Leak pegs desktop Broadwell, Skylake for mid-year

During our live podcast on Wednesday, a viewer asked about the launch schedule for unlocked Broadwell desktop processors. I guesstimated that the chips would arrive mid-year, around the Computex trade show, which is when motherboard makers traditionally roll out new products. The latest leaked roadmap suggests a slightly earlier release, and it maps out Intel’s plans for the rest of the year and into 2016.

VR-Zone’s Chinese alter-ego published the graphic, which schedules unlocked Broadwell chips for mid-Q2. The roadmap doesn’t mention Broadwell-K specifically, but it does refer to unlocked chips with an LGA package and 65W thermal envelope.

If the leak is accurate, the next K-series parts will be based on Skylake-S silicon. They’re slated for the end of Q2, shortly after Broadwell hits the desktop. These unlocked chips will have 95W thermal envelopes, according to the roadmap, and they’ll be joined by non-K variants with 65W and 35W TDPs. Broadwell and Skylake should be able to coexist on the desktop, since the former will work in existing 9-series motherboards, while the latter will presumably require a new chipset.

The mobile variant of Skylake will reportedly debut just before the desktop chip. 15W and 28W flavors of Skylake-U are on tap, both with BGA packages. There’s no mention of a lower-power replacement for the Core M, though. The lowest-wattage chip on the roadmap is the Braswell replacement for Bay Trail-D, which is supposed to arrive early in Q2. That SoC is listed with a 10W thermal envelope, and it seems to be targeted at low-power desktops. Intel is already shipping Cherry Trail, its Bay Trail successor for tablets.

The last item on the roadmap is Broadwell-E, which is scheduled for the first quarter of next year. That CPU should plug into existing X99 motherboards, and it’s listed with the same 140W thermal envelope as Haswell-E.

Comments closed
    • Prototyped
    • 5 years ago

    Ah, the new successor to Core Duo as a stopgap. At least Stopgap Duo had a run of about 6 months (between Dothan Pentium M and Merom Core 2 Duo). Sounds like Broadwell will be “Stopgap Quad”, with its life as the current-generation desktop processor being less than a quarter year. (A bit longer as a mobile processor, but essentially at best September 2014 through June 2015.)

    • tsk
    • 5 years ago

    Hurry up! I’m building a computer as soon as R390x/titan II, a decent 21:9 freesync/gsync monitor and skylake becomes available. Altho I might go with haswell-e if skylake is late to the party.

    • themattman
    • 5 years ago

    When these processors are released, I would love to see some comparisons between these new processors with some older 2008-2011 era tech.

      • UberGerbil
      • 5 years ago

      It would be nice for the graphs to include examples going back all the way to Sandy Bridge. The trouble is, all the software gets updated multiple times in the interim; a patched and updated game or multiply-rev’d benchmark running on Win10 with 2015 drivers is hardly comparable to the much older code running on Win7. And if Scott (or his minions) were to go back and retest on the old hardware with the new software, it vastly multiplies the testing burden — assuming he even still has the old hardware available.

      But the low-level synthetic tests might be doable, and comparisons on those to much older hardware have sometimes shown up in past reviews.

        • VincentHanna
        • 5 years ago

        I agree with this, Sandy not only had higher base clocks, but is still the undefeated champ of the OC. Its not uncommon to have a decently cooled sandy OC to 5ghz for daily use.

        As far as I’m concerned, its the standard bearer for the intel -k community.

        • MadManOriginal
        • 5 years ago

        You’re right, but a few representative examples of popular ‘mainstream enthusiast’ CPUs from a given era would be fun.

      • Ninjitsu
      • 5 years ago

      I think TR or Tom’s or AnandTech did this not too long ago, probably was TR.

    • Ninjitsu
    • 5 years ago

    Skylake + DDR4 + cheap 980 equivalent = I’m saving for it.

    • PrincipalSkinner
    • 5 years ago

    Can’t wait. Also, can’t wait to see what benefits it brings to laptops.

      • stdRaichu
      • 5 years ago

      Indeed – I’m sure we’ll soon be seeing laptops that have the same 3hr battery life but are now only 5mm thick.

        • Generic
        • 5 years ago

        People will not truly understand the genius behind thin laptops until 2035 when they move to solar power and drop the on board battery completely. The OEMs are just planning ahead I tellz ya.

        And. Bonus. By then you will find yourself adoring the zero-travel keyboard and razor sharp edges the printed circuit sheet metal provides.

          • Voldenuit
          • 5 years ago

          The Daystar! It burns!

          • stdRaichu
          • 5 years ago

          Zero-travel keyboards are old hat – the new iphone is just a prototype for bendable touchscreen keyboards.

          Just use a touchscreen as a keyboard, put the solar cell behind the touchscreen and the light from the top not-a-keyboard touchscreen will power the laptop so we’ll finally have solar power that works in the dark (as long as your hands aren’t too big and you don’t use the keyboard too much).

          A razor-edge laptop would also be perfect for knocking gillette off their monopoly perch for mens shaving products, meaning businessmen on the go no longer need to waste valuable luggage space on a handheld razor and can thus just smear their laptop in foam instead. Sharp edges technology will also give a big boost to luggage companies and elestoplast so not seeing any downsides here.

    • albundy
    • 5 years ago

    Another new intel performance chipset, eh? Nothing from AMD since 2011. When does ridicule turn into pity?

      • chuckula
      • 5 years ago

      That all depends on how our friend Zen turns out, although Zen + integrated graphics isn’t coming until 2017, so ridicule can continue for a couple more years 😉

        • derFunkenstein
        • 5 years ago

        did I miss the latest on wccftech? What happened to 2016 and FM3 socket? TR’s writeup said there’s no mention, but do we really think AMD won’t push their GPUs into every product going forward?

          • chuckula
          • 5 years ago

          Communications issues:
          1. I’m calling anything next-gen from AMD “Zen” because we don’t have any other solid names BUT: Zen is technically just a label applied to a microarchitecture and not to a specific chip with specific features.
          2. “Zen” [b<]is[/b<] coming in 2016 (barring potential disasters at GloFo). However, the 2016 Zen does *not* appear to include graphics. Instead, it's a CPU-only chip ala an FX-8350. 3. The version of "zen" that gets integrated graphics is 2017, however.

            • derFunkenstein
            • 5 years ago

            OK, I guess it’s just at point #2 we diverge: I live under the assumption that everything everywhere will have integrated graphics going forward. I looked at the last bit of rumor/news as “they didn’t say no graphics, so I assume there’s graphics”

            • VincentHanna
            • 5 years ago

            I really do too. I understand a lot of enthusiasts are always screaming “kill the integrated graphics and use that TDP for something else, all day long… And I understand that a system with a discrete GPU doesn’t need another 7-12 GPGPU cores hanging around for no reason, but I still maintain that CPUs that don’t have a GPU attached are feature lacking.

            1) Safe-mode/clean-boot/ fresh installs– I can’t tell you the number of times I’ve booted up a clean PC and found that MSFTs drivers didn’t even acknowledge the existance of a discrete GPU.

            2) Operating systems, web browsers, even office apps these days DO use GPU acceleration. Best case scenario, we can keep that kindof stuff off of the front side bus.

            3)GPUs die. I know, I know, it sounds like a fairy tale, but it has happened.

            4) I just spent upwards of $200 on a chip! I want it to have a GPU on it, dag blame it!

            5) The intel/AMD double standard. I EXPECT Intel to wipe the floor with AMD on the CPU side. I EXPECT AMD to steal Intel’s lunch when it comes to on die GPUs… Heck, I’m still waiting for PS4 quality APUs to manage to get themselves socketed onto a 14″^2 mobo somehow. <<–I want this chip!(with a real processor attached of course) Nvidia promised us something similar would come out of project Denver, but it never happened.

    • anotherengineer
    • 5 years ago

    Neely is getting warm fuzzy thoughts.

    I know it!!

      • NeelyCam
      • 5 years ago

      Neely is feeling conflicted. I’ve been itching to upgrade my laptop for… well, years. And now I’m wondering if I should pull the trigger as soon as Broadwell 15-28W chips are widely available (e.g., Dell XPS 13), or if I should wait until yet another Black Friday in hopes of getting a Skylake system?

      Grr. This cycle never ends…

        • Ninjitsu
        • 5 years ago

        Wait for the 10-series platforms to be announced/leaked/detailed, otherwise I’d say Broadwell.

        • MadManOriginal
        • 5 years ago

        At this rate, the year of Linux on the desktop will actually happen before you replace your laptop.

        • Stochastic
        • 5 years ago

        I too was conflicted as I had been waiting to upgrade my laptop for years. In the end I got the new Dell XPS 13 with the $100 online coupon. Aside from a slight increase in performance/watt, I don’t think the Skylake version that will be available at the end of this year will make all that much of a difference in my day-to-day use of the machine. When I upgrade again in four years time, hopefully 4K OLED displays will be standard along with NVMe SSDs and post-Cannonlake CPUs from Intel. Until then the XPS 13 should do just fine.

    • bfar
    • 5 years ago

    Looking forward to seeing what Skylake can do 🙂

      • anotherengineer
      • 5 years ago

      You can see electrons move inside silicon!!!!!!!!!!!!!!!!!!!!!!!!

    • Shambles
    • 5 years ago

    Skylake can’t get here soon enough for me. The only thing that is holding me back from my NAS build is an affordable platform that supports DDR4 and virtualization.

      • DPete27
      • 5 years ago

      So Skylake will be DDR4?

        • derFunkenstein
        • 5 years ago

        Now that DDR4 is shipping and widely available, it’s wise – memory makers aren’t going to want to do both forever, or even for a long time.

        If history repeats itself, I think it’s a pretty safe bet. DDR3 appeared in X58 in late 2008, and over the course of 2009 basically everything else from Intel went DDR3. Core 2 chipsets supporting DDR3 appeared first, and then Lynnfield. By the end of that year everything was DDR3. Same will likely be true of DDR4, though maybe a little longer. Zen is also rumored to be DDR4, available late next year.

          • the
          • 5 years ago

          To nitpick a little, the socket 775 chipsets support DDR3 arrives in mid 2007. The transition to DDR3 went rather quickly after Nehalem started shipping in late 2008 as it was the chip to get for gaming.

          The reason I see the DDR4 transition taking a bit longer than the DDR3 transition is due to costs. I’m not expecting DDR4 to drop in price much throughout 2015. Memory manufacturers are pleased that DDR4 is a high end part used mainly in servers (Haswell-EP). The customers buying those systems are businesses that can maintain a price premium. Even when SkyLake launches into the consumer mainstream, I see it being packaged with DDR3 except in premium devices here power consumption is more important than price.

            • derFunkenstein
            • 5 years ago

            my bad – didn’t realize DDR3 came so early. So Intel’s transition was complete within about 18 months, tho.

            • willmore
            • 5 years ago

            I think we may be seeing DDR4 drop in price faster than you think. Look at what we’re seeing with planar Flash. Past 2X nm, it’s just not practical to make devices. So, everyone is moving to 3D Flash. What’s going to happen to all of that 2X nm fab capacity? You can make DRAM on those production lines, but you can’t make most other kinds of chips.

            So, we’ve got a glut of DRAM/Flash fab capacity. What kind of DRAM do you think they’ll retool to make? DDR3 or 4? It seems an obvious choice to me.

            • the
            • 5 years ago

            You answered your own question: flash. The demand is still there regardless of the process geometries.

        • Ninjitsu
        • 5 years ago

        90% yes, from what we know.

      • Flying Fox
      • 5 years ago

      NAS should not need that much memory bandwidth provided by DDR4. What kind of VMs are you thinking of running on that box?

        • stdRaichu
        • 5 years ago

        To be honest there are very few uses outside of “[i<]HOW much?![/i<]" servers where you can actually use the bandwidth available from modern CPU/memory effectively. IIRC DDR4 gives you something like 200GB/s whilst most DDR3 was somewhere in the 100GB/s ballpark but we've been well into the area of diminishing returns since the DDR2 days - going for super-expensive RAM hasn't made economical sense or even much of a performance increase for quite some time now. Buy memory for capacity first, latency second, and only buy for extra bandwidth if you can't conceivably fit any more silk toilet roll in the toilet for your skiing cabin in Dubai 😉

          • auxy
          • 5 years ago

          Quad-channel DDR4-2133 is only ~70GB/sec.

          A typical Haswell machine has 25.6GB/sec memory bandwidth.

          • tipoo
          • 5 years ago

          DDR3, 100GB/s? DDR4 at 200? Whachya talkin bout? Dual channel DDR3 at 1600MHz is 25.6 GB/s, and DDR4 is not dramatically faster yet. Even quad channel DDR4 at its highest clock would probably not get you near 200.

    • chuckula
    • 5 years ago

    [quote<]If the leak is accurate, the next K-series parts will be based on Skylake-S silicon.[/quote<] Well.. not quite. More like "the next socketed desktop parts will be based on Skylake-S silicon" [quote<]The mobile variant of Skylake will reportedly debut just before the desktop chip. 15W and 28W flavors of Skylake-U are on tap, both with BGA packages. [/quote<] ???? Broadwell-U is just shipping now. There's no way Intel is upstaging it that quickly (and even if Intel did, the OEMs would never go for it*). Do you mean that Skylake-U is shipping AFTER Skylake-S (like late this year into next year)? Mid Q2 for Broadwell-K is a bit earlier than I had expected. I was assuming an unveiling at Computex in early June with availability later that month, which is basically the tail end of Q2. If the leak is right, it might be referring to when Broadwell-K ships en masse to retailers and OEMs with the on-sale date being a bit later. * You might ask: What about Skylake-S and Broadwell-K on the desktop? Well the validation for desktop parts is much easier than for full-blown notebooks. Additionally, Broadwell-K is using existing Z97 motherboards, so the OEMS only really have to validate a new motherboard for Skylake and do a relatively minimal SKU update for Broadwell-K chips. That validation is much easier that juggling inventory for two sets of notebooks that launch within only a few months of each other.

      • the
      • 5 years ago

      This is the fallout from all of Broadwell’s delays. The Broadwell-K parts were originally supposed be available in 2014 so they’re close to 6 months late. The one thing Intel decided to do to make up for the delay is releasing a desktop version with the eDRAM for L4 cache. That could be the high end Broadwell part that’ll hang around for the later half of 2014. Realistically, if a Haswell user wanted to upgrade to a Broadwell, that’d be the chip to get as the upgrading just for power saving isn’t going to be worthwhile.

      The real surprise is the quick release of SkyLake-S. I thought Intel would give at least 6 months of space for Broadwell-K on the desktop but it looks like they last a quarter before being replaced. This small time span is odd and I suspect some particular OEM needs are being met here. It is also good to see unlocked quad core Sky Lake parts arrive this year. Hopefully it’ll be enough to get the many Sandy and Ivy Bridge owners to upgrade.

      The other thing that concerns me is that Broadwell-E has slipped in 2016. I was hopping to see it launch in late 2015Q3 or 2015Q4. I was under the impression that this year there wouldn’t be much of a launch time difference between Broadwell-K and Broadwell-E on the desktop. Intel indicated that these parts weren’t delayed due to the TSX errata so I’m scratching my head as to why they’re now launching in 2016.

        • Ninjitsu
        • 5 years ago

        From what I know, Broadwell was not hitting the desktop in the -K form, which would have been left to Skylake.

        In fact, I still think we’ll get socketed Broadwell-H series parts (ones with the L4), while the -K series goes to Skylake, hence the closely spaced launch.

        I’m assuming Broadwell-E either:
        A) Won’t launch, but give way to Skylake
        B) But Intel usually ships two -E generations on the same socket, so A) is less likely, and the reason is probably that they aren’t satisfied with Haswell-E sales.

        IIRC there was a two year gap between SNB-E and IVB-E, that too during a tick cycle, so maybe something similar is happening this time.

        I’m also wondering whether Broadwell provides enough of a justification in terms of performance for an -E part. IVB-E wasn’t as “new” as SNB-E or HSW-E were launched, and consequently didn’t have much of a point except compatibility. Maybe Intel’s just going with (A) after all.

          • chuckula
          • 5 years ago

          [quote<]I'm also wondering whether Broadwell provides enough of a justification in terms of performance for an -E part. [/quote<] It does merely because any -E series part is just a slightly tweaked Xeon chip... and Broadwell Xeons are most certainly going to be a big thing due to power savings even if they don't perform any differently than their predecessors.

            • the
            • 5 years ago

            The Broadwell-EP chips that make up the Xeon’s will likely have an increased core count due to the die shrinks. Intel has been using two or three different dies for the Xeon E5 chips, with the smallest die becoming Core i7’s for consumers. (Yeah, that Core i7 59760X is ‘small’ compared to the massive 18 core Xeon E5 2699 v3.)

            Broadwell-E will either increase clock speeds at the same power envelop or reduce power consumption well below the 140W max it currently ships at.

            • Ninjitsu
            • 5 years ago

            Solid points, both. Thanks.

          • MDBT
          • 5 years ago

          Ninjitsu the way I read your post you seem to say that there will be no Broadwell-K and that you doubt the existence of Broadwell-E. I fail to see how you came to either of those conclusions as previous leaks and roadmaps appear to contradict both points.

          In early 2013 it was reported that there would be no socketed Broadwell desktop variants. Last year when Haswell-K refresh was shipped it was clear to the industry that socketed desktop broadwell-K was coming and that the 9 series chipset it was supposed to launch with would support Broadwell-K processors when they launched in 2015. Roadmaps also show Skylake-K coming after Skylake-S. This leak does not contradict either point, except for the poor wording chuckula already pointed out. The exact Broadwell-K timing however has been a shifting variable as Intel has experienced some delays.

          E series parts have traditionally been one full generation behind the mainstream desktop socketed chips so having Broadwell-E release some time in 2016 with Skylake-S already available and Skylake-K still missing in action would not be out of character. As for “not having much of a point” the same could be said of every chip on the “tick” side of the cycle is only intended to be a process shrink. Ivy shrunk Sandy, Broadwell will shrink Haswell, etc.

            • Ninjitsu
            • 5 years ago

            No, I’m merely doubting Broadwell-K because:

            1. As you mention, Broadwell wasn’t supposed to socketed at first, focusing on mobile only.

            2. Broadwell desktop seems to be too close to Skylake desktop, and given (1), it seems a possibility that Intel simply releases a Broadwell-H variant (for the 9 series boards) along with Skylake-S,K. Or, there may be no -K series at all this year.

            3. The shifting variable can only shift so much before it’s scrapped altogether. Broadwell-K will likely not perform or overclock much better than Haswell-K, so to launch it alongside Skylake-S may not be the best idea*. Unless of course, Skylake-S comprises the below 65W variants (e.g. i7-6770S) only and isn’t some new nomenclature for parts.

            *who do you sell it to? Devil’s Canyon or other Haswell-K users won’t find the 5% increase worth it, and new buyers can get similar performance and a much newer platform with Skylake (+DDR4, which will be another marketing point). Only people that remain are 9-series owners with Core i3s or lower.

            I don’t have enough information to doubt Broadwell-E, as you can see I’m going back and forth on it. I’m just wondering if, given a late 2016 release and IVB-E’s lukewarm reception, combined with a then 2-year old platform, it may be more suitable to simply transition to Skylake-E directly, because then again, the release cycles will coincide.

            Last of all, ticks* usually bring, as a result of the process shrink :
            1. improved GPUs
            2. better efficiency, lower power or both
            3. Some newer platform features

            Ticks are usually good for mobile, especially broadwell, but “the” and chuckula make good points above increasing core count for Xeons.

            *I’m making them sound like lice. :p

        • UnfriendlyFire
        • 5 years ago

        Some of the laptop OEMs mentioned that they plan on sitting out on Broadwell or doing a limited release, in preparations for Skylake.

        Aka, the demand for Broadwell is going to be a lot smaller than Skylake if they’re both launching around the same time. And Intel isn’t going to give AMD or Qualcomm/ARM any ground.

        • VincentHanna
        • 5 years ago

        If Intel wants us sandy people to upgrade, all they need to do is come out with a new chipset that is actually worth getting.

        The last time I upgraded, it wasn’t because of the processor, it was because of the chipset. I know that sounds weird, but its true. Its been a long time since Intel did anything really original in the CPU arena besides boosting the clock speed(which, in fact I would probably take a performance hit here if I did upgrade), boosting the number of cores(which is not terribly useful due to a lack of software support), or creating new, ever-more confusing ways of pseudo-product segmentation.

        Fortunately though, PCIE 4.0 is coming, as well as USB 3.1… those 2 things alone offer enough incentive for a lot of people to seriously consider upgrading.

          • UnfriendlyFire
          • 5 years ago

          NVM Express or M.2 connector. No more SSD bottleneck.

          • the
          • 5 years ago

          I’m one of those Sandy Bridge owners myself. I will say that I’ve sat out Ivy Bridge and Haswell as there hasn’t been much overall improvement on the desktop. Sure Haswell is faster per clock but seemingly doesn’t clock as high. I’m hoping Sky Lake provides some benefit per clock and still at Haswell’s overall clock speed.

          Though I get where you are coming from with the chipset though. Z97 is interesting to me for NVMe support. SSD’s are fast to the point in day to day activities there isn’t much difference between models. I hope NVMe changes that or at the very least provides the feeling of something faster than a SATA based SSD.

          PCIe 4.0 is nice but more PCIe lanes are better. I’m hoping that Intel increases the DMI speed and increases the number of lanes on the chipset, even at 3.0 speeds. SATAe is something that needs raw lane count. If Intel or a 3rd party figures out how to do hardware RAID over PCIe, that’s worthy of an upgrade in my book.

          USB 3.1 is neat but I hope that either AMD and/or nVidia start to integrate USB 3.1 into there GPU’s and put Type C connectors on their GPUs. So even with an older system, I dream of an awesome upgrade path this way. Seriously, why not?

            • VincentHanna
            • 5 years ago

            [quote<]PCIe 4.0 is nice but more PCIe lanes are better.[/quote<] More PCIe lanes is more expensive for the Mobo mfg(and therefore the consumer) Its also literally 2x as fast as PCIe 3.1, meaning that 8x pcie 4.0 = 16x pcie 3.0 and 16x pcie 4.0 = 32x pcie 3.0. If you amicable to compromise though, how about more + faster? ^_)^

            • the
            • 5 years ago

            The thing is that even with double the bandwidth, you still need one or two lanes per peripheral. Two lanes per SATAe minimum. What PCIe 4.0 does in this situation is let SATAe and M.2 go from 4 lanes to 2 lanes while retaining bandwidth. The catch is that if you want multiple devices in a system you’ll quickly run out of lanes. Say a dual GPU setup (8x each) and two drive RAID 0 array using SATAe on a motherboard with USB 3.1 and 10G Ethernet. That’d be more lanes than what the current socket 1150 and Z97 chipset can provide today even if all lanes were capable of PCIe 4.0 bandwidth. Ultimately including some of these IO functions on-die or in the chipset (USB 3.1 etc.) would help with the lane deficiency. Similarly throwing a few more onto the chipset and boosting DMI link speed also resolve the issue.

            Socket 2011(-3) solves this problem by including far more lanes integrated into the CPU socket. What’d I’d like to see is Intel release a companion PCIe controller chip that links to socket 2011(-3) via QPI that just provides raw extra lanes. Though the only real consumer usage would be for multiple GPU setups. (Servers on the other hand have more options to use that extra IO bandwidth.)

            Although I like the compromise of [url=https://www.youtube.com/watch?v=221AWmsuX84<]more and faster.[/url<]

          • MDBT
          • 5 years ago

          I also feel that a chipset feature is more likely to get me to upgrade than another incremental performance boost on the CPU.

          Unfortunately the Sunrise point 10 series chipset set to debut with skylake will have neither PCI-E 4.0 nor USB 3.1 (natively). I personally would find value in additional cores but Skylake-S/K won’t be offering anything we haven’t already seen in Sandy/Ivy/Haswell/Broadwell.

          SSD bottlenecks are largely irrelevant to me since I can’t even tell the difference between my SATA II SSD and my SATA III SSD anyway. I’ll try an M2 now that I have a slot but I’m going to do so out of convenience of the form factor, not an expectation of noticeable speed improvement.

            • the
            • 5 years ago

            The move from SATA II to SATA III was bandwidth increase mostly. Unless you’re dealing with large files all the time, the bandwidth increase isn’t going to be that noticeable in day to day usage. The main benefit SSD’s brought over traditional hard drives as the latency drop to perform operations by several orders of magnitude. Even at SATA I bandwidth or even Ultra ATA/100, the latency difference is very noticeable.

            That’s why I have high hopes for NVMe as it addresses the latency aspect again. Removing the legacy storage stack with something thin built around the PCI software stack and optimized for SSDs will do wonders. Granted that current SSD’s already have low latencies but this is a big enough change I’m optimistic that I can feel the difference in regular day-to-day usage.

            • MDBT
            • 5 years ago

            That certainly explains why I see what I do and the virtue of at least one interface change. Thanks for the clarification!

Pin It on Pinterest

Share This