Intel Broadwell Thread

[cbn]

The 5820K and 4790K are about $45 retail apart, therefore I am predicting $425 or so for the 6800K.

The trend for hex cores has been continually cheaper prices though. And if 10 core becomes the new top SKU I would expect that trend to continue.

With that mentioned, i7 6800K (according to that leaked chart) does have 100 Mhz higher clocks than i7 5820K, but I am thinking that would not be enough to stop a price decrease.

Another thing to factor in is that at the time of launch i7 6800K will be based an older architecture than i7 6700K. This was not true for the big socket when i7 5820K launched (re: i7 4790K was also Haswell). Furthermore, X99 was actually the newer platform at the time.

P.S. Would be glad to see the low end of the big socket return to the near $300 price level as we saw in the past.

 

[crashtech]

Maybe if Intel offers a locked version, there will be something closer to price parity, but for an unlocked hexcore, I believe there will be a premium to be paid for the extra unlocked cores.

 

[cbn]

Maybe if Intel offers a locked version, there will be something closer to price parity, but for an unlocked hexcore, I believe there will be a premium to be paid for the extra unlocked cores.

Why do think the i7 6800K will increase in price over the i7 5820K?

 

[crashtech]

Why do think the i7 6800K will increase in price over the i7 5820K?

Because the i7-6700K debuted for more than the 4790K did.

 

[cbn]

Because the i7-6700K debuted at more than the 4790K did.

i7 6700K and i7 4790K have the same price on Intel Ark.

 

[crashtech]

I don't believe that reflects reality on the street, but we'll see. It's not like I want the CPUs to be more expensive, it's just a best guess.

 

[tenks]

If you want a K vs K comparison then look at i7 4820K which was cheaper than both the i7 4770K and i7 4790K.

you're also comparing a quad to a quad.

Its now a hex to a quad. Big difference


Also I don't forsee the 6800k being more expensive then $400. The 5820k its replacing launched at $389

 

[RussianSensation]

Broadwell-E
i7-6950X 10C/20T 3.0GHz 25M
i7-6900K 8C/16T 3.3GHz 20M
i7-6850K 6C/12T 3.6GHz 15M
i7-6800K 6C/12T 3.4GHz 15M

http://www.xfastest.com/thread-165075-1-1.html

This is awesome, moving 8-core down and replacing it with a 10-core. If Intel keeps the pricing structure similar and the 10-core starts at $999, this means they are trying to attract even more enthusiasts away from the mainstream Z170 platform. With Broadwell, the architecture differences will ensure that IPC is even closer to the 6700K. Even now it's not a clear cut choice to go with 6700K vs. 5820K but with 6800K, it'll be even harder to justify i7 6700K. I can see a lot more PC gamers just paying $100 extra for the X99 + 6800K combo to have 2 more cores over HT and maybe a 6-7% IPC difference. Can't wait to see how well BW-E overclocks.

I think the existing mainstream i7 6700K is really an i7 in the marketing name only. If we look at the i7 as representing high end performance of the past, the current i7 6700K will not reflect that at all considering 6-10 core i7 offerings on X99. If BW-E moves to a 6/8/10-core setup, I will instantly consider the i7 6700K a glorified i5 regardless of the marketing name.

http://wccftech.com/intel-broadwell...7-6900k-core-i7-6850k-core-i7-6800k-detailed/

Anyway, I am all for Intel re-focusing its efforts on the high-end CPU segment. Give me more cores over the useless IGP and HT for $100 extra. :thumbsup:

I knew for years that Intel was just holding back adding more cores to maximize profits. This means they are finally at a point where they can go 6-10 core and maintain similar or even greater profits -- not surprising considering how much of the die space is taken up by the crappy IGP on Skylake on 14nm. :sneaky:

I also expect a reduction in power usage for the 6-8 core parts (i.e., 4930K vs. 3930K):

The day BW-E launches is going to be a bad day for all AMD engineers who worked tirelessly to catch up to Intel only to find out Intel is going 6/8/10 core, which would ultimately mean at least 6/8/10-core SKL-E in 2017.

Also I don't forsee the 6800k being more expensive then $400. The 5820k its replacing launched at $389

Even if it's $20-30 more expensive, would still be a great deal against the $385 i7 6700K. 2 full cores, the same motherboard offering the option of getting an 8-10 core down the line, and probably a very small deficit in IPC against 6700K will make it almost a no-brainer to keep over 4-5 years of ownership.

 

[cbn]

you're also comparing a quad to a quad.

Its now a hex to a quad. Big difference

The issue was the numbering scheme related to price---> http://forums.anandtech.com/showpost.php?p=37831788&postcount=3067

But as mentioned in this thread, the hexcore does have lower clocks and no iGPU. And with i7 6800K vs.i7 6700K it will be an older architecture as well.

P.S. i7 4820k did have 200 Mhz higher base clocks and 2MB more cache than i7 4770K, but it still cost less money than i7 4770K. Of course, like the i7 6800K vs. i7 6700K comparsion it is based on an older architecture (Ivy Bridge) compared to the Haswell on the i7 4770K.

 

[guskline]

Happy with my 5960X. Can't wait to see how the 10C/20T overclocks!

 

[cbn]

The day BW-E launches is going to be a bad day for all AMD engineers who worked tirelessly to catch up to Intel only to find out Intel is going 6/8/10 core, which would ultimately mean at least 6/8/10-core SKL-E in 2017.

On ebay the E5 2670 (Sandy Bridge 8C/16T, 2.6 Ghz base/3.3 Ghz turbo) is now dropping in price. Currently going for $179.00 shipped and available in volume. This is a pretty dramatic change from a few months ago when these priced much higher. The higher clocked SB E5 8C/16T SKUs should follow at a later point. It might even be we see some Ivy Bridge E5 Xeons available in surplus quantities by the time of the Zen 8C/16T launch.

 

[cbn]

I don't believe that reflects reality on the street, but we'll see. It's not like I want the CPUs to be more expensive, it's just a best guess.

Ok, but what do you think will happen with the Intel Ark price?

 

[jpiniero]

Intel moving the enthusiasts to the server line would make a bunch of sense, especially if Cannonlake mainstream can't hit the high clock speeds.

Still think 6950X is going to be more than $999. While the yields are getting better there's still been talk of the die cost being still high. Milking out more money out might be one way of making that up.

Even if it's $20-30 more expensive, would still be a great deal against the $385 i7 6700K. 2 full cores, the same motherboard offering the option of getting an 8-10 core down the line, and probably a very small deficit in IPC against 6700K will make it almost a no-brainer to keep over 4-5 years of ownership.

But would you be able to overclock the 6800K to make up the clock speed deficit? The gap between Broadwell and Skylake is higher than you would think because Broadwell-C has the EDRAM. You can get most of what you mentioned today with the 5820K.

 

[Arachnotronic]

Intel moving the enthusiasts to the server line would make a bunch of sense, especially if Cannonlake mainstream can't hit the high clock speeds.

I think that HEDT and mainstream will essentially converge by 1H 2017 (KBL and SKL-E) and at some point it will make no sense for the mainstream socketed platform to exist.

Mainstream will go embedded and HEDT will have a wider range of products derived from Xeon.

 

[Arachnotronic]

The day BW-E launches is going to be a bad day for all AMD engineers who worked tirelessly to catch up to Intel only to find out Intel is going 6/8/10 core, which would ultimately mean at least 6/8/10-core SKL-E in 2017.

There are some of us on this board who told those who were getting all excited about AMD's efforts that there are huge advantages to the fact that HEDT is derived from Intel's Xeon server lineup. AMD is coming in with an 8 core Zen? No problem. Relabel a "Xeon" with more cores (and Xeon EP for Broadwell goes all the way up to 22; Skylake goes to 28) as a HEDT processor and unlock the multiplier.

Intel can play AMD like a fiddle and anybody who thinks that Intel's marketing team is dumb enough to let AMD "one up" Intel in such a lucrative segment probably needs to rethink their understanding of the industry :thumbsup:

Still pleasantly surprised that this is basically a pre-emptive strike rather than a reactionary move with SKL-E.

 

[RussianSensation]

On ebay the E5 2670 (Sandy Bridge 8C/16T, 2.6 Ghz base/3.3 Ghz turbo) is now dropping in price. Currently going for $179.00 shipped and available in volume. This is a pretty dramatic change from a few months ago when these priced much higher. The higher clocked SB E5 8C/16T SKUs should follow at a later point. It might even be we see some Ivy Bridge E5 Xeons available in surplus quantities by the time of the Zen 8C/16T launch.

That's what I love about the workstation chipset -- you have the option to go multi-core/much faster CPUs later on as they drop in price while the mainstream socket is completely dead end. Look at the X58 users enjoying 4.3-4.6Ghz 6-core goodness on their boards? Still rocking. Unless something dramatic happens in the mainstream platform, for my next upgrade I am going workstation, especially if it only costs $100-125 extra to do that and in the process get more cores.

But would you be able to overclock the 6800K to make up the clock speed deficit? The gap between Broadwell and Skylake is higher than you would think because Broadwell-C has the EDRAM. You can get most of what you mentioned today with the 5820K.

Skylake has barely higher IPC vs. Haswell, so I am going to estimate 8% advantage over BW-E at best. Let's assume 6800K can hit 4.5Ghz vs. 4.8Ghz on the 6700K.

4.8Ghz * 1.08 / 4.5Ghz = 15%

If BW-E can hit 4.6-4.8Ghz, it's completely game over for the 6700K. I would personally already take 5820K over 6700K, so 6800K would be automatic for me over 6700K. 15% IPC vs. 50% more cores and workstation upgrade path option. I made a mistake of buying the i5 cripple and not doing that again. Most games hardly benefit from higher Skylake IPC over Haswell while at least I could find a use case for a 6-8 core CPU outside of games. Not to mention I'd rather have extra cores in reserve than a measly 15% single core reserve. If you need more cores, you can't make them out of thin air. If i7 6700K was a $250-270 US processor, that would be a different story but it's selling for $500 CDN before tax => $565 with taxes for a quad-core FFS. At that point I am inclined to pay extra and get 6-8 cores, even if the architecture is 1 gen behind because increases in IPC don't matter to the extent they did with C2Q -> Nehalem/Lynnfield -> Sandy -> Haswell.

Intel can play AMD like a fiddle and anybody who thinks that Intel's marketing team is dumb enough to let AMD "one up" Intel in such a lucrative segment probably needs to rethink their understanding of the industry :thumbsup:

Still pleasantly surprised that this is basically a pre-emptive strike rather than a reactionary move with SKL-E.

I never expected AMD to catch up in the $350+ segments. Where I thought they had any shot was the $50-350 ones. That's where Intel is particularly weak, especially with the weak sauce i3s and i5s.

 

[Arachnotronic]

That's what I love about the workstation chipset -- you have the option to go multi-core/much faster CPUs later on as they drop in price while the mainstream socket is completely dead end. Look at the X58 users enjoying 4.3-4.6Ghz 6-core goodness on their boards? Still rocking. Unless something dramatic happens in the mainstream platform, for my next upgrade I am going workstation, especially if it only costs $100-125 extra to do that and in the process get more cores.

Here's the fly in the ointment so to speak when it comes to Skylake...

With Skylake-E platform (Basin Falls) the PCH will be the same for both mainstream (Kaby Lake) and HEDT. HEDT will be Socket 2011 (v4) with quad channel memory while the Xeon E5 will use the Lewisberg PCH and Socket P (LGA 3467-ish).

So I don't think you'll be able to drop in Xeons in future HEDT platforms but it does look like Intel is beginning to be more aggressive about bringing more cores to the HEDT platform faster.

 

[RussianSensation]

Here's the fly in the ointment so to speak when it comes to Skylake...

With Skylake-E platform (Basin Falls) the PCH will be the same for both mainstream (Kaby Lake) and HEDT. HEDT will be Socket 2011 (v4) with quad channel memory while the Xeon E5 will use the Lewisberg PCH and Socket P (LGA 3467-ish).

So I don't think you'll be able to drop in Xeons in future HEDT platforms but it does look like Intel is beginning to be more aggressive about bringing more cores to the HEDT platform faster.

Thanks for that info. I haven't been keeping up on the technical side on SKL-E since I noticed rumors pointing to it shifting towards Q1 2017. In any case, even if we ignore the Xeon upgrade path, you could still buy the 6-core now and the 10-core in 5 years. It's at least an option. What about Z170? Quad-core crippled for life.

Also, with both diminishing IPC increases and diminishing user benefits per each new IPC architecture (i.e., we are quickly moving into an era where desktop PC performance is primarily GPU+I/O: PCIe limited, not CPU limited), to me the focus in the enthusiast segment should shift to more cores. PC enthusiasts would be much more receptive of upgrading/paying $ to get 6/8/10/12/16 cores vs. a measly 10% IPC increase every 2 years. If I am upgrading, I want to feel that I am getting more IPC and more cores which is why an i7 6700K mainstream platform is disappointing right off the bat. Glad to see that someone at Intel is realizing there is an untapped market which is willing to pay more as long as Intel offers more cores.

Still don't know if Intel will put the 10-core at $999 or introduce a higher priced $1199-1499 segment.

 

[IndyColtsFan]

If I am upgrading, I want to feel that I am getting more IPC and more cores which is why an i7 6700K mainstream platform is disappointing right off the bat.

I definitely agree with your assessment. I've been struggling with upgrading to a 6700K for a couple of months (my rig is in my sig). Yes, a GPU update next year will really help my current system, but it is well over 4 years old and I have an upgrade itch. The ridiculously discounted 5820K at Fry's really has my interest but I really wonder if I should just hold out a little longer and go straight to Broadwell E. It is definitely an interesting time in the HEDT arena.

 

[tential]

I literally buy high end CPUs because I like having them. I cannot rationally justify a 10 core CPU for any of my workloads (I play games), so I will not try to

This is the reason I'm moving to the HEDT platform as my next build. Just because =D

 

[Sweepr]

Well here's my prediction.

6800K at 5820K price or slightly lower, $370-390 (28 PCIe lanes).
6850K @ $449-499 (40 PCIe lanes).
6900K @ 5930K price, $580-600 (40 PCIe lanes).
6950X @ 5960X price (40 PCIe lanes).

6900K would be my pick as a future-proof chip to pair high-end 16nm dGPUs with.

There are some of us on this board who told those who were getting all excited about AMD's efforts that there are huge advantages to the fact that HEDT is derived from Intel's Xeon server lineup. AMD is coming in with an 8 core Zen? No problem. Relabel a "Xeon" with more cores (and Xeon EP for Broadwell goes all the way up to 22; Skylake goes to 28) as a HEDT processor and unlock the multiplier.

Talk about stealing Zen's thunder. Worst case scenario Skylake-E is a higher-clocked 10C/20T chip in a brand new platform. Best case, considering they are going from 22 to 28 cores on servers, maybe a 12C/24T Skylake-E? Forget Haswell-E, the target has moved.

I can imagine some very agressive 6800K pricing right before Summit Ridge's launch. Wouldn't hurt the more expensive parts too much because enthusiasts would still want 8 or 10 cores.

 

[myocardia]

Well here's my prediction.

6800K at 5820K price or slightly lower, $370-390 (28 PCIe lanes).
6850K @ $449-499 (40 PCIe lanes).
6900K @ 5930K price, $580-600 (40 PCIe lanes).
6950X @ 5960X price (40 PCIe lanes).

6900K would be my pick as a future-proof chip to pair high-end 16nm dGPUs with.

I would agree with every one of these, except I'm gonna bet that Intel will add exactly $100 to the top slot. The fastest HEDT chip has always been the one that people who aren't worried at all about price have bought, and since they will be getting more cores, as well as the platform upgrades, I believe Intel will take advantage of that.

Forget Haswell-E, the target has moved.

Indeed.

 

[cbn]

There are some of us on this board who told those who were getting all excited about AMD's efforts that there are huge advantages to the fact that HEDT is derived from Intel's Xeon server lineup. AMD is coming in with an 8 core Zen? No problem. Relabel a "Xeon" with more cores (and Xeon EP for Broadwell goes all the way up to 22; Skylake goes to 28) as a HEDT processor and unlock the multiplier.

Intel can play AMD like a fiddle and anybody who thinks that Intel's marketing team is dumb enough to let AMD "one up" Intel in such a lucrative segment probably needs to rethink their understanding of the industry :thumbsup:

IMHO AMD is probably better serverd by focusing on performance per watt with good enough single thread for mobile and high core count server*. With this strategy they at least stand a chance to go head to head with Intel on anything but high performance desktop.

So it will be interesting to see how they evolve the Zen series? Do they target a certain single thread performance level (ie, frequency x IPC) and then optimize purely for performance per watt with each successive Zen generation? Or do they follow the pathway we typically see for big core CPUs where single thread is increased for every new generation

*This is something that cat cores were not able to achieve (for the most part). Construction cores did have decent single thread, but performance per watt wasn't there.

 

[jpiniero]

Wouldn't hurt the more expensive parts too much because enthusiasts would still want 8 or 10 cores.

Will they? Games are getting better at core scaling, but it seems that 16 threads is going to be way overkill. Espeically when DX12/Vulcan games are on the horizon... the need for CPU power is going to be less. Surely there will be people demanding it because of bragging rights but I think it may just be that. The server line does offer more bandwidth; and that seems to be more useful.

It would be interesting if Intel stuck the EDRAM on Broadwell-E. Don't bet on it though.

IMHO AMD is probably better server by focusing on performance per watt with good enough single thread for mobile and high core count server

Except that's not what Zen is. That's what K12 (was?)

 

[cbn]

Except that's not what Zen is. That's what K12 (was?)

Wasn't there a news post with Jim Keller saying K12 would have a bigger engine than Zen?

If so, I wonder if part of the plan was/(is?) for K12 to emulate x86 for the high core count servers.

EDIT: Here is the news post I remember with the info:

http://techreport.com/review/26418/amd-reveals-k12-new-arm-and-x86-cores-are-coming

Keller even outright said that "the way we built ARM is a little different from x86" because it "has a bigger engine."