GPU dropped in new Sandy bridge CPUs

Remember those IGP-less Core i5 CPUs we mentioned last month? They’ve now appeared on Intel’s official price list (PDF). The chips come with their GPUs disabled and can be identified by the model numbers 2550K, 2450P, and 2380P. That first CPU features a fully unlocked upper multiplier, so it’s the one you’ll want.

Versus the existing Core i5-2500K, the 2550K offers 100MHz increases in base and Turbo frequencies, bringing the chip up to 3.4GHz with a 3.8GHz Turbo peak. Those speeds match the Core i7-2600K, which offers additional cache and Hyper-Threading support missing from the i5 series. Intel is selling the 2550K at $225—$92 less than the 2600K and a $9 premium over the 2500K.

The 2550K looks like a good match for Intel’s P67 chipset, which doesn’t tap into the Sandy Bridge IGP. That combo won’t be able to take advantage of Intel’s QuickSync video transcoding tech, though. Were I building a system today, I’d recommend sticking with the 2500K and pairing it with a Z68 motherboard to get QuickSync support. Intel seems intent on further segmenting its offerings, and the ones with the highest clock speeds may not always be the most desirable.

Comments closed
    • Bensam123
    • 9 years ago

    I’m personally not that interested in Quicksync… but the energy saving features of being able to use the on die GPU with Lucid Virtu when my normal GPU isn’t needed is well worth buying IMO. That is something that will actually save you money in the long run.

    I’m surprised we haven’t seen a power comparisons for it here on TR… or really any information about it in graphic card reviews…

      • sschaem
      • 9 years ago

      I thought modern GPU had super, super low 2D operations. Specially with a 28nm GPU VS a 32nm CPU

      I Guesstimate, switching would save ~10watt in active 2d mode and <3 watt in idle, and maybe nothing in long idle mode?…
      but its also messy at the driver/app level, it works but its a nasty hack.

      Thats a saving of maybe 72watt over 24h if the system is running in desktop mode.

      Not worth it for desktops.

    • Dposcorp
    • 9 years ago

    Wow, I didn’t realize so many people don’t trickle down machines as time marches on. Some of us keep reusing items as needed and even after, pass them down to people less fortunate. Also, a video card isnt like a $5 NIC or $5 sound card.

    I know plenty of people who are still running their 5850s/5870s, but have changed their board and CPU out many times.

      • BobbinThreadbare
      • 9 years ago

      You can get a brand new [url=http://www.newegg.com/Product/Product.aspx?Item=N82E16814127614<]video card[/url<] for $30. If you take a little time you can get used cards for even less.

      • A_Pickle
      • 9 years ago

      [quote<]I know plenty of people who are still running their 5850s/5870s, but have changed their board and CPU out many times.[/quote<] Those [i<]poor[/i<] people...

      • Bensam123
      • 9 years ago

      You can get video cards under $20 on newegg with a MIR. The AMD 5xxx series is pretty new still…

    • puppetworx
    • 9 years ago

    AMD makes far superior integrated graphics. Intel isn’t even a consideration unless you are opting for discrete graphics.

    Anyone who disagrees is a) wrong and b) wronger.

      • chuckula
      • 9 years ago

      No you are wrong: Not everybody plays 3D games, and Intel graphics trump AMD in many applications such as low-power non-gaming mobile applications, and workstations where the GPU in Sandy Bridge is more than adequate.

      Repeat this line 5 times a day for the next week: “There a lot more people in the world than me. Not everybody cares about <insert name of game here>. In fact, only a small minority of people care about <insert name of game here>. I should get a grip.”

        • Deanjo
        • 9 years ago

        Exactly, it all depends on what the user needs. The way that it looks right now intel is also coming out with integrated graphics to challenge the entry CAD graphics crowd. I wonder if AMD has any ambitions to bring their entry fire line into a version of opteron.

        [url<]http://vr-zone.com/articles/intel-graphics-to-challenge-nvidia-amd-in-cad-pro-space-/14613.html[/url<]

    • FormCode
    • 9 years ago

    I don’t know if it’s just me, but the naming schemes keeps getting more and more ridiculous. Next time someone asks advice about what cpu he should be i’ll just say “pick a number between 2300 and 2550 and you’ll be fine!”

    • moose17145
    • 9 years ago

    I am probably going to receive flak for this, but whatever…

    To me this kind of product actually makes a lot of sense from a marketing and business standpoint. Chances are these chips just have a flaw in the GPU portion of the die, while the rest of the chip is completely unaffected. So by disabling the GPU entirely and giving it a slight speed bump allows them to sell the defective dies that would otherwise be tossed (as one or two other people have mentioned). Personally, if I were in the market for a new i5, one of these is actually what i would likely go for. I do not do a lot of video encoding, so quicksync isn’t really something I am interested in, and I play games, so I will always be using a discrete GPU. Basically I just don’t have a need for a integrated GPU, so I just do not see getting a chip with it being disabled as being a big deal. I kind of feel like I cannot be the only one who doesn’t really give a dang about losing the GPU on the chip… if I was never going to use it anyways, and by having it disabled can get a 100mhz speed bump for the same price, then why not? (yes I am aware that 100mhz is like spitting into the ocean and thinking you changed it’s volume by a significant amount… but again… chances are I would never have ended up using the GPU anyways, so I am fine getting a minimal speed bump instead of paying for a function that I would never use anyways. I would have been paying the 200ish dollars for the rest of the chip that I will make use of… )

    That being said I probably wouldn’t be looking at anything outside of the 2011 socket SB-E’s anyways if I was in the market for a new CPU and motherboard. I like to tinker and overclock… but at the same time I need the VT stuff in the chip to be functional as well due to running virtual box’s and stuff. So for playing with overclocking and still having the VT stuff enabled I am kind of forced into the high end 2011 socket, which sucks… but that’s how the market segmentation crumbles and all that…

      • Stargazer
      • 9 years ago

      [quote<]I kind of feel like I cannot be the only one who doesn't really give a dang about losing the GPU on the chip... if I was never going to use it anyways, and by having it disabled can get a 100mhz speed bump for the same price, then why not?[/quote<] I don't mind losing the IGP either, and the IGP on my 2500K is disabled in UEFI. However, are you *really* actually getting a 100 MHz speed bump? If you're planning to overclock your 25*0K, there's no indication that the 100 MHz higher stock clock will result in any higher achievable overclock (and most people won't target the maximum achievable overclock anyway, instead settling for a lower power usage). So, *for me*, given equal price, there would essentially be a toss-up between the 2500K and 2550K, with the edge going to the 2500K because it has the *option* of using the IGP if I should for some reason find a use for it in the future, while the 2550K has no discernible advantage whatsoever. But, the 2550K actually costs *more*. Why would I pay *more* for *less*? Sure, if you have no intention of overclocking, then the 2550K will be a bit faster, so until (if) a non-K version of the 2550 pops up, I suppose it does have its place as a processor for people who don't want to overclock but plan to use discrete graphics cards (and that's presumably a quite significant market), but... Eh. It just feels lackluster. (and if a non-K 2550 pops up, the 2550K just plain doesn't make sense at its current price point)

        • moose17145
        • 9 years ago

        well 90% of the time my machine is running at stock speeds. I do overclock it from time to time if i am playing a game and wanna get a little bit more out of my system… but really I just overclock as a hobby / passtime / just because I am curious what I can eek out of my machine / because I find it relaxing to just tinker around with settings and is something I enjoy. Once I am done tinkering for the night or go back to doing ACTUAL work I usually set my machine back to stock just for sake of stability being required over moar megahurtz (that and for actual work I would hardly say that my current 1366 socket i7 machine is so slow that I need to overclock it). So the 100mhz would be more useful for me over the IGP… but if you plan on overclocking and leaving it there, then yes I can understand why you would want the version with the IGP still enabled over the extra 100MHz simply for the option being available should you need it or should something happen to your discrete GPU, since these likely do not overclock any better or worse than a 2500 with the IGP still enabled (unless the way they disabled the IGP on these has resulted in lower overall temperatures and power consumption compared to the IGP on a chip that is still functional but simply being ‘disabled’ / ‘turned off’ in the bios…. but I doubt it…)

    • DeadOfKnight
    • 9 years ago

    I understand the idea of marketing a cheaper GPU-disabled chip, but I would expect the price to go down to $200 or less, even with the 100Mhz increase. Maybe if the enabled hyper threading this would make sense.

      • Ifalna
      • 9 years ago

      Why would the price go down if the IGP is still there? If they’d produce new DIEs w/o the IGP it would affect yield rates so the price could go down. But this seems nothing more than a marketing gag to me.

        • DeadOfKnight
        • 9 years ago

        That’s only partially true. You clearly know nothing about binning.

          • OneArmedScissor
          • 9 years ago

          Thanks for educating everyone.

            • Ifalna
            • 9 years ago

            Binning = Component sergregation.

            2700K: can go at least X MHZ over spec.
            2700 normal: can stably run at spec but isn’t great to overclock.
            2700 with defective Graphics unit: Processor is still fine -> can be sold as igp less.

            NET Loss of components is LOWER than when they only sell fully functional devices, thus their Yield effectively increases and prices should theoretically come down a tiny amount.

            However in reality: You prefer to increase the processor speed by a harmless amount and then can use that psychological factor of “Yay it better so of course it’s more expensive” and make a profit. -> Hence why I called it a marketing gag.

    • shank15217
    • 9 years ago

    Woohoo, we’re back to Intel of the 90s (took only 2 years), it gets only worse from here on out folks.

    • Firestarter
    • 9 years ago

    But if you’re not using the IGP, how is that different from it being permanently disabled? Looks like if you’re OCing anyway you’d be shooting yourself in the foot, paying more for a chip that does less.

    • Dposcorp
    • 9 years ago

    This sucks for re purposing a CPU down the road in a Z68 chipset.
    You know, when you trickle the system down to your HTPC, server, or for a family member who does not need a discrete GPU.

    You will HAVE TO get a cheap video card for it.

    I have a cheap SINGLE CORE Sandy Bridge Celeron in a cheap Biostar S1155 board, and even it has a GPU built in.

      • ImSpartacus
      • 9 years ago

      Hell, if you need a discrete graphics cards [i<]that[/i<] much, I'll mail you my old 8800GT. Start the bidding at fifty bucks.

        • Dposcorp
        • 9 years ago

        My point was with the new chips I would have no choice but to use a add in card…….with the current chips, I can go 1U, 2U, etc…without needing any card, let alone video card that would fit into a small case.

          • chuckula
          • 9 years ago

          Hold on.. you aren’t one of those guys who thinks that every time a company announces a new product that all the existing products will be incinerated and death squads will be dispatched to execute those who bought the old products are you? You do know that Intel will still sell systems with the IGP right? You can step back from the ledge on the 18th floor now.

          Seriously, get over it. LOTS of CPUs have been sold without built-in graphics including every Bulldozer chip ever made from AMD, AND a bunch of new “Athlon IIs” that are actually… you guessed it… Llanos with the graphics disabled.

          If you desperately need to get a system for repurposing to an HTPC just choose one of the VAST selection of Intel or AMD chips that uh… actually includes an IGP. Wow that wasn’t too hard was it?

            • moose17145
            • 9 years ago

            To add to this….

            Intel also never once said they were going to stop selling CPUs without discrete GPUs enabled. All of these GPU-less models still DO in fact have the GPU built into them. In fact the silicon under the hood is identical to that of any other i5 that has the GPU enabled. It has simply been disabled, most likely due to a manufacturing defect that occurred which rendered the GPU portion of the chip inoperable, but otherwise left the rest of the chip at 100% and fully functional just as any other i5 out there. In order to help regain some of the yields that are lost from a manufacturing defect in the GPU, Intel simply totally disabled the GPU portion and bumped the speed of the CPU cores by 100MHz as a form of “compensation” for the loss of the GPU. Via this method are then able to sell them at essentially same price point as the GPU enabled versions.

            But no where did Intel or the article say that all future i5’s are going to have disabled GPUs. If you need a functional integrated GPU, then buy the i5 2500 that has a functional GPU. All this is, is just another option, mainly an option targeted at people like me who would never use the integrated GPU anyways and would actually get more use out of the 100 megahurtz than the integrated GPU.

          • BobbinThreadbare
          • 9 years ago

          This seems like a very silly concern to me.

          Oh no, I won’t be able to use my $300 CPU in the future without a $20 graphics card.

      • NeelyCam
      • 9 years ago

      Then buy a CPU with a GPU in it instead.

    • lilbuddhaman
    • 9 years ago

    I got money on these IGP disabled ones OC’ing poorly.

      • Arclight
      • 9 years ago

      I think they should overclock the same, hence there is no point not to get the cheaper model with IGP. Intel be crazy…

    • ImSpartacus
    • 9 years ago

    You have to pay [i<]more[/i<] to get an IGP-less version? At the very least, I expected the price to be the [i<]same[/i<]... On a side note, does anyone think it's funny that Intel is still using alphabetic suffixes even after they have plenty of digits to work with? They could've released all the "normal" CPUs as 2*50, the K versions as 2*70 and the P versions as 2*30. EDIT- I'm wrong. I'll leave this post unedited as a testament to my wrong-ness.

      • entropy13
      • 9 years ago

      [quote<]You have to pay more to get an IGP-less version? At the very least, I expected the price to be the same...[/quote<] You might have missed the 100MHz speed bump.

        • Stargazer
        • 9 years ago

        [quote<]You might have missed the 100MHz speed bump.[/quote<] When comparing two unlocked parts? Whee.

          • travbrad
          • 9 years ago

          Yeah, considering 2500/2600K will do 4.5-5ghz+, a 100mhz speed bump is pretty “meh”

            • swaaye
            • 9 years ago

            Not if you want sleep to work. I got stuck at 4.3 GHz because of that sleep issue with the PLL overvoltage.

            • bthylafh
            • 9 years ago

            I had that problem too. I wonder if Asus ever followed my suggestion to at least put a notification in the EFI and OC utility telling you about that if you try to use it.

            • swaaye
            • 9 years ago

            On my P8Z68, ASUS has PLL overvoltage set to ‘auto’ by default in the recent BIOS, which from what I’ve found means that it gets automatically enabled at ~4.4 GHz and disabled below that. My 2500k won’t boot Windows at 4.4 GHz with it disabled.

            However, the automatic overclocker feature of the mobo always sets me to 4.3 GHz. I think this indicates that they avoid entering the PLL overvoltage zone with that function.

          • Game_boy
          • 9 years ago

          Then you’re paying for the absolute guarantee it will be able to go that 100MHz higher.

            • Stargazer
            • 9 years ago

            If you’re not going to overclock and not going to use the IGP, you should be looking at the 2500 instead of the 2500K.

            If you *do* plan to overclock, I would personally be horribly disappointed if a 2500K failed to overclock by even 100MHz.

            Is there any record of a 2500K (that works at stock speed) that has failed to overclock by 100MHz?

            • travbrad
            • 9 years ago

            [quote<]Is there any record of a 2500K (that works at stock speed) that has failed to overclock by 100MHz?[/quote<] I'd be surprised if any 2500K failed to reach 4ghz, let alone 3.4ghz

        • ImSpartacus
        • 9 years ago

        It’s still stupid that they use the wrong names. The 2380P is just a 2400 without the IGP. So it [i<]is[/i<] cheaper than the 2400, but it's called a 23*0. And the 2550K is just a 2600K without the IGP, but for some batsht reason, the 2550K is an "i5" and the 2600K is an "i7". Like I said in the first post, Intel has four digits to work with, yet they still can't get a coherent system going even when they resort to alphabetic suffixes. EDIT- Wrong again. I'm already in the corner. Crying will start in a minute or two.

          • dpaus
          • 9 years ago

          [quote<]Intel has four digits to work with, yet they still can't get a coherent system going even when they resort to alphabetic suffixes[/quote<] Did it ever occur to you that that's deliberate?

            • ImSpartacus
            • 9 years ago

            If it’s intentional, then explain the naming system for [url=http://en.wikipedia.org/wiki/Sandy_Bridge#Mobile_platform<]mobile Sandy Bridge[/url<]. The alphabetic suffixes are actually the least confusing part.

            • dpaus
            • 9 years ago

            [quote<]Explain the naming system for mobile Sandy Bridge[/quote<] I can't. I don't think [i<]anyone[/i<] can, except maybe the guy who came up with it (and I'm not even sure about him). I've even noticed that vendor websites now include statements to the effect that 'Intel product naming schemes do not imply performance or speed capabilities' I suspect that they do it deliberately to mimize the kind of spec-price comparisons that sites like this thrive on (and from there, to eliminate 'comparison shopping' based on any factors other than the ones they want the comparison to be made on).

            • ImSpartacus
            • 9 years ago

            I understand now. In your original post, I thought you implied that alphabetic suffixes contributed to a more coherent naming system.

            • dpaus
            • 9 years ago

            My bad – I worked all weekend, and I’m at the stage where I’m not very coherent either. I don’t think any amount of alphabetic suffixes will help right now.

            EDIT: but if was going to try to use alphabetic suffixes, in addition to Intel’s ‘K’ suffix (perhaps implying that I can be ‘unlocked’ from my desk?) (please?), I could always go with GM’s famous ‘SS’ (for ‘Super Sport’), which would make me ‘dpausSSK’ – which would completely explain my lack of coherence.

            Bazinga!

            • derFunkenstein
            • 9 years ago

            just looking at teh table with the specs makes my head hurt, let alone the model names.

            • ImSpartacus
            • 9 years ago

            Yup. And the funny thing is that is is definitely doable with 4 digits.

            Let the first digit be the generation. Let the second be the TDP (1 for 17W, 3 for 25W, 5 for 35W, 7 for 45W, etc.). Let the third be clock speed or general CPU effectiveness. Then let the fourth digit handle GPU clock modifications. The i3,i5,i7 notation already handles stuff like hyperthreading, turbo and cache stuff.

            • BobbinThreadbare
            • 9 years ago

            I would switch 2nd and 3rd digits, but either way, it is easy.

          • Stargazer
          • 9 years ago

          The 2600K is an i7 because it has Hyper-Threading and a larger L3 cache. The 2550K doesn’t.

    • BobbinThreadbare
    • 9 years ago

    Is quicksync really that great?

      • TheEmrys
      • 9 years ago

      Yeah, it kinda is. I transcode video from 1080p to 720p or even down to 320×240 and its a very noticable difference.

        • geekl33tgamer
        • 9 years ago

        But I would rather have my Radeon 6870’s do it even faster. I’m sure that the majority of people who end up with these CPU’s have them alongside significantly faster GPU hardware…

          • ImSpartacus
          • 9 years ago

          If you’re multitasking, sometimes you want your GPU to be able to do other things while you’re watching a video.

          It’s only icing on the cake when I mention that Quick Sync is also the fastest decoder on the consumer market. Yes, faster than many (if not all) GPUs.

          If you don’t ever watch HD video, then you don’t need Quick Sync. Me? I like to watch movies sometimes.

            • rechicero
            • 9 years ago

            Do you really think most users encode video? I suppose I’m some kind of statistical oddity, as I only know one person who do that (once in a month, maybe?).

            From my experience, I’d say QuickSync is useless for most users. Great for people who encode (and doesn’t mind some of the downsides of QuickSync), of course.

            • burntham77
            • 9 years ago

            True, but “most users” probably don’t read this website. Or even know their CPU model.

            • ImSpartacus
            • 9 years ago

            Encoding is irrelevant, [b<]De[/b<]coding is where the most tangible advantage lies. If you don't like watching HD videos, then that's cool. You won't need Quick Sync.

          • Arclight
          • 9 years ago

          [quote<]Inte’s Quick Sync is untouchable though. It’s 48% faster than NVIDIA’s GeForce GTX 460 and 71% faster than the Radeon HD 6970. I don’t want to proclaim that discrete GPU based transcoding is dead, but based on these results it sure looks like it. What about image quality?[/quote<] Source: [url<]http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/8[/url<]

            • khands
            • 9 years ago

            I’ll be very interested to see if this holds up once AMD’s new stuff on GNC goes live.

            • sschaem
            • 9 years ago

            The AMD 7 serie is changing all that… well soon enough, hopefully.

            I think people planning to get a discreet GPU in the next 6 month have no need for quicksync.

            So Intel IGP in gamer machine are becoming like human appendix, it brings no benefit to the system.

            • geekl33tgamer
            • 9 years ago

            I take back my earlier comment and will source out some humble pie. I’ve just looked at this:
            [url<]http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/8[/url<] Even a Phenom II X6 in software transcodes video faster than a 6870, and the QuickSync is indeed untouchable. Impressive stuff, but im not sure how / why transcoding on a X6 is faster than the Radeon tho??? In theory, the 6870 should beat it easilly as it's got a lot more geometery computing power and bandwidth at it's disposal locally than the CPU...?

            • BobbinThreadbare
            • 9 years ago

            Is bandwidth important for video encoding? I would think that there is hardly any branching code and modern CPUs can prefetch everything very easily.

          • bcronce
          • 9 years ago

          I would love to see some benchmarks. When Intel’s IGPs first came out, they were getting 1/5th the FPS in game, but were twice as fast for transcoding.

          Your 6870 might not actually be faster. You’d have to bench it first.

        • BobbinThreadbare
        • 9 years ago

        How many options are there? When I encode video, I really like to poke around and mess with the various things.

          • Voldenuit
          • 9 years ago

          Quick Sync will not encode high profile H.264. For me, that’s a deal-killer.

          The encoder in the 7970 is supposed to be more configurable, but I don’t want to spend ~$600 on a graphics card either. Hopefully, Pitcairn (7800) and Cape Verde (7700) will have the same hardware encoder as the 7970.

            • BobbinThreadbare
            • 9 years ago

            I’m thinking a 7700 series card is going to be my upgrade from my 4870. I’m starting to get a little jealous when I see DX11 options I can’t turn on.

      • Goty
      • 9 years ago

      The quality isn’t particularly good (compared to software transcoding), so no, not really.

        • NeelyCam
        • 9 years ago

        For some people, the quality is more than good enough (i.e., those watching them on tablets/phones).

        I don’t use QuickSync, because I don’t like any of the software tools supporting it. I tried it a couple of times and it was way fast, but I can transcode with the CPU just as well, leaving it crunching overnight with some handbrake-based freeware.

          • Deanjo
          • 9 years ago

          Ya for some people it is “good enough”. Much like 128k mp3s are good enough for some.

          Quicksync to me is way overrated and I would prefer to have quality over speed (especially since transcoding using x264 is faster then realtime with even an AMD processor).

            • NeelyCam
            • 9 years ago

            I think it’s quite a bit better than 128k mp3s… maybe 192k.

            But I agree with you – CPUs are fast enough.

          • BobbinThreadbare
          • 9 years ago

          Avidemux is an amazing free software. I strongly recommend it for any encoding purposes.

      • pogsnet
      • 9 years ago
    • flip-mode
    • 9 years ago

    This doesn’t make sense to me. So these just have the IGP fused off? What’s the advantage? More overclocking headroom? Less power consumption?

      • derFunkenstein
      • 9 years ago

      100 extra MHz for the same dollar is about all I can tell. They may OC better but they’re not widely available yet to test out.

      • Waco
      • 9 years ago

      If you run with the GPU disabled anyway, I don’t see any difference at all other than the clock speed and less options down the road if your discrete GPU dies.

      Segmentation like this is retarded.

        • nanoflower
        • 9 years ago

        That would depend on what the yield is like on their chips. If they end up with a large number of chips that don’t have fully functional GPUs but otherwise work then this makes a lot of sense as it gives them a way to make money on those chips while apparently giving the consumer something extra (the 100MHz bump.)

          • flip-mode
          • 9 years ago

          If they’re rescuing yield then it would be nice to see a little price cut to go with that feature cut rather than a price boost. It’s some pretty mentally retarded product segmentation IMO.

            • crabjokeman
            • 9 years ago

            You’ll buy it and you’ll like it! Silence!

            • paulWTAMU
            • 9 years ago

            If it was 20-30 bucks less I’d be all over it–I don’t use IGPs anyhow, so I wouldn’t miss it, but I object to paying the same or more for less (and 100 mghz speed bump is worthless these days).

Pin It on Pinterest

Share This