670 (one SM disabled), GTX 660Ti (more hardware disabled) and the professional cards Quadro K5000 and tesla K10.

          That doesn’t sound so much like “binning” which is a term that I believe comes from produce farmers, but rather if Toyota just made one model car, and if you want a lower-priced model, they’ll take off two wheels and break the windshield. Want a lower-price still? They’ll bust the transmission so it only goes in 2nd gear.

          It would be one thing if we’re talking about a chip that meets spec goes for full price and one that does not meet spec goes for a lesser price, but that’s not what you described.

          • Elementlmage says:

            More like how Hyundai has various models of the Sonata, just in reverse. The Sonata and Sonata Turbo both use the same engine, chassis, and interior. But, the Sonata Turbo has 50 bucks in scrap iron bolted to the side of the engine and some tweaking done to the engine settings to produce 80 more horsepower than the base model Sonata. In turn, they slap an extra 10 grand on to the price tag.

            What Nvidia and AMD do, is they go ahead and produce the Sonata with the Turbo and engine settings already in the car, because at the micro-level, it’s cheaper to have a monolithic production line. Every so often, one of the Turbos doesn’t work quite right so they take it off, or the engine doesn’t balance quite right at higher loads, so they take off the turbo still, and sell it as a cheaper model Sonata.

          • jalf says:

            Nah, it’s just plain old segmented pricing. Like I said above, it’s exactly like what happens when you buy an airline ticket.

            You can end up paying $100, $150, $200 or $400 for the exact same goddamn seat on the exact same goddamn plane, being served the exact goddamn food. But they invent some arbitrary limitations and circumstances that allows them to sell some tickets cheaply and charge a small fortune for others. (Oh, you’re buying the ticket only 4 days before leaving? That’ll cost ya! And oh, I’m sorry, we only set aside 10 tickets in the cheapest pool, and those have already been sold. Oh, and the *other* cheap ticket form is only available on thursdays.”

            But the end product is *exactly* the same. They’re just exploiting that some people are willing to pay a lot, and others can live with a few constraints and limitations if it means they get the product at a lower price.

            Sure, *sometimes* hardware manufacturers do actual binning based on product quality (this GPU only works if we reduce its speed by 10%, and this one has a few defective shader cores that we’ll have to disable, so we’ll sell those cheaper), but most of the time, it has literally nothing to do with the hardware, and they’re just disabling bits and pieces on the cheaper models to allow them to sell at multiple different price points.

      • Dozer says:

        There was a printer sold a decade or two ago which came in two versions: basic and professional. It turned out the ‘basic’ version was the ‘professional’ version with an extra chip ADDED, which made the printer PRINT SLOWER.

        Market segmentation. You want to sell the full-price thing to the people who are willing to pay full price, and cheap things to people who will only buy cheap, but you want to prevent the people who are willing to pay full price from buying the cheap version. By crippling the cheap product, or introducing region locks, or making it convoluted to buy the cheap version, or stigmatising the cheap product with ugly branding (Tesco Value/Sainsburys Basics etc).

        Early railways had three classes. Second class was a basic railway carriage. First class was luxurious, with nicer seats and more space. Third class carriages had no roof. This wasn’t because it was expensive to put a roof on a railway carriage, but it was to scare 2nd-class passengers away from buying 3rd-class.

        And that is why it’s bad for the NHS to need to compete with private healthcare providers.

    • czerro says:

      The fab process, especially for new architecture, has tiny yield rates. The number of chips that perform at the goal frequency and stability is small, but the remaing parts can still be used stably at lower clock rates or with defective areas disabled. Same thing with locked cores on CPUs. They’re locked cause they don’t function to the standard expected, or perhaps not at all. The manufacturer isn’t crippling the chip for no reason and selling it at a discount, the chip is already crippled and unusable as intended. The manufacturer then examines the yields and determines from there what the substandards are that will result in them being able to use the majority of the remaining yield in lower clocked less feature-rich mid-range -> entry-level parts.

    • Wonkyth says:

      This is a general reply and thanks to you and everyone who has replied to you. I now understand how it works. ^_^

  3. Alfius says:

    Time was I used to spend months carefully researching my next rig before pulling the trigger. Now, thanks to RPS I know when to hold fire without having the buckets of free time I used to as a student. Good news for my employer, good news for my wallet, good news all round.

    I may just buy myself a new PC to celebrate – just not quite yet.

    • Screamer says:

      Do your own research as well, I bought a 670 in Feb on the recommendation in the “chin wag” article that there wouldn’t be anything new in a while, now 2 months down the line…… :/

      • Jeremy Laird says:

        To be fair, none of these mooted cards are new (well, the 780 is a new config of GK110, but anyway…). They’re just a bit cheaper. So assuming this turns out to be fully accurate, and as I said in this piece, your 670 remains very much current. Your 670 was based on Nvidia’s top real-world GPU. That won’t change with what is a relatively minor repositioning of a few GPUs. But tech getting a bit cheaper over time is obviously inevitable.

        Think of it this way – the point of the other piece was to say it didn’t look like that Nvidia would launch a new generation of GPUs, thus rendering the likes of the 670 redundant. So it was safe to buy in the knowledge you’d maintain cutting edge tech for the foreseeable. That appears to remain correct. So isn’t this really good news for you?

        • Screamer says:

          Yes you are correct :), and I know this is how technology work, especially PC technology, Its basically old the minute you buy it. Luckily I have the 670 OC Power Edition which runs about the same as a vanilla 680. It makes the “I could have bought a better one for the same price” pill a bit better to swallow ;)

      • Guvornator says:

        I take it you’re new to this PC business ;) Old scars, my friend, old scars…

      • Maxheadroom says:

        Same here. I bought one of those Samsung SSD’s on the back of an RPS recomendation and about 3 weeks later thay ran a follow up article saying “Hey y’know that SSD we told you to run out and buy? Well now they’re half price!”

        Yeah I know it’s not their fault/nature of the beast and all that, but it’s monday and I want a whinge :)

      • MattM says:

        Don’t forget the 700’s still aren’t out now.

      • ernierock says:

        I did the exact same thing, also a 670, just only 1 month ago. :(

    • analydilatedcorporatestyle says:

      Maybe the PR department fed ‘untrue’ stories to the press to shift the remaining stock before the new product line launch.

      That’s what I’d do if it was my responsibility to shift units before a rebadge/new line launch.

      Maybe a bit of cynicism could be applied by journalist but is that really their job?

      I hope those prize winners are enjoying their new kit?!

      • PopeRatzo says:

        Maybe a bit of cynicism could be applied by journalist but is that really their job?

        Yes, cynicism is exactly their job. If they’re not being a little cynical and looking sideways at every marketing brochure that comes across the transom marked, “Press Release” then they are NOT doing their job.

  4. remon says:

    So, where’s the report about the AMD 7990? Or are you only reporting Nvidia stuff RPS?

    • Jeremy Laird says:

      The 7990 is essentially existing AMD GPUs Crossfired on a card. While Titan is no more relevant in terms of pricing (both are thoroughly OTT), it was at least based on a novel GPU, which now appears set to trickle down a slightly more attainable price point. Titan’s therefore quite a bit more interesting and newsworthy than the 7990.

      If you’re after some AMD-centric coverage on RPS, we did an interview recently:
      link to rockpapershotgun.com

      And here where I make the case for reconsidering AMD CPUs:
      link to rockpapershotgun.com

  5. amateurviking says:

    Looks like I’ll be sticking with my trusty 560ti for the foreseeable then. I suspect I won’t need to upgrade until I switch up the monitor from 1680×1050.

    • Christian says:

      Yeah, I’m in the same boat, although I’m using 1920×1200, (which on a side-note I can only recommend instead of 1920×1080..never understood why someone would put up with that..) and can tell you it works mostly fine with that resolution as well.
      I’ve been thinking about an upgrade for a while now but really can’t find anything that gives me enough ‘more’ for the money. Maybe the GTX 760 Ti? We’ll have to see how quiet they stay if we’re just talking overclocked cards here though..

      • MacTheGeek says:

        Another 560Ti / 1920×1200 user here. My primary reason for contemplating an upgrade is that the 1GB of VRAM on the 560Ti just doesn’t get the job done anymore, with texture sizes really having taken off in the past couple years.

        Here’s hoping for some massive price drops on both sides of the fence.

        • kaloth says:

          Seems like a popular card, I have one too!

          I’m actually considering trying to find a matching card for the one I’ve got now and run it SLI as a stop-gap solution til the new chip architecture comes out. The 560ti’s are super cheap now, and running two cards SLI will give me roughly a 660ti/670 equivalent (couldn’t find an exact benchmark, but found that 560ti sli is a little more powerful than a 580, and the 580 is just a hair above the 660ti and just below the 670).

  6. Memph says:

    Eep and I’d just gotten my heart all set upon a Gigabyte OC GTX660Ti. I knew new stuff was coming, but not quite so soon. cheers for the heads-up and info.
    Meanwhile my olde 460 will have to soldier on…

    • Jupiah says:

      I just bought an ASUS GTX660 Ti a month ago. If these new cards are significantly cheaper for the same power I’m gonna feel like a bit of a sucker… but I guess I can’t complain too much since I still own an awesome graphics card.

      • ernierock says:

        Yup, same thing here with a 670. It hurt so much to see “so the 770 will essentially be the same price as the 670 but as fast as the 680”. But indeed, you can never know in the PC world and I still own an epic card :)

    • Darlizzle says:

      soldier on with that card, come on mate, i’m still running a GTX 275 lol, that thing sounds like a whirlwind when running FC3: Blood Dragon haha.

  7. stahlwerk says:

    Any word on RAM loadouts? From what I gather that will be the key thing, coming the next console gen, with the GPU part of the APU having access to multiple GBs of fast, unified RAM, i.e. shared in full with the CPU part.

  8. povu says:

    New graphics cards just as we get a new CPU release from Intel, info on the next Xbox, and the release of Metro: Last Light? Well timed.

  9. Wut The Melon says:

    On the somewhat confusing thing with AMD not releasing HD8000 series: as far as I can understand it, AMD isn’t going to release the 8000 series for desktops, but instead will work on the HD 9000 series which might still see a release this year.

    So it is almost as if nVidia and AMD agreed to have another bullshit-naming-generation (700 and 8000) before continuing development on a proper new generation of GPUs.

    Source: link to tomshardware.com

    • Jeremy Laird says:

      Yeah, I’m afraid it does rather work like that – AMD and Nvidia seeming to conspire with one another. Sort of off-the-record armistice. Think that’s what happens when they have fairly evenly matched products, as they do at the moment. Its in both their interests to cool off a little and keep the cash rolling in on existing products for a while.

      Think the AMD 9000 series is dependent on TSMC’s 20nm node, which actually seems to be on track, so yes, we might just get 9000s at the arse end of the year. Doubt it will be much earlier than that as I think 20nm wafer starts only kick off at the end of Q2, if you care about that sort of thing.

      • Demontain says:

        If AMD are releasing new cards this year, which they are supposedly, it will probably be a 28nm refresh. I can’t find the exact article but I do remember AMD saying they will be sticking to 28nm a while longer considering moving to new processes are becoming more and more expensive with less benefits. This might also be the reason why they recently released Bonaire (7790), which could’ve been a Cape Verde replacement for the 8000 series but they used it to fill the gap between the 7770 and 7850 instead.

        I’d say we’ll see a new 28nm line-up at the end of the year and the 20nm line-up late 2014.

        Also, that article talking about Volcanic Islands, which tech sites love to copy from wherever it got leaked from, is most likely a fake. It pretty much shows a big APU with a 16-core Bulldozer/Piledriver/Steamroller and 16 GCN CUs.

  10. Scroll says:

    Good. My 560ti will do the job for now as I’m already bottle necked by my cpu, the longer I can hold off on a full upgrade the better.

    • nrvsNRG says:

      Same here.
      I have a feeling we’re in the same boat as quite a few other 560Ti based rig owners.
      I had a new set up already planned but I’ll hold on and see what the prices are like in a couple of months.
      …actually I’m not even convinced how much better things will be unless I upgrade my monitor first as I currently have a 1920X1080.

    • CantankerousDave says:

      My desktop machine is still rocking a Q6600 and Radeon 4870. The only thing I can upgrade without replacing *everything* is the graphics card, but the rest of the system would be a bottleneck if I did. Then again, I rarely turn it on these days, working and gaming on a mid-range laptop that’s much more powerful.

      • Scroll says:

        I’m on the same cpu, the cost of upgrading that is out of the question right now. It’s not worth overhauling the whole system until I come across a game that simply doesn’t work on my system.

      • Arglebargle says:

        Still running a q6600 here as well, though my 4870 finally bought the farm, now got a warrenty replacement of the odd duck 6790. Given my screen size, there’s no real reason to upgrade right now. When the time comes, pretty much eveything will have to change. Though I do like my old school case.

        Once the new consoles hit stride, and developers start developing seriously, and porting games from them, it’ll start to get tougher though. Still, that’s probably a year or so away.

      • sabrage says:

        I’m on an e6750 coupled with a Radeon 6850, and only then because my 8800GTX bit the dust. An SSD is enticing, but I won’t be getting any other new hardware until my CPU flat-out can’t cut it.

      • edna says:

        I think you can safely upgrade your GPU actually. I’ve just paired a 650 Ti Boost with my Q6600@3GHz and I get about the same fps as the review sites, so I don’t seem to be bottlenecked. If it isn’t a problem for Skyrim and Bf3, then it shouldn’t be for anything else, I wouldn’t have thought. Certainly a nice improvement over the 5770 that it replaced.

  11. kael13 says:

    I’ve thought about upgrading my rig a few times, but I just can’t see any real reason to. Really I should save, but I think I’m going to end up splooshing on one of them fancy, thin Apple murchines this Summer.

  12. Crainey says:

    Good timing on this article, I was just about to update by 460 (I know) to a MSI 660ti PE. Looks like I will hold back and see what Nvidia has to offer.

  13. golem09 says:

    560ti owner here as well. But I guess I’ll make a big leap next year for my consumer oculus. With whatever is available then, hopefully the 800 generation.

  14. Iskariot says:

    I am still running a 295. It holds up quite well.
    I am currently in the process of researching a new system and have been waiting for the 700 series to get released. I am still not quite sure what will be my next card. Perhaps a 780. Out of pure financial recklessness I am even considering the Titan..

  15. darkChozo says:

    Random advice request: right now, I’m running a 2500K/560Ti and have started to look at maybe upgrading my rig, both because I’m starting to see performance issues and because I have significantly more money now than I did two years ago. I’m thinking I want to advance a couple generations and a price tier or so, probably putting me at a 4770K/770 or thereabouts.

    I feel like staggering my purchases is somewhat more fiscally prudent then dropping $1000 or so all at once, so I’m thinking of getting a 4770K+mobo whenever that gets released (my worst-running games are CPU monsters like ARMA and PS2, plus I tend to leave a bunch of processes running in the background), and grab a [ 7/8][70/80] sometime in the future.

    Anything unreasonable in my lightly-researched and not-well-thought-out plans?

    • MacTheGeek says:

      This is basically what I did, starting several years ago.

      In 2009, I built a rig with an E6850 and an HD 4890. In 2011, I upgraded the video to my current 560Ti; in 2012, I upgraded the CPU/RAM/mobo with a 2500K setup. (All of my castoff parts have since been reassembled, in a computer that my wife now uses for nursing school. That’s the hidden benefit to upgrades: hand-me-downs!)

      If you’re budget-conscious, upgrading in pieces is the way to go. You can shop around and wait for killer deals on parts, buying them when the individual prices are lowest (instead of buying everything at the same time).

  16. TechnicalBen says:

    Relabeling/badging last years product to be “new” should be made illegal. Relabeling/badging last years product as “new” and upping the price by £50 should be punishable by a slap to the face with a large “tickling feather” (cricket bat with “tickling feather” label/badge on it).

    • identifierad says:

      Yah. Point is, though, they’re *lowering* the price.

  17. Carra says:

    So my 1 year old Geforce 670 will still run everything, yey :)

  18. csmaster2 says:

    So why are we talking about 7970 while theres 7990 ????

  19. Gailim says:

    perhaps it might finally be time to update my old GTX 260

    1080p gaming at med-high since 2008

  20. MrLebanon says:

    i’d be interested in seeing how two “760 ti’s” run in SLI

    • grundus says:

      Probably about the same as two 670s in SLI.

      What I want to know is: Can I SLI a 680 with a 770, or would it be cheaper/better to get two 760Tis instead? If only I had a backup graphics card, I’d sell my 680 now and sit on the cash until the 760s are out…

      • Brun says:

        Unless something has changed SLI only works with two identical cards – or rather, it’s only officially supported when used to link two identical cards. I think people have hacked it to work with other configurations, but there were stability issues and it’s generally not a good idea.

        Also, conventional wisdom is that a single, more powerful GPU will provide better bang for your buck than two less powerful GPUs in SLI/CrossFire. SLI support has improved greatly in the past few years but it can still cause problems that single GPUs do not have.

  21. SuicideKing says:

    Hey I read something a few days ago about Volcanic Islands and a 9970 from AMD launching later this year.

  22. uh20 says:

    i was planning on buying the 670 this month but due to my child’s budget, im going to wait and buy a 760Ti or a 770

    its kind of sad that i have to completely disregard AMD, they make some good cards, but their linux support is horrifying, so i can’t really buy any of them.