Broadwell-E 10 cores?

[RussianSensation]

It's 6/6/8/10 cores for Broadwell-E, which leaves open the possibility that the 6/6/8 will retain similar configurations and pricing to Haswell-E, and the 10 core is $2,000.

Some people will still buy the 6950X just because it raised the bar for performance, but it's not exactly the offering that even HEDT buyers are looking for.

That's a good point. We still don't know the actual prices to assume that 6950X is $999. It could be more expensive. $2000 though seems way too steep.

This sucks. I would feel bad buying a new "enthusiast" CPU with a two-year old architecture no matter how many cores it has, but everyone will make you feel bad for buying the mainstream i7 actually of the latest architecture because it's a quad-core. Can't win no matter what.

Mainstream hexa-core; why can't Intel give us that...

I think you can "win" depending on your own needs. If you need the absolute fastest gaming performance, an i7 6700K @ 4.8-4.9Ghz will beat any BW-E in almost all modern games. If you need great gaming performance + multi-tasking for things other than games, then you choose BW-E. What makes it tricky is predicting gaming performance in future games, including DX12 games. Compared to launch prices, Z170 boards are dropping fast. The very high-end Gigabyte Gaming 7 Z170 is only $170 but is stacked to the moon and back in features. Right now you cannot get an X99 board for $170 that has USB 3.1, dual Ultra M.2 slots, Creative SoundCore 3D chip that blows every other X99 motherboard away, besides possibly the Asus X99 Rampage boards.

Source

All of a sudden, if you are using speakers, not headphones, to match this you'd need to get an X99 board + a dedicated sound card or you sacrifice sound quality. Many PC gamers will deny deny that onboard sound is good enough but facts are facts. It's good enough but it's not great enough.

Also, if don't need more than 1 GPU, you can get a good Z170 motherboard for $110-115. For a lot of Z170 buyers, this board should meet their needs easily.
http://www.newegg.com/Product/Produ...cm_re=Z170_motherboard-_-13-157-634-_-Product

What about X99? If you start nit-picking detail-by-detail, you cannot get a $115 X99 board, and it's going to be impossible to get a $170 X99 board with USB 3.1, dual Ultra M.2 slots and awesome sound.

So I guess I was too critical of the 6700K to some extent; there are some advantages that could be a lot bigger than they first appear.

Also, with a very basic $25-30 cooler, you can max out the 6700K without any effort. It's going to be much harder (if not impossible) to contain the temperatures of a 6-10 core BW-E @ 4.6-4.8Ghz on such a budget cooler. I am not going to lie though as there are $40 coolers that could do it hehe.

Source

Of course some of these factors I mentioned don't matter if a gamer already has a high-end CPU cooler that can be re-used, if a gamer is using headphones where the X99's amp is doing the work / if a gamer carries over a dedicated sound card from a previous build. If these points do matter though, the X99 setup will cost more since you'd need a higher end CPU cooler and a higher-end mobo with better sound or a better sound card. The Z170 user can just use that $ saved and get a 950 Pro 256GB PCIe SSD or step-up from the GTX980 to the 980Ti as an example. It's places like MicroCenter that make X99 so much more attractive.

It is frustrating that Intel is splitting the platforms like that. I'd take a 6-core mainstream CPU on the latest architecture, but a lot of people on these boards resent such an idea...for some reason. :\

 

[Aristotelian]

It's 6/6/8/10 cores for Broadwell-E, which leaves open the possibility that the 6/6/8 will retain similar configurations and pricing to Haswell-E, and the 10 core is $2,000.

Some people will still buy the 6950X just because it raised the bar for performance, but it's not exactly the offering that even HEDT buyers are looking for.

Thanks for the chart in your previous post - which seems to indicate that Broadwell-E will be the only -E class available in 2016. I intended to wait for Skylake-E and go all out, but at this rate that'd mean I have to wait to build that new machine until...mid 2017 at the earliest?

I'm on an overclocked 2600k and am itching to do a new overkill build - the 2600k has really served me well since 2010-11 when I built this rig, so I guess a 10 core broadwell-E should serve me quite a while too.

 

[beginner99]

What makes it tricky is predicting gaming performance in future games, including DX12 games.

Agree. Especially because a DX12 game could make use of the iGPU meaning higher FPS on mainstream platform. I assume we will see games of both kind: such that benefit from 6+ cores (albeit a bit lower clocked) and such that benefit from an iGPU.

Because users with iGPU is far greater than 6+ core CPUs as an game engine developer I would clearly invest in making use of the iGPU.

 

[Arachnotronic]

I'm on an overclocked 2600k and am itching to do a new overkill build - the 2600k has really served me well since 2010-11 when I built this rig, so I guess a 10 core broadwell-E should serve me quite a while too.

I think a 10 core Broadwell-E will easily last you 5+ years.

 

[aigomorla]

are these supposed to work in the X99?

or is intel slapping a new board for them?

Dayam i miss the days i was sponsored by a trusted partner... i probably would already have a sample. :\

 

[Arachnotronic]

are these supposed to work in the X99?

or is intel slapping a new board for them?

Dayam i miss the days i was sponsored by a trusted partner... i probably would already have a sample. :\

X99.

 

[ShintaiDK]

are these supposed to work in the X99?

or is intel slapping a new board for them?

Dayam i miss the days i was sponsored by a trusted partner... i probably would already have a sample. :\

X99. Kaby Lake PCH will span 2 generations of E as well.

 

[tenks]

SKYLAKE-EP roadmaps from the summer showed 2h 2017 at the earliest. This was before broadwell-e got delayed again to 2nd qtr.

I don't see Skylake-E hitting until Fall 2017.

 

[Aristotelian]

A bit of a newbie question - why would they limit these CPUs to 40 PCIe lanes - it always results in a tri-sli configuration of, at most, 16x, 16x, 8x. I guess it's because the difference in the third slot of 8x and 16x is small - but why can't the CPU have 48 PCIe lanes?

And how will this work if someone wants to use PCIe drives, which seem to be increasing in speed at a drastic rate.

I'm asking this question within the context of someone ready to pay EUR 1000 ish for a CPU but still finds the built-in limitations a bit odd in that price bracket.

 

[Arachnotronic]

A bit of a newbie question - why would they limit these CPUs to 40 PCIe lanes - it always results in a tri-sli configuration of, at most, 16x, 16x, 8x. I guess it's because the difference in the third slot of 8x and 16x is small - but why can't the CPU have 48 PCIe lanes?

And how will this work if someone wants to use PCIe drives, which seem to be increasing in speed at a drastic rate.

I'm asking this question within the context of someone ready to pay EUR 1000 ish for a CPU but still finds the built-in limitations a bit odd in that price bracket.

48 lanes is coming with Skylake-E, you will soon get your wish

The reason that it's limited is probably due to the die size constraints at 32nm/22nm; a reticle is only so big so it's a question of whether Intel and its customers would trade off PCIe lanes for more cores/cache or not.

At 14nm and 10nm, there is a lot more headroom so Intel can grow the number of PCIe lanes it includes while at the same time growing core count.

 

[ShintaiDK]

A bit of a newbie question - why would they limit these CPUs to 40 PCIe lanes - it always results in a tri-sli configuration of, at most, 16x, 16x, 8x. I guess it's because the difference in the third slot of 8x and 16x is small - but why can't the CPU have 48 PCIe lanes?

And how will this work if someone wants to use PCIe drives, which seem to be increasing in speed at a drastic rate.

I'm asking this question within the context of someone ready to pay EUR 1000 ish for a CPU but still finds the built-in limitations a bit odd in that price bracket.

Remember the CPUs comes from the server segment. So as such its entirely irrelevant what people on the HEDT platform wants.

These CPUs works in x99 boards. So whatever limit that was, these will have too.

With Skylake-E you may see 48 lanes. Since we know some SKUs at least supports it. However if the Skylake-E will get it as well is another matter.

 

[Arachnotronic]

With Skylake-E you may see 48 lanes. Since we know some SKUs at least supports it. However if the Skylake-E will get it as well is another matter.

100% that Skylake-E will have 48 lanes, at least in the top SKU.

 

[Aristotelian]

Thanks guys. So someone building a broadwell-E system both for games and wider usage where cores are usable may be better off going SLI (2x16) and leaving 8 lanes free for a primary drive such as the Intel P3608 as a primary drive.

I get that they come from the server segment - I suppose I'm just realising that the true enthusiast segment (all out) is probably never catered for - I've seen users on this forum of 4-way SLI etc, but even that alone uses all the lanes, preventing pcie ssd use.

Sorry to ask a bit of a broader question on this but I'm really looking forward to this E release.

 

[tenks]

Sorry to ask a bit of a broader question on this but I'm really looking forward to this E release.

I think a lot of people are. If the 8core launches at $550-$700, I think a lot people that were planning on buying a 6c like the 5820k or it's successor, wouldn't mind paying $200-$300 more for an 8c...I know I wouldn't. Which would bring more cache and PCIe lanes too. win/win/win

 

[RussianSensation]

100% that Skylake-E will have 48 lanes, at least in the top SKU.

I'd rather have 32-40 PCIe 4.0 lanes:

PCIe 4.0 x8/x8 for the GPUs and PCIe 4.0 x8/x8 for RAID PCIe NVMe (40 PCIe 4.0 would leave room for USB3.1, etc.)

Head spinning speed" is on tap to those using the Z Turbo Quad Pro, says HP. The firm claims that its PCIe Gen3 solution offers data speeds up to 16x faster than a standard SATA SSD – that's up to 9.0GB/s sequential read speeds. Furthermore an "industry leading NVMe controller technology specifically architected for low latency for SSD," provides 3x better random read performance compared to a traditional SATA SSD product.

http://hexus.net/tech/news/storage/88049-hp-launches-z-turbo-quad-pro-workstation-storage-solution/

Yaaaaaaaaaaaaaaaa :wub:

 

[Arachnotronic]

I'd rather have 32-40 PCIe 4.0 lanes:

PCIe 4.0 x8/x8 for the GPUs and PCIe 4.0 x8/x8 for RAID PCIe NVMe (40 PCIe 4.0 would leave room for USB3.1, etc.)

Head spinning speed" is on tap to those using the Z Turbo Quad Pro, says HP. The firm claims that its PCIe Gen3 solution offers data speeds up to 16x faster than a standard SATA SSD – that's up to 9.0GB/s sequential read speeds. Furthermore an "industry leading NVMe controller technology specifically architected for low latency for SSD," provides 3x better random read performance compared to a traditional SATA SSD product.

http://hexus.net/tech/news/storage/88049-hp-launches-z-turbo-quad-pro-workstation-storage-solution/

Yaaaaaaaaaaaaaaaa :wub:

I think we won't see PCIe 4.0 until Cannonlake at the very least, if not Icelake

Skylake-E is PCIe 3.0.

 

[Arachnotronic]

I think a lot of people are. If the 8core launches at $550-$700, I think a lot people that were planning on buying a 6c like the 5820k or it's successor, wouldn't mind paying $200-$300 more for an 8c...I know I wouldn't. Which would bring more cache and PCIe lanes too. win/win/win

I am really glad that Intel is making BDW-E more compelling like this. $1k for yet another 8 core would have just been a waste of everybody's time, IMO. People don't mind paying big for high performance CPUs (especially for people for whom this is their primary hobby), but at least give us a solid improvement gen-to-gen.

Before this leaked, I had zero intention of buying BDW-E. Now I am wondering how I am going to time the sale of my 5960X to maximize resale value in order to nab one of those 6950X chips at fairly minimal cost.

SKL-E better have 12 cores

 

[Arachnotronic]

BTW, here is the reason Intel seems to all of a sudden give a hoot about enthusiasts:

26% growth in gaming desktops in a market that's declining. Fast growing, high margin market? Intel is finally waking up to enthusiasts

 

[ShintaiDK]

NVidia and AMD is seeing the same, more high end cards sold. The question is if it can hold. It is very volatile after all depending on games to raise the bar.

But highend Skylake CPUs are selling like hotcakes for now. And so will NUCs

 

[jpiniero]

BTW, here is the reason Intel seems to all of a sudden give a hoot about enthusiasts:
26% growth in gaming desktops in a market that's declining. Fast growing, high margin market? Intel is finally waking up to enthusiasts

Is that people buying the 8-core? Or is it people buying a PC with 970s to play Dota 2?

The 10 core is really about bragging rights and not practicality. It will be priced as such.

 

[ShintaiDK]

Is that people buying the 8-core? Or is it people buying a PC with 970s to play Dota 2?

The 10 core is really about bragging rights and not practicality. It will be priced as such.

The growth is not from HEDT. But higher end LGA1151 chips sold contra previous. So ye, its basically GTX970+i5 or so.

 

[Arachnotronic]

Is that people buying the 8-core? Or is it people buying a PC with 970s to play Dota 2?

The 10 core is really about bragging rights and not practicality. It will be priced as such.

I think demand for 6700K and 5820K is probably quite robust. 5930K is a weirdo that's overpriced for the value it delivers and 5960X is basically for people who want the fastest CPU just because.

 

[RussianSensation]

NVidia and AMD is seeing the same, more high end cards sold.

Do you have a source for that? The GPU market's volume unit sales have dropped by about half. I think what you are discussing is that NV/AMD have increased ASP over time, which is not the same at all as claiming that there are actually more high-end cards sold in terms of units.

Back in the glory days of GPUs, 100-130 million annual GPU sales was normal. Today, it's barely 65 million.

Also, when the market is made up of 100%, if in the past the market share between both firms was roughly 50/50 or 60/40 and suddenly one firm gets 20%+ to 80%+, it's no wonder it has record revenue/profits, but that doesn't tell us much about the market as a whole because we'd need to compare the entire market. I highly doubt that high-end graphics in terms of overall units are selling as well as in the olden days. In the past, one had to upgrade every 12-18 months to play the latest games. This isn't even remotely the case anymore where a $650 card can last 4-5 years on console PC ports. That surely has dropped the demand for high-end cards overall.

The growth is not from HEDT. But higher end LGA1151 chips sold contra previous. So ye, its basically GTX970+i5 or so.

I think Intel would disagree with the above.

"While there were year-over-year declines in desktop and notebook volume, we saw record Core desktop mix due to growth in the high-end segment and record Core i7 mix overall for the PC business,” http://www.kitguru.net/components/c...core-i7-processors-set-records-in-q2-company/~ source

i5 is not high end by Intel's own definition, it is the mainstream segment, which only underlines the quote above that there is greater growth in % terms in the enthusiast sales, coming from the i7 and HEDT (i.e., Premium and Extreme segments).

I am really glad that Intel is making BDW-E more compelling like this. $1k for yet another 8 core would have just been a waste of everybody's time, IMO. People don't mind paying big for high performance CPUs (especially for people for whom this is their primary hobby), but at least give us a solid improvement gen-to-gen.

Same but it sets a risky precedent since if you move from 8-core HW-E to a 10-core BW-E and you already expect 12-core SKL-E, it means if Intel stalls increasing the number of cores, you won't even upgrade and skip a generation. Up until now, did you skip a generation?

I think demand for 6700K and 5820K is probably quite robust. 5930K is a weirdo that's overpriced for the value it delivers and 5960X is basically for people who want the fastest CPU just because.

Yup, certain sites like TechReport still keep crapping on the 5820K for not having enough PCIe lanes. But anyone with real tech knowledge who has done the math knows the extra lanes on 5930K are basically marketing for most people. If you are using that many PCIe lanes, you are running some seriously expensive hardware - far beyond beyond dual-SLI GPU (PCIe 3.0 8x/8x) + a pair of 950Pro PCIe NVMe (PCIe 3.0 x4/x4). At that point I cannot imagine not having enough money to spend on a 5960X. 5930K was a "sucker's marketing CPU" unfortunately. Will be interesting to see where they price the 6850K relative to the 6800K and 6900K. If it's too high up from 6800K, just another marketing CPU.

 

[Arachnotronic]

Same but it sets a risky precedent since if you move from 8-core HW-E to a 10-core BW-E and you already expect 12-core SKL-E, it means if Intel stalls increasing the number of cores, you won't even upgrade and skip a generation. Up until now, did you skip a generation?

I moved to an 8 core HSW-E from a 4930K. The resale value of these HEDT chips is quite good so once you take the initial "hit," it's not too terrible to upgrade at an annual basis. However, I will generally only upgrade if I think I am getting something materially better than what I had before.

Let's suppose I do buy a 10 core BDW-E to replace my 8 core HSW-E...if SKL-E comes in at "only" 10 cores, then do I really want to go through the trouble of ripping and replacing my motherboard/system for +5% perf/clock + AVX-512 + 8 more PCIe lanes? Nope.

But if they offer a 12 core chip, then I know I'm getting all of the improvements above + two more CPU cores. That, at least to me, is worth the hassle of trying to sell my old CPU + mobo a few months ahead of launch and having to basically rebuild my system on top of the additional cash outlay. Fortunately SKL-E won't require new memory like moving to HSW-E did!

tl;dr - if SKL-E is 10 core only, and if I upgrade to a 10 core Broadwell-E, then I wouldn't upgrade; I'd rather put that money towards more GPU horsepower or something. If it's 12 cores, then I'd be much more tempted

Yup, certain sites like TechReport still keep crapping on the 5820K for not having enough PCIe lanes. But anyone with real tech knowledge who has done the math knows the extra lanes on 5930K are basically marketing for most people. If you are using that many PCIe lanes, you are running some seriously expensive hardware - far beyond beyond dual-SLI GPU (PCIe 3.0 8x/8x) + a pair of 950Pro PCIe NVMe (PCIe 3.0 x4/x4). At that point I cannot imagine not having enough money to spend on a 5960X. 5930K was a "sucker's marketing CPU" unfortunately. Will be interesting to see where they price the 6850K relative to the 6800K and 6900K. If it's too high up from 6800K, just another marketing CPU.

I'm right with you there. 5820K is a great processor, especially if you overclock. It's hard to beat a hex core @ 4.3-4.4GHz overclocked for <$400. It is, IMO, the best "value" today for gamers. :thumbsup:

 

[ShintaiDK]

Do you have a source for that?.

NVidia financial statements. Plus we can see the same with Intel. 26% gaming growth.

I think Intel would disagree with the above.

Nope.