Broadwell/Skywell Release - New Intel Details

lazz

lazz

Limp Gawd
Joined
Apr 15, 2007
Messages
324
Hmm. If this is really true to me the only point of Broadwell-K is lga1150/ddr3 support.
 
95watt TDP looks promising for overclockers, looking forward to seeing reviews on this. It could potentially be my next upgrade even though I probably won't "need" one.

Intel should also IMO cut some costs on the K series by not including their crappy HSF. Only reason to own a K series is to overclock and if you're overclocking you aren't using the intel HSF.
 
^ agreed. I'd much rather see Intel sell the K sku's without the OEM HSF and put that money into soldered IHS's.
 
Broadwell-E....again enthusiasts get a neutered version of the smallest die (aka "the shaft")....

Thanks Intel...:rolleyes:
 
lutjens said:
Broadwell-E....again enthusiasts get a neutered version of the smallest die (aka "the shaft")....

Thanks Intel...:rolleyes:
Click to expand...


How do you figure that?

6-8 cores with HT'ing, fully unlocked, large Cache, good chance of solder under the IHS, potentially 40-PCIe lanes for all BDWL-E options, compatible with existing LGA-2011v3 MoBo's, etc...That means we lowly "mainstream" users of i5/i7-K aren't going to enjoy a lot of those features for another 3-7 years when they migrate over to that product segment.
 
These slides are legit.

Personal opinion though...I don't think Skylake is going to land according to this schedule. I think this roadamp is their idealized "hey I hope we can do this".
 
DejaWiz said:
How do you figure that?

6-8 cores with HT'ing, fully unlocked, large Cache, good chance of solder under the IHS, potentially 40-PCIe lanes for all BDWL-E options, compatible with existing LGA-2011v3 MoBo's, etc...That means we lowly "mainstream" users of i5/i7-K aren't going to enjoy a lot of those features for another 3-7 years when they migrate over to that product segment.
Click to expand...

To get those 6-8 cores, they'll use the smallest die size of Broadwell-EP (which will almost certainly have more cores than the 8 core die for Haswell-E). They'll neuter it down to make the 8 and 6 core i7s. Cache size likely won't change from what it is now. The PCI-E lane count of 40 lanes has been available since Nehalem was out, so there's nothing new or exceptional there either. All things considered, Broadwell-E users will get nothing but Intel's Xeon leftovers with ZERO truly high end options (which is exactly what we've been given since Sandy Bridge-E). This rates as "the shaft" in my book.

The largest die will be for Broadwell-EX and may be up to 24 cores (somewhere between 18 and 24). A nice chip with great potential, but again, they will be completely hard locked. Given that the current E5-2699 V3 (when overclocked by the measly 4 MHz that you can do with the locked BCLK straps) runs with a Vcore of 0.900V at nearly 3.0GHz, the incredible clock speed potential of these large die CPUs is obvious.

Bottom line...again there will be no unlocked processors available that have more than 8 cores, not for any price, unless they expand the E5-1600 V4 series to include chips with higher core counts (that they actually decide to sell to the channel). It wouldn't matter if the top Xeon SKUs for each core count were unlocked, as one could simply purchase the core count of their choice and then run it at the top speed the chip and the cooling used would support. But as they're all hard locked, again there will be no way to get a high core count chip that's unlocked, not for any price.

And that's a crying shame in my book...:(
 
drescherjm said:
I expect lutjens wants an overclockable 12+ physical core cpu
Click to expand...

If that is what he wants because he actually has a beneficial use for that many cores, then there are plenty of Xeon options available with up to 14 cores for the LGA2011-v3 platform, not to mention SMP capability to expand the number of cores even further. Not OC'able, but who really wants to risk system instability when running such processor-intensive tasks; or invest the money in what it takes to try and pull that much heat off of 14 OC'd single die cores?

I realize it would be fun and cool to have 10+ fully unlocked cores on an enthusiast platform, but the need doesn't outweigh the feasibility at this point, imo.
 
DejaWiz said:
If that is what he wants because he actually has a beneficial use for that many cores, then there are plenty of Xeon options available with up to 14 cores for the LGA2011-v3 platform, not to mention SMP capability to expand the number of cores even further. Not OC'able, but who really wants to risk system instability when running such processor-intensive tasks; or invest the money in what it takes to try and pull that much heat off of 14 OC'd single die cores?

I realize it would be fun and cool to have 10+ fully unlocked cores on an enthusiast platform, but the need doesn't outweigh the feasibility at this point, imo.
Click to expand...

What I'm saying is that people who pay for a high core count Xeon chip should be able to use the huge amount of headroom that's present in these chips. As I said above, the E5-2699 V3 when run at 105MHz bus speed (2940MHz) run at 0.9V and run quite cool. I know such a chip wouldn't match a chip with a lower core count in pure clock speed, but a modest clock speed between 3.5GHz and 4.0GHz with a slight voltage increase and should be easily obtainable under water without risking damage or instability.
 
lutjens said:
What I'm saying is that people who pay for a high core count Xeon chip should be able to use the huge amount of headroom that's present in these chips. As I said above, the E5-2699 V3 when run at 105MHz bus speed (2940MHz) run at 0.9V and run quite cool. I know such a chip wouldn't match a chip with a lower core count in pure clock speed, but a modest clock speed between 3.5GHz and 4.0GHz with a slight voltage increase and should be easily obtainable under water without risking damage or instability.
Click to expand...

a) People who buy these chips aren't interested in running them above spec. They're interested in 99.99999% uptime.

b) Power consumption doesn't increase linearly with voltage. I'd bet the same effects we see with 8 cores would probably be compounded with 18 cores. Instead of .25V being a huge jump, maybe .05V becomes a huge jump.

c) Heat dissipation doesn't scale linearly with core count. There's thermal transfer within the silicon, which is why cores sandwiched in the middle are usually running hotter than cores at the edges bordered by IO and cache and shit. At 18 cores you're talking there's entire columns of cores that are going to be sandwiched, and no heatsink is going to be able to pull the heat off fast enough.

d) The market for people who want to spend $6k for a cpu for personal use is non-existant. There's "enthusiasts" on here who bitch out about upgrading their 4 year old i7-920 to something newer, and that only costs a few hundred.
 
dr/owned said:
a) People who buy these chips aren't interested in running them above spec. They're interested in 99.99999% uptime.

b) Power consumption doesn't increase linearly with voltage. I'd bet the same effects we see with 8 cores would probably be compounded with 18 cores. Instead of .25V being a huge jump, maybe .05V becomes a huge jump.

c) Heat dissipation doesn't scale linearly with core count. There's thermal transfer within the silicon, which is why cores sandwiched in the middle are usually running hotter than cores at the edges bordered by IO and cache and shit. At 18 cores you're talking there's entire columns of cores that are going to be sandwiched, and no heatsink is going to be able to pull the heat off fast enough.

d) The market for people who want to spend $6k for a cpu for personal use is non-existant. There's "enthusiasts" on here who bitch out about upgrading their 4 year old i7-920 to something newer, and that only costs a few hundred.
Click to expand...


a) True...most people who buy an 18-core aren't going to overclock them. However, if the chips were unlocked, nothing would change for servers and workstations using them. They'd still run at default speed happily and default TDP. But people who want to use such a chip on an enthusiast platform and overclock it should be able to.

b and c) Again true, but the chip should still be capable of the overclock range I stated above. They're not going to match a 6 or 8 core CPU, but should still be capable of a decent increase when good water cooling is used.

d) The E5-2699 V3 is not $6K. Regardless, by unlocking the top SKU of Xeon, it would satisfy both the small ultra high-end market, as well as those who use them in the typical high-end server environment. One chip to serve both markets.
 
Skylake-K by the end of 2015? Sign me up, although I'm very doubtful.

The only good thing about Broadwell is that the growing pains(or shrinking pains in this case) of 14nm seem to have been sorted before the new architecture drops. They really should cancel Broadwell for desktops and just give us Skylake because the entire enthusiast community is waiting for it anyway.
 
Michaelius said:
Ehh for god's sake - why when they have broadwell chip in hand can't they make a test against Haswell at same clock speeds ?
Click to expand...

Probably some stipulation from Intel for getting a preview before release. Given that it actually used more power than the 4500u that it replaced(but was faster and gave better perf/watt).
 
so what are we looking at for these 14nm desktop Broadwell's in terms of speed? 6Ghz oc'd?
 
So Broadwell-E is going to be compatible with current X99 motherboards?

Hmmm.....
Might have to reconsider.
 
TroyX said:
so what are we looking at for these 14nm desktop Broadwell's in terms of speed? 6Ghz oc'd?
Click to expand...

Nicholars said:
At least 10Ghz as an estimate
Click to expand...

I LoL'd...Its been nearly impossible to get to to 5Ghz with anything after Sandy unless you got a good Ivy chip and are using top of the line (read HIGH end AIO/Custom Loop) cooling..Haswell was a joke, and DC fixed *SOME* of that, but again most chips top out ~200Mhz over their Turbo speed of 4.4Ghz...

As core die size keeps shrinking, it is harder to pull the heat from them..If you look at the die shot of a Broadwell dual core (was the only die shot out at the time), ~95% of the die is dedicated to shitty onboard graphics.:(:(
 
Still very hard to make sense of these leaked roadmaps.

When will standard desktop Broadwell arrive?
When will K-series desktop Broadwell arrive?
When will standard desktop Skylake arrive?
When will K-series desktop Skylake arrive?

I doubt they all will arrive Q2 2015.
 
It looks to me like Broadwell-k in mid April. And Skylake-k in June. And the only way this makes sense is they are for 2 different incompatible sockets.
 
Last edited:
Can we assume Broadwell-K have lower TDP than Haswell-K? I see 65W, but also 95W.
 
dr/owned said:
.

d) The market for people who want to spend $6k for a cpu for personal use is non-existant. There's "enthusiasts" on here who bitch out about upgrading their 4 year old i7-920 to something newer, and that only costs a few hundred.
Click to expand...

I know for a fact that you are wrong. ��
 
DejaWiz said:
^ agreed. I'd much rather see Intel sell the K sku's without the OEM HSF and put that money into soldered IHS's.
Click to expand...

You've always been able to buy them in an OEM SKU if you wanted.

It would be a foolish move to ditch the Retail SKU on the K's.

What about all those people who buy the K's not to overclock NOW, but do it later? Surely they must not exist, right? No, not possible. :rolleyes::rolleyes::rolleyes:
 
TheGamerZ said:
You've always been able to buy them in an OEM SKU if you wanted.

It would be a foolish move to ditch the Retail SKU on the K's.

What about all those people who buy the K's not to overclock NOW, but do it later? Surely they must not exist, right? No, not possible. :rolleyes::rolleyes::rolleyes:
Click to expand...

Not enough of them to matter. How many do you personally know? I don't know any.
 
Please give me something to upgrade to Intel. Pretty please. This is getting sad.
 
TheGamerZ said:
You've always been able to buy them in an OEM SKU if you wanted.

It would be a foolish move to ditch the Retail SKU on the K's.

What about all those people who buy the K's not to overclock NOW, but do it later? Surely they must not exist, right? No, not possible. :rolleyes::rolleyes::rolleyes:
Click to expand...

Well, that's your opinion...

Besides, where did I mention that selling a BOXED RETAIL K-SKU without the shitty OEM HSF = ditching retail SKUs? It would be the same as the retail versions of the High-End Desktop Processor line: 5820K, 5930K, 5960X...all with no HSF in the box and the full 3 year warranty, unlike 1 year (or less) on OEM/Tray processors.
 
Last edited:
dr/owned said:
5960X. More than twice as powerful as what you've currently got.
Click to expand...

5960x is a little out of my comfort zone on price, but 5820k has certainly caught my eye. Wish I had a Microcenter around here.

Getting pretty bored with S1155 even if the performance is still fine. Had the same conversation after 4 years on Q6600.
 
Nicholars said:
At least 10Ghz as an estimate

Parja said:
Since the past few new processes have had such impressive gains in clock speed headroom, right?
Click to expand...
Click to expand...

10 GHz is exactly what we were told back in 2000 that Netburst architecture would deliver by the year 2005. About 10 years later, and we've just now seen Intel deliver their first factory-clocked 4 GHz processor (i7-4790K).

...Albeit one with exponentially higher IPC which will run circles around the fastest factory clocked Netburst CPU, even when this latest Haswell is at 1/4 of its factory clock speed. :p
 
You must log in or register to reply here.
Back
Top