updateBroadwell-E Supports Intel Turbo Boost Technology 3.0

[NTMBK]

Well broadwell-C isn't Broadwell-E.

I've been saying this for months.

WWYBYWB

 

[Arachnotronic]

Haha. I think the HEDT line being a core+ behind definately makes the decision even harder for everyone. You can have more cores, cache and bandwidth. Or the newest core, on the newest platform with the newest tech. Not both! F'n Intel!

I think things would be way more cut and dry if we HEDT and DT came out in the same quarter or relative time frame. There would be no hindsight of "i dont have the newest uArch, did I make the right decision?" or "I got the latest and greatest, but should I have got more cores?"

I wonder if Intel does this to us on purpos so that the DT chips will always have a value being a generation ahead.





Yea..I was almost with you on that. Decided to go all in on BDWE

Sadly I think this is a case of Intel not being able to release these chips (DT and HEDT) at the same time. HEDT is based on server processors which have entirely different uncores and, soon, will have different, more complex CPU cores as well. This means validation time is much longer and so HEDT will always be behind DT in terms of CPU core generation.

The one exception we had in recent years was HSW-E and HSW, but this was only because BDW never (really) came to desktops in the first place.

 

[frozentundra123456]

Sadly I think this is a case of Intel not being able to release these chips (DT and HEDT) at the same time. HEDT is based on server processors which have entirely different uncores and, soon, will have different, more complex CPU cores as well. This means validation time is much longer and so HEDT will always be behind DT in terms of CPU core generation.

The one exception we had in recent years was HSW-E and HSW, but this was only because BDW never (really) came to desktops in the first place.

I dont see the problem. Just design a mainstream desktop chip with six cores. Certainly should be plenty of die space at 14nm.

 

[Arachnotronic]

I dont see the problem. Just design a mainstream desktop chip with six cores. Certainly should be plenty of die space at 14nm.

It's not that easy as this would require additional engineering resources. Why do this when you can recycle a server chip to do the same job?

The perf/clock difference isn't huge between Intel CPU cores as some are so fond of pointing out, so if somebody really wants more cores, they can get 50%+ more cores, 50% more memory bandwidth, etc. for a trade-off of maybe 5-10% in perf/clock.

If you genuinely need/want the additional cores, this is a no brainer trade-off. If 5-10% IPC difference is going to kill you, then you obviously don't have many workloads that benefit from the additional MT performance and would be better suited to a higher clocked, higher IPC quad.

 

[frozentundra123456]

I see your point, but it would be nice if Intel just made the best possible product instead of segmenting everything to death to try to eke out the last bit of margin. Plus it would be on the latest platform and would offer an upgrade down the line from dual or quad core chips.


BTW, I used to "debate" this strongly with Shintai. Where is he these days, see very few posts from him.

 

[Arachnotronic]

I see your point, but it would be nice if Intel just made the best possible product instead of segmenting everything to death to try to eke out the last bit of margin. Plus it would be on the latest platform and would offer an upgrade down the line from dual or quad core chips.

It's not even about that though, I am sure that Intel could price a mainstream hex core for much higher than a 4+2 and per-device margins would be lovely.

The problem is that there are significant R&D costs that go into designing, taping out, validating, etc. a new die. In order to justify that additional work, the company needs to be assured that it'll be able to sell enough of those chips to justify putting their engineers to work on building that chip rather than working on another project (future mainstream CPUs, future server CPUs, etc.)

The problem with a hex core mainstream chip is that the market for it would be very limited. It would be a no-go in all-in-one PCs, no-go in high performance laptops, and even most tower desktop buyers wouldn't be interested. It really would be a chip designed solely for people who typically buy 5820K-class products, and it's really hard to recoup investment on a chip so incredible niche.

The HEDT chips have a known, high volume market in servers/workstations, so Intel doesn't have to worry about getting a good return on those. The 4+2 chips are used all over the place, so it doesn't take much to cherry pick the very best ones and sell them as 6700K.

Anyway, a mainstream 6+2 die would just make little economic sense from an R&D/manpower perspective.

Now I know what you'll say next...AMD has an 8C/16T Zen coming soon...this is true, but do you think this die was designed solely for HEDT? I can almost guarantee you that it is a server die that's mounted onto the same package as a Bristol Ridge APU.

BTW, I used to "debate" this strongly with Shintai. Where is he these days, see very few posts from him.

On ban. He'll be back once the ban lifts.

 

[jhu]

On ban. He'll be back once the ban lifts.

Again?

 

[CHADBOGA]

Again?

He is getting banned for the most trivial things imaginable.

Horrible moderating.

This post doesn't belong here and you calling out and baiting the mods will only just give you a vacation.

-Rvenger

 

[Sweepr]

He is getting banned for the most trivial things imaginable.

Horrible moderating.

IMHO this is quickly turning into another SA with one-sided opinions. Just look at the number of 2016 accounts with a very similar posting pattern, that already tells a lot. I wonder to myself if mods don't see these posts?

I don't get why the Pascal Thread was moved to a subforum that no one visits either. I mean, we have 13 Pascal related thread @ VC&G (first page) now, so the objective wasn't to keep the news in one place.

 

[2blzd]

Can wait for Turbo 3.0!

 

[DrMrLordX]

Seems like Shintai is already back, ban lifted, or am I wrong? I dunno. People have to be careful in the VC&G subforums, I've seen people get warned/infracted in the AMD subforum for bringing up Nvidia. It's just the way it is.

Perhaps CPUs & Overclocking will become similarly-fragmented. Kind of a shame, but it was a necessary evil over on the VC&G side.

So, it's interesting to see how many differences there are between Broadwell-E and Broadwell-C (eDRAM and iGPU notwithstanding). What other tricks does Intel have up their sleeve besides MSR199?

 

[2blzd]

Seems like Shintai is already back, ban lifted, or am I wrong? I dunno. People have to be careful in the VC&G subforums, I've seen people get warned/infracted in the AMD subforum for bringing up Nvidia. It's just the way it is.

Perhaps CPUs & Overclocking will become similarly-fragmented. Kind of a shame, but it was a necessary evil over on the VC&G side.

So, it's interesting to see how many differences there are between Broadwell-E and Broadwell-C (eDRAM and iGPU notwithstanding). What other tricks does Intel have up their sleeve besides MSR199?

Most people don't realize that when Intel delayed the chip, they continued to work on and improve the core. It's not like Intel took the Broadwell-C core and plopped into an HEDT uncore (I feel like most people really thinks this is what happens), and said wala we're done!....then sat there for 11 months, idly waiting for yields to improve. Far from the truth. FUGGER over at Xstreme systems hinted at this in his talks with Intel last august. The extra time let them get the 10 core sku in for consumers (yields) and also develop "turbo 3.0". He also mentioned, at the time (august 2015), they were working on making the turbo "3.0" backward compatible with all its modern cores, including the already released skylake dt.

 

[Arachnotronic]

Most people don't realize that when Intel delayed the chip, they continued to work on and improve the core. It's not like Intel took the Broadwell-C core and plopped into an HEDT uncore (I feel like most people really thinks this is what happens), and said wala we're done!....then sat there for 11 months, idly waiting for yields to improve. Far from the truth. FUGGER over at Xstreme systems hinted at this in his talks with Intel last august. The extra time let them get the 10 core sku in for consumers (yields) and also develop "turbo 3.0". He also mentioned, at the time (august 2015), they were working on making the turbo "3.0" backward compatible with all its modern cores, including the already released skylake dt.

Interesting. I would imagine Turbo is done via firmware, so updating older CPUs to include this feature may be viable.

 

[frozentundra123456]

Seems like Shintai is already back, ban lifted, or am I wrong? I dunno. People have to be careful in the VC&G subforums, I've seen people get warned/infracted in the AMD subforum for bringing up Nvidia. It's just the way it is.

Perhaps CPUs & Overclocking will become similarly-fragmented. Kind of a shame, but it was a necessary evil over on the VC&G side.

So, it's interesting to see how many differences there are between Broadwell-E and Broadwell-C (eDRAM and iGPU notwithstanding). What other tricks does Intel have up their sleeve besides MSR199?

I have to pretty much avoid the VC and G forum. I get angry every time I go there. I dont really see the point of splitting it up into subforms. The general VC and G forum still seems pretty much like it was before. Splitting it up certainly didnt make the general forum any less partisan. Cant imagine what the AMD forum is like.

 

[Azuma Hazuki]

Caaaareful guys, that all looks suspiciously like the dreaded Mod Callout ()

 

[zentan]

IMHO this is quickly turning into another SA with one-sided opinions. Just look at the number of 2016 accounts with a very similar posting pattern, that already tells a lot. I wonder to myself if mods don't see these posts?

I don't get why the Pascal Thread was moved to a subforum that no one visits either. I mean, we have 13 Pascal related thread @ VC&G (first page) now, so the objective wasn't to keep the news in one place.

Yep,can't understand the reason why a new architecture GPU thread was moved to that subforum. Also for some reason the old VC&G mod Stahlhart is not active it seems.

 

[zentan]

Since they moved from 22nm to 14nm with same platform tdp envelope of 140W and also similar SKUs apart from 10c/20t one it was very expected that there would be a better turbo implementation or higher full load all core turbo clocks.

 

[2blzd]

Sounds like a gimmick to get better scores in single threaded benchmarks.

If you've ever had more than 4 cores you would know a lot of them aren't utilized very well, if at all. With this you can run/park apps directly on one core/whatever core you want, choose the clock speed and move on. This helps you avoid idle cores and truly maximize your ability to MT.

For example: On a hexcore - running excel on 2 cores @ 2ghz while gaming on 4 cores @ 3.5ghz.

Or whatever you want to imagine.

Just after talking to a few people who were at IDF in august, it seemed there was a general excitement about BD-E and that it wasn't just another HEDT iteration. Who knows if that actually will translate into retail product, knowing Intel's ability to crush dreams, but the fact that we got a 10core sku after the many roadmaps that showed the same 6/6/8 setup, is promising. I hope it overclocks as well as some have lead on.

but who knows, it could really just be another BLEHEDT chip.

 

[TheRyuu]

If you've ever had more than 4 cores you would know a lot of them aren't utilized very well, if at all. With this you can run/park apps directly on one core/whatever core you want, choose the clock speed and move on. This helps you avoid idle cores and truly maximize your ability to MT.

For example: On a hexcore - running excel on 2 cores @ 2ghz while gaming on 4 cores @ 3.5ghz.

Or whatever you want to imagine.

Just after talking to a few people who were at IDF in august, it seemed there was a general excitement about BD-E and that it wasn't just another HEDT iteration. Who knows if that actually will translate into retail product, knowing Intel's ability to crush dreams, but the fact that we got a 10core sku after the many roadmaps that showed the same 6/6/8 setup, is promising. I hope it overclocks as well as some have lead on.

but who knows, it could really just be another BLEHEDT chip.

Yea but requiring manual intervention at the OS level from Intel's program/service/driver makes the situation seem a bit weird to me. If they have the capability to do dynamic frequency control on an individual core basis why is that not done automatically? Just let the OS scheduler do its job and have it turbo boost whatever cores are being used.

To me this almost sounds like the technology simply wasn't mature enough to do that. I would have to think they would want to make this kind of frequency scaling automatic if they can so I'm not sure this will last for more than this generation if it actually becomes a thing. We already see the power/frequency control moving to the CPU in Skylake.

 

[ShintaiDK]

Intel and Microsoft works very closely together in such matters. Skylake is a prime example of this. So I wouldn't worry.

 

[2blzd]

This is speculation of what little info I pieced together over the year and trying to match it up with the final PR/marketing terms. I could be completely wrong for all I know. Its just once CSBIN posted the description of turbo 3.0 everything just started to click and make sense. But who knows

 

[Markfw]

IMHO this is quickly turning into another SA with one-sided opinions. Just look at the number of 2016 accounts with a very similar posting pattern, that already tells a lot. I wonder to myself if mods don't see these posts?

I don't get why the Pascal Thread was moved to a subforum that no one visits either. I mean, we have 13 Pascal related thread @ VC&G (first page) now, so the objective wasn't to keep the news in one place.

Just to enlighten everyone, a few tips.

First, when posters "appear" to have an agenda, we can't guess, assume or accuse without proof.

As for the subforum, VC&G became a horrible place with nothing but fanboys, flamefests, and troll posts. So there was an attempt to fix that by giving everyone a subforum where they could feel safe from the "other side", and now there are 3 places for this to happen.

And last, VC&G mods come and go fast, as the community makes it so hard to moderate, they quit because of all of the above. (we are volunteers, not paid)

As to why the thread got moved (just a guess) was if its there, you can't talk about AMD, so that cuts down on the trolls that must have been ruining the thread. I don't even know which one you are talking about,

And to everyone else, lets get this thread back on topic, or I will lock it, this is the CPU forum.

 

[Unoid]

He is getting banned for the most trivial things imaginable.

Horrible moderating.

This post doesn't belong here and you calling out and baiting the mods will only just give you a vacation.

-Rvenger

In my opinion, it's great moderating, He's the only one on my ignore list, I've thought about including you too.

too many intel/nvidia trollbaiters doing this on purpose.

 

[frozentundra123456]

If you've ever had more than 4 cores you would know a lot of them aren't utilized very well, if at all. With this you can run/park apps directly on one core/whatever core you want, choose the clock speed and move on. This helps you avoid idle cores and truly maximize your ability to MT.

For example: On a hexcore - running excel on 2 cores @ 2ghz while gaming on 4 cores @ 3.5ghz.

Or whatever you want to imagine.

Just after talking to a few people who were at IDF in august, it seemed there was a general excitement about BD-E and that it wasn't just another HEDT iteration. Who knows if that actually will translate into retail product, knowing Intel's ability to crush dreams, but the fact that we got a 10core sku after the many roadmaps that showed the same 6/6/8 setup, is promising. I hope it overclocks as well as some have lead on.
but who knows, it could really just be another BLEHEDT chip.

Well, actually, it was meant primarily in jest. I understand the purpose, but the problem is that most BW-E chips will be overclocked anyway, so it will show little if any benefit. Could be helpful I suppose in productivity apps when one does not want to overclock, and for servers of course.

As far as being excited for BW-E, I wish I could be but just cant. Only thing even remotely exciting about desktop Broadwell was the edram and IGP performance, neither of which will apply to HEDT. Actually, nothing exciting has come on 14nm for desktop. I mean Skylake is OK, but after all the modest improvements, I was hoping for a 15% or more performance improvement with Skylake. I know they have supposedly been working on the process and uncore, but I am expecting little to zero improvements except for the 10 core, if one can use the cores.

 

[krumme]

Just to enlighten everyone, a few tips.

First, when posters "appear" to have an agenda, we can't guess, assume or accuse without proof.

As for the subforum, VC&G became a horrible place with nothing but fanboys, flamefests, and troll posts. So there was an attempt to fix that by giving everyone a subforum where they could feel safe from the "other side", and now there are 3 places for this to happen.

And last, VC&G mods come and go fast, as the community makes it so hard to moderate, they quit because of all of the above. (we are volunteers, not paid)

As to why the thread got moved (just a guess) was if its there, you can't talk about AMD, so that cuts down on the trolls that must have been ruining the thread. I don't even know which one you are talking about,

And to everyone else, lets get this thread back on topic, or I will lock it, this is the CPU forum.

Keep up the hard and good work guys. Its a tough place but without you it would be a war zone. Good you are here to remind us when we go flaming. Thumbs up.