I plan on switching from AMD to Intel CPU. Skylake with its new socket seems most 'future proof' for the near-term.

As much as I want to replace my aging motherboard, I'm going to wait and see if Skylake is worth the cost of new ram, CPU, and motherboard.

I may end up doing a Skylake build, just because o want something new to play with! Been on Z77 for 3 years now! It's not like anything new is coming anytime soon after Skylake, kaby lake and canonlake! All the same architecture just the latter will be 10nm. Won't see a new architecture until 2018 no doubt! Who knows perhaps a new arch on 10nm will bring 8 cores to the mainstream in 2018!

I may end up doing a Skylake build, just because o want something new to play with! Been on Z77 for 3 years now! It's not like anything new is coming anytime soon after Skylake, kaby lake and canonlake! All the same architecture just the latter will be 10nm. Won't see a new architecture until 2018 no doubt! Who knows perhaps a new arch on 10nm will bring 8 cores to the mainstream in 2018!

Something new to play with, that's always the thing that gets me too, although I managed to resist until now probably - but like the previous poster before you, I too am waiting to see what Z170 and Skylake can do, but like you, I'd like something new, rather the motherboard and features than anything else, because I've been on this platform for 4.5 years now and it's just...getting old. And let's face it, my current motherboard isn't exactly top of the line, it's okay, but not all it can be. And if Kaby Lake will bring something amazing (doubt it) one can always chuck one of those in down the line. I don't think we'll see mainstream over 4 cores for a long time; what will Intel do for their E platform then? 12 cores maybe, come to think of it 🙂 But yeah, it would be nice to have 6 or 8 cores for the mainstream lineup.

Well -- a few things. I don't think AMD could have released a card without HBM that's competitive with Nvidia. It's not like they would have had a new architecture if they pushed HBM back a year. If they had gone GDDR5, you would have got a Fury X Fiji chip, with a 315w TDP instead of 275 and it would have performed slightly worse. I don't think HBM will be overclockable at all in the near future. Partner companies like EVGA/MSI/ASUS/Etc all recycle GPU cores for RMA purposes if a memory module gets burned out. I haven't looked recently but I know in the past that GDDR DRAM failure is like the highest cause of RMA problems with cards. Now that HBM is integrated onto the interposer, you can no longer recycle the chip if it fails. I also think that due to the nature of manufacturing HBM with over 10K micro connections per stack, the voltage tolerances are probably a lot more strict vs the more robust GDDR. Couple all this with how fast HBM is, I don't even think overclocking is necessary. I don't know about what the boxes say but actual bandwidth numbers are actually even lower than 480GB/s. http://techreport.com/r.x/radeon-r9-fury-x/b3d-bandwidth.gif Current maximum bandwidth is actually only 387GB/s. I'm not sure if that's due to the ROP count or some other DX11 limitation, random texture is still higher than Nvidia, so I personally think the ROP count is interfering here. Regardless it's not coming close to the theoretical maximum on HBMv1. But again, all this being said, I don't think AMD had a choice. They couldn't have released a competitive GDDR5 card. They might as well cut their losses as much as possible, get the move to HBM out of the way, get some experience with it and move forward with a die shrink + hopefully new architecture next year.

Not quite on topic, but a decent read still. As for the matter of the new CPUs and chipsets, I hope you guys won't be disappointed. I'm not sure what extra features you guys are looking for, but then again, the older your computer gets, the more you want to go for a new one, I know that feeling! Just don't expect your new hardware to do magic that Haswell (-E) can't do already, because iirc, there have been benches showing that the 6700K isn't that much faster than a 4790K...

Maybe it's like cars, you get a new one that isn't necessarily better but you just want a new one after so many years?

Well -- a few things. I don't think AMD could have released a card without HBM that's competitive with Nvidia. It's not like they would have had a new architecture if they pushed HBM back a year. If they had gone GDDR5, you would have got a Fury X Fiji chip, with a 315w TDP instead of 275 and it would have performed slightly worse. I don't think HBM will be overclockable at all in the near future. Partner companies like EVGA/MSI/ASUS/Etc all recycle GPU cores for RMA purposes if a memory module gets burned out. I haven't looked recently but I know in the past that GDDR DRAM failure is like the highest cause of RMA problems with cards. Now that HBM is integrated onto the interposer, you can no longer recycle the chip if it fails. I also think that due to the nature of manufacturing HBM with over 10K micro connections per stack, the voltage tolerances are probably a lot more strict vs the more robust GDDR. Couple all this with how fast HBM is, I don't even think overclocking is necessary. I don't know about what the boxes say but actual bandwidth numbers are actually even lower than 480GB/s. http://techreport.com/r.x/radeon-r9-fury-x/b3d-bandwidth.gif Current maximum bandwidth is actually only 387GB/s. I'm not sure if that's due to the ROP count or some other DX11 limitation, random texture is still higher than Nvidia, so I personally think the ROP count is interfering here. Regardless it's not coming close to the theoretical maximum on HBMv1. But again, all this being said, I don't think AMD had a choice. They couldn't have released a competitive GDDR5 card. They might as well cut their losses as much as possible, get the move to HBM out of the way, get some experience with it and move forward with a die shrink + hopefully new architecture next year.

You do have a point and I think you're right overall. From a business point of view it probably made more sense. But on the long term, that card could soon get smashed to pieces in 4k considering games are growing hungrier in terms of VRAM. Not sure when, but I have a feeling it's going to happen quite soon. We already have examples of greedy games like SOM. And oh my, that memory bandwidth is quite lower than I expected. Like you said, the ROP count is a good candidate for the limited performance. I had a feeling something was going on with the memory bandwidth ever since I first saw the modified specs from 640GB/s to 512GB/s.

Not quite on topic, but a decent read still. As for the matter of the new CPUs and chipsets, I hope you guys won't be disappointed. I'm not sure what extra features you guys are looking for, but then again, the older your computer gets, the more you want to go for a new one, I know that feeling! Just don't expect your new hardware to do magic that Haswell (-E) can't do already, because iirc, there have been benches showing that the 6700K isn't that much faster than a 4790K...

While I agree, history has made me weary of leaked benches. Remember the Titan X/Fury X leaked benches a few months back? That's when all the 'titan killing' began and the hype escalated massively.

While I agree, history has made me weary of leaked benches. Remember the Titan X/Fury X leaked benches a few months back? That's when all the 'titan killing' began and the hype escalated massively.

Oh you're absolutely right, I don't take those benches for being 100% correct. But they were disappointing, and somehow I don't see Skylake being much faster compared to Haswell, than Haswell compared to Ivy Bridge, no 50% performance boost. That's what I meant with my comments, that people should not expect too much of Skylake just because it's fabricated in a different node.

The only reason for me to switch for Skylake would be UEFI PCI-exp SSD support (system boot drive). However maybe better GPU can put some pressure on my 2500K.

Oh you're absolutely right, I don't take those benches for being 100% correct. But they were disappointing, and somehow I don't see Skylake being much faster compared to Haswell, than Haswell compared to Ivy Bridge, no 50% performance boost. That's what I meant with my comments, that people should not expect too much of Skylake just because it's fabricated in a different node.

Yeah I know what you mean. We'll just have to wait for AMD Zen. If it's somehow going to outperform current processors I can bet Intel is going to perform another Conroe and everyone will be happy because competition will flourish.

Yeah I know what you mean. We'll just have to wait for AMD Zen. If it's somehow going to outperform current processors I can bet Intel is going to perform another Conroe and everyone will be happy because competition will flourish.

Hehe pretty much so 😀 And AMD needs a winning product these days, more than ever I'm afraid...

Something new to play with, that's always the thing that gets me too, although I managed to resist until now probably - but like the previous poster before you, I too am waiting to see what Z170 and Skylake can do, but like you, I'd like something new, rather the motherboard and features than anything else, because I've been on this platform for 4.5 years now and it's just...getting old. And let's face it, my current motherboard isn't exactly top of the line, it's okay, but not all it can be. And if Kaby Lake will bring something amazing (doubt it) one can always chuck one of those in down the line. I don't think we'll see mainstream over 4 cores for a long time; what will Intel do for their E platform then? 12 cores maybe, come to think of it 🙂 But yeah, it would be nice to have 6 or 8 cores for the mainstream lineup.

Yeah, it gets me sometimes too lol I really don't know why people seem to need to justify building a new rig sometimes, if you want a new toy to play with and have the disposable income just do it lol Can my current i5 3570K do everything I want, yes, will a Skylake 6700K massively improve my gaming experience, no, do I care? no lol I want a new toy to play with. I think your right with your car analogy, once you've achieved what you want with it and had it a while you want a new project right, I see nothing wrong with that. You just always seem to see people moaning about IPC and how their Haswell or Devils Canyon chip will last a few more years blah blah blah, well obviously it will lol If you don't have the cash laying around then yeah you've got to be more frugal, how frugal I guess depends your personal set of circumstances i.e Wife, Kids, Mortgage etc I just don't see the point in people going on when they run a modern system and all they generally do is game and they moan about 10-15% IPC improvement when what they already have does the job just great lol Its like they seem to think that Intel thinks everyone wants to upgrade every generation or something lol No I don't think we'll see 8 cores soon, however 2018 is 2.5 years away, thats a long time in Computers. That should be the year we see a brand new Architecture on 10nm (not Canonlake) hoping that could be when mainstream gets more cores. I mean can you seriously see in 2018 people are still buying quad cores that have been around since like 2007!

I'd go for a X99/5820K build now over waiting for another quad core with Skylake tbh.

I heard 5820k not that good for gaming as it's a waste? Also what do they overclock like compared to 2600/2700k @ 4.8ghz?

The people who have outright told you that are a little misguided. The 5820k is a six core, it'll clock to 4.6ghz + without much of a hassle, assuming you have adequate cooling. The per core performance is a tiny bit worse than Devils Canyon, but nothing to write home about. The fact is, those extra two cores are going to give your platform a much longer lifespan and considerably better performance in games that support those extra cores. That's going to become more common with DX12. If you're looking for a platform update I'd say that going from a quad to a quad is a waste, at least when you're running Sandy + level processors. I'm guessing the main reason people were arguing against the 5820K was the price, which was mainly down to DDR4. We've seen DDR4 prices drop, and the 5820K is a very well priced chip. You're not looking at a huge difference in cost between that and a DC or Skylake i7.

Hehe pretty much so 😀 And AMD needs a winning product these days, more than ever I'm afraid...

It doesn't matter much how well Zen performs if AMD can't get it to market on time and without over-hyping it. Zen could be leaps and bounds ahead of Skylake, but if it's launched 3-6 months late, it won't matter. If it's over-hyped (like AMD has a history of doing) then it won't matter either. AMD has to do 4 things near perfect. Launch on time, provide competitive performance, provide the marketed performance and price competitively. So far, AMD has failed at 3 of the 4 repeatedly. Competitive pricing is the only thing they've been consistent with.

Skylake looks cool and all that but what i really want to see are the new Zen CPUs! C´mon AMD, show us you´re still alive...

The people who have outright told you that are a little misguided. The 5820k is a six core, it'll clock to 4.6ghz + without much of a hassle, assuming you have adequate cooling. The per core performance is a tiny bit worse than Devils Canyon, but nothing to write home about. The fact is, those extra two cores are going to give your platform a much longer lifespan and considerably better performance in games that support those extra cores. That's going to become more common with DX12. If you're looking for a platform update I'd say that going from a quad to a quad is a waste, at least when you're running Sandy + level processors. I'm guessing the main reason people were arguing against the 5820K was the price, which was mainly down to DDR4. We've seen DDR4 prices drop, and the 5820K is a very well priced chip. You're not looking at a huge difference in cost between that and a DC or Skylake i7.

Thanks for the responce dude. Are we going to see 6-8 core Skylake?

Not for awhile at least, that's one of the things I think a lot of people were hoping for. You'll need to wait for Skylake-E before you see more than four cores, which is again a higher end platform. Considering the performance difference, if you're hankering for an upgrade you might as we just go for X99 now and get it over with.

Well I have just been looking up reviews on Overclockers and Scan and people were saying they don't overclock very well, that they only overclock to 4.1-4.2ghz. I only every use the BIOS inbuilt overclocking menu. Do you think 4.1 would be the equivalent of 4.6 as my Sandy is?