Terbinator
Member
You need to sell to make profits and no one is going to buy if your top-end CPU is rocking up with IPC from 5 years ago and costs over £250.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Core i7-6950X (Broadwell-E) to be first Intel consumer CPU with 10 cores (rumor)
- Thread starter Durante
- Start date
You need to sell to make profits and no one is going to buy if your top-end CPU is rocking up with IPC from 5 years ago and costs over £250.
dr_rus
Member
That's why I hope that Zen will deliever - not only to force Intel to release something good for a reasonable price but to help AMD to get some money from CPUs.
I thought about upgrading to both of those platforms last year, but ultimately decided to stick with my 2500K a little longer. I have been itching to upgrade for a long time, but I don't like throwing money down the drain. Most of the things I use my computer for are primarily bound by single-threaded performance, and the CPU doesn't seem to matter much in the vast majority of games yet unless you want to play at console-level resolutions and > 120 FPS.
Instead I have focused on upgrading the parts of my system where it would make a tangible difference for my use - the GPU, displays, and my computer chair.
Caayn
Member
https://downloadcenter.intel.com/pr...sor-Extreme-Edition-25M-Cache-up-to-3-50-GHz-
Support page for the cpu has gone up.
25MB cache and a clock speed of 3.5GHz are basically confirmed now. Nothing about core count yet.
Support page for the cpu has gone up.
25MB cache and a clock speed of 3.5GHz are basically confirmed now. Nothing about core count yet.
frontieruk
Member
http://www.hardwarebbq.com/intel-reveals-i7-6950x-hedt-extreme-edition-cpu-info/ relevant?
edit ...Same info as above with speculation
Intel has updated its Management Engine software which is now available for download. As its pointed out in the changelog, this new version also works with a particular upcoming HEDT processor- a Broadwell-E based Intel i7-6950X Extreme edition processor, along with a couple of information about this CPU.
Read more at: http://www.hardwarebbq.com/intel-reveals-i7-6950x-hedt-extreme-edition-cpu-info/
edit ...Same info as above with speculation
Intel has updated its Management Engine software which is now available for download. As its pointed out in the changelog, this new version also works with a particular upcoming HEDT processor- a Broadwell-E based Intel i7-6950X Extreme edition processor, along with a couple of information about this CPU.
Read more at: http://www.hardwarebbq.com/intel-reveals-i7-6950x-hedt-extreme-edition-cpu-info/
dr_rus
Member
Same as 5960X but with two additional cores. Should be a good CPU for those who need them cores. A faster clocked 6 core one will probably be a better option for gaming.
Projectjustice
Banned
Im waiting for Zen. AMD always does very well with multi-core CPUs.
AMD CPUs are great for tight budgets, but you're otherwise much better off going with an i5 or i7.
dr_rus
Member
Rumors point to Computex launch, yeah, so - in about a month from now.
I'm struggling to think of a time when they've actually did very well with multi-core CPUs - I think it was back in Athlon 64 X2 going against the last P4s ten or so years ago? Anyway, I'm willing to give Zen a benefit of the doubt and really hope that it will be on Skylake level in core-for-core performance.
I know I'm responding to these posts a good half a year late, but as SimonChiu has already bumped the post, I may as well give my two cents if Blu and Durante are still paying attention.
I would be willing to wager good money that this is all about yields. Basically, if your interconnect topology is of dimension greater than one, it's impossible to disable a single core in a way which leaves the chip topologically consistent (i.e. in a way which is invisible to software). A ring is a 1-dimensional torus, which makes it the lowest-latency topology where cores can be disabled individually to improve yields. If you go up to a 2-dimensional topology, say a simple 2D mesh (e.g. Epiphany) or a 2D torus (e.g. KNL) then you have to disable a full row or column of cores when a single core fails, and likely both a row and a column if two cores fail.
When you're producing large dies on immature low-yield processes, like Intel do with their big Xeons, maximising the number of usable dies per wafer, and maximising the value of those dies, is probably going to be more important than implementing the most efficient interconnect. To take a simple example, consider a 16-core Xeon die implemented either with a ring bus or a 2D torus. With a single faulty core, they can sell the ring-bus version as a 15-core chip, and the 2D torus version as a 12-core chip. With two faults, that reduces to 14-core and 9-core respectively. Three faults, 13-core and 6-core, etc. Intel can make more money selling higher core-count chips with less efficient interconnects than vice versa, so that's what they'll do.
Xeon Phi is different, both because the large number of cores make the interconnect bottleneck bigger, and because the large number of cores make the cost of disabling a full row or column proportionally lower (as it's the square root of the number of nodes). With Knights Landing they're going with a 6x6 torus of paired cores for 72 cores total on the full chip. I don't think they've talked about a binned version, but when they do I'm guessing it's going to have one row and one column disabled for 50 usable cores, allowing them to sell dies with two faulty nodes (or more if they're lucky with their relative positions).
The move to a more yield-sensitive topology may well partly explain why we're seeing KNL arrive so late. The 22nm KNC was fully available only around 6 months after the first 22nm Ivy Bridge CPUs, yet we're now around a year and a half after the first 14nm Broadwell chips, and we've only just seen the availability of KNL-based development systems start to trickle out. Of course, Intel's general yield issues on 14nm certainly won't be helping.
I would be willing to wager good money that this is all about yields. Basically, if your interconnect topology is of dimension greater than one, it's impossible to disable a single core in a way which leaves the chip topologically consistent (i.e. in a way which is invisible to software). A ring is a 1-dimensional torus, which makes it the lowest-latency topology where cores can be disabled individually to improve yields. If you go up to a 2-dimensional topology, say a simple 2D mesh (e.g. Epiphany) or a 2D torus (e.g. KNL) then you have to disable a full row or column of cores when a single core fails, and likely both a row and a column if two cores fail.
When you're producing large dies on immature low-yield processes, like Intel do with their big Xeons, maximising the number of usable dies per wafer, and maximising the value of those dies, is probably going to be more important than implementing the most efficient interconnect. To take a simple example, consider a 16-core Xeon die implemented either with a ring bus or a 2D torus. With a single faulty core, they can sell the ring-bus version as a 15-core chip, and the 2D torus version as a 12-core chip. With two faults, that reduces to 14-core and 9-core respectively. Three faults, 13-core and 6-core, etc. Intel can make more money selling higher core-count chips with less efficient interconnects than vice versa, so that's what they'll do.
Xeon Phi is different, both because the large number of cores make the interconnect bottleneck bigger, and because the large number of cores make the cost of disabling a full row or column proportionally lower (as it's the square root of the number of nodes). With Knights Landing they're going with a 6x6 torus of paired cores for 72 cores total on the full chip. I don't think they've talked about a binned version, but when they do I'm guessing it's going to have one row and one column disabled for 50 usable cores, allowing them to sell dies with two faulty nodes (or more if they're lucky with their relative positions).
The move to a more yield-sensitive topology may well partly explain why we're seeing KNL arrive so late. The 22nm KNC was fully available only around 6 months after the first 22nm Ivy Bridge CPUs, yet we're now around a year and a half after the first 14nm Broadwell chips, and we've only just seen the availability of KNL-based development systems start to trickle out. Of course, Intel's general yield issues on 14nm certainly won't be helping.
Oh wow, so much for hoping the new 8 core part would come in at the 5930K's launch price.
Schnozberry
Member
I'll have to wait and see how these chips overclock. My 5820k hits 4ghz at 1.07v and has been a really good performer. I'm really struggling to see what the point of upgrading would be.
DouglasteR
Member
Wow,
So much more money for much less performance.
Hope AMD delivers.
So much more money for much less performance.
Hope AMD delivers.
Man the CPU and GPU markets both are so fucked right now.
Possumowner
Member
Can't wait to see the pricing
Damn them and their pricing.
OminoMichelin
Member
The price deltas are...impressive :lol +~50%, ~60%, ~70% for each subsequent tier.
I guess they feel confident they can squeeze as much as they want from the consumers and no one will complain.
I guess this means that the prices for older CPUs will only go up.
I guess they feel confident they can squeeze as much as they want from the consumers and no one will complain.
I guess this means that the prices for older CPUs will only go up.
Damn... that's a lot and I wanted to replace my 3930k.
BigTnaples
Todd Howard's Secret GAF Account
Nice. Would be a good step up from my 3770k no?
I'll probably upgrade my CPU and Ram sometime next year.
I'll probably upgrade my CPU and Ram sometime next year.
dr_rus
Member
We'll have Skylake-E next year and I'm hoping that Zen will affect these prices. But I myself will probably go for 6850K in June as my current system is dying on me.
For a while the 5820k was a lot cheaper than even the 6700K. You could also go with the middle ground and look at Zen, it seems promising.
Schnozberry
Member
Plenty of overclock headroom on the 5820k. With Intel downplaying performance gains to focus on energy efficiency, Skylake-E probably won't be a panacea either.
"plenty" is relative. You can take the average 5820K up to 4.5Ghz but you have to give it 1.3V for it to be stable.
At that point that thing gets *hot*. Those Broadwell numbers on paper don't look at all impressive in comparison, but it is liable to overclock better at lower voltages.
I'm just hoping that Skylake-E will offer a 8c/16t configuration at the ~USD$600 price point (or better, but that's rather pie-in-the-sky).
I wish that were the case here, haha. I'd like to be even cautiously optimistic about Zen, but I'm expecting AMD to underwhelm yet again.
Schnozberry
Member
I guess it depends on how well you do with the silicon lottery. I OC conservatively at 4ghz and am able to do it 1.07v. I'm usually less than 40c on air doing normal desktop tasks.
Schnozberry
Member
If Zen is competitive and cost effective than we might get lucky and see prices driven downwards. The AMD/Nvidia monopoly on enthusiast level performance has been a real albatross for consumers.
blu
Wants the largest console games publisher to avoid Nintendo's platforms.
That's actually an interesting thought, and just like Durante I had never considered the economy of interconnects vs cores.
BTW (Durante should like this : ) did you know intel are selling these days an add-on board populated with 3x E3 Xeons as an ocl accelerator, and have named it.. *wait for it* Intel VCA (yes, the exact same acronym as NV's Visual Computing Appliance).
Yeah, I can get that on stock voltages, but at 4.4Ghz I need 1.16V to be stable. Past that I need to start ramping up a lot more to be stable.
lmimmfn
Member
IPC from 9 year old OCd Bloomfield is fine if you're not expecting 120FPS+ and you're on 1440p with a decent GPU, don't knock it
If AMD can achieve IPC of 5 years ago but with more cores, high overclock and cheaper price they're onto a winner. I'd love to see AMD get back in the game.
Lactose_Intolerant
Member
I hope they make it 8c too. I might be able to get my boss to pay for it.
Schnozberry
Member
Those are best processors. The ones other people pay for.
kruis
Exposing the sinister cartel of retailers who allow companies to pay for advertising space.
I had hoped that the new ten core CPU would push prices down for the 8 and 6 core CPUs. Obviously not the case. My plan was to upgrade from a 6 core 980x to an 8 core 6900K, but now I'm considering the cheaper 6850K. I'm equally disappointed by the prices of Nvidia's top offerings. I may postpone my whole PC upgrade and wait yet another year ...
Sir Abacus
Member
Oh oh oh. A new turbo boost mode! Let's hope it allows voltage constrained boosting!
Super Rookie
Member
Mines on 4.6GHz\1.3V at ~45C on air cooling with normal dektop usgae and ~60C while gaming or CPU intensive tasks.
I just let my ASUS motherboard auto clock it for me and that's what it ended up and it's super stable, no crashes yet.
Try running a power virus stress test with IBT or Prime95. It is probably stable, but you'll see your temps go through the roof.
Of course, that isn't a typical load, but those are the conditions that I'm calling "stable".
not even folding@home get that kind of heat development
it is nonsensical benchmark for heat that has no real world bearing
StereoVsn
Member
Well, this is more then I expected for certain. I will see what happens and might spring for 6850 toward the year end (Black Friday maybe). It would be over $1K for CPU, Motherboard, 16 or 32 GB RAM and might need new cooler. Let's hope Zen does well and pushes prices down (doubt it).
d00d3n
Member
What are the real world implications of 28 v. 40 pci express lanes? 28 will be fine for a single GPU and a pcie ssd, right?
I was actually curious to jump in on one of the higher tiers this time, but the two top models seem incredibly expensive. The 1100$ model can at least be considered, but it seems strange to pay so much and still not get "the best"
I was actually curious to jump in on one of the higher tiers this time, but the two top models seem incredibly expensive. The 1100$ model can at least be considered, but it seems strange to pay so much and still not get "the best"
dr_rus
Member
Yeah, if you don't want to use SLI/CF 28 lanes should be enough.