Intel Broadwell Thread

[Head1985]

Its all about Ram speed.If broadwell-E cant utilize Fast DDR4 3600+ then Skylake with 3600-4000Mhz DDR4 will crush it(in games).Same with broadwell with L4.
That L4 ADDs 20-40% IPC over haswell(games)

Also overclocking is big question.Most 6700k can OC to 4.7-4.8Ghz.

But if broadwell-E can scale with DDR4 like skylake then it will be decent cpu.Still overpriced but decent for games.

 

[Headfoot]

I can't way to get my Broadwell-E so that I can exclusively play games with it.

I lol'd

 

[Worthington]

Well, there may be some games that will be faster on overclocked six or maybe even 8 core BW-E. I doubt the games will utilize the 10 core, and it probably will not overclock that well either. Hard to say what DX-12 will bring on cpu demands.

Problem is, BW-E is starting at a 5% or so IPC disadvantage, and probably will not overclock quite as high as 6700k, so a game will have to use the cores really well to match or beat overclocked 6700K.

This is what is holding me off. With the minor updates to CPUs each new generation i generally keep a CPU for a while (still using a 920). DX12 is looming and part of me thinks that 8 cores at ~4.2ish will age better than 4 cores at 4.7ghz.

 

[Aristotelian]

Its all about Ram speed.If broadwell-E cant utilize Fast DDR4 3600+ then Skylake with 3600-4000Mhz DDR4 will crush it(in games).Same with broadwell with L4.
That L4 ADDs 20-40% IPC over haswell(games)

Also overclocking is big question.Most 6700k can OC to 4.7-4.8Ghz.

But if broadwell-E can scale with DDR4 like skylake then it will be decent cpu.Still overpriced but decent for games.

Could you please provide us with some benchmarks showing that ram speed (ddr4 above 3GHz) has a non-trivial impact on fps in games?

 

[AdamK47]

Haswell-E and Broadwell-E both utilize quad channel memory. They also include 20+ MB of L3 cache. That negates the need for 3600+ memory and a L4 cache. Not sure why you would even need a level 4 cache.

 

[tenks]

i love the spin/fud being put out there. Haswell-E can keep up with Skylake just fine...yet somehow, a broadwell-e will be worse on a smaller process. Got it!


lol, some people

 

[tenks]

@Worthington , Without the L4 that elevated the Broadwell-C chips over Skylake in some instances, Broadwell-E is going to be a fairly predictable ~5% IPC increase, not too much to get excited about except for those needing to increase their core count. I'd build a Haswell-E system today if needed, it's not like the CPU can't be sold later, it's well established that Intel CPUs hold their value fairly well in these days of such small improvements.

Check out the clock for clock graphs on Anandtech. Even though the results are "poisoned" a bit by the existence of the L4, trends can be extrapolated pretty well:

http://www.anandtech.com/show/9482/intel-broadwell-pt2-overclocking-ipc/3

So you're reccomending someone to buy a Haswell-E chip, when broadwell-E is less than 60 days out, and they can spend the SAME MONEY, on a better, newer chip. Horrible recomendation.

 

[frozentundra123456]

Haswell-E and Broadwell-E both utilize quad channel memory. They also include 20+ MB of L3 cache. That negates the need for 3600+ memory and a L4 cache. Not sure why you would even need a level 4 cache.

Fallout 4 is the major example of huge performance increases on Skylake with fast memory. I know the benchmarks have been posted in these forums but I am at work and dont have time to look them up. If I recall correctly, it can be a 30 or 40 percent gain in FPS with fast memory over DDD4 2133. But that is kind of a corner case, and most other games show a smaller gain.

It would be very interesting to see Haswell-E/Broadwell-E vs Skylake at max overclocks and different memory speeds in the same test, but I dont think it has been done.

 

[frozentundra123456]

So you're reccomending someone to buy a Haswell-E chip, when broadwell-E is less than 60 days out, and they can spend the SAME MONEY, on a better, newer chip. Horrible recomendation.

BW-E will certainly be newer. How much "better" is yet to be seen. It all depends on overclocking and maybe lower power use. But yea, I agree it makes sense to wait now, all else being equal. But if a great deal popped up on HW-E, I would not hesitate to buy it.

 

[crashtech]

So you're reccomending someone to buy a Haswell-E chip, when broadwell-E is less than 60 days out, and they can spend the SAME MONEY, on a better, newer chip. Horrible recomendation.

To each their own, but in tech there are always new products coming out. One could just as easily pick up a good OCing Haswell-E today as pick up a poor OCing Broadwell-E in two months, in that case with buying Haswell the performance difference is nil, and the expenditure less. You don't have a crystal ball any more than I do, chances are Broadwell-E will be a very minor improvement, but not enough to wait for if someone is ready. Again, it's not as if an updated CPU can't be dropped in the X99 platform, the CPUs we talk about here are socketed for a reason.

 

[tenks]

To each their own, but in tech there are always new products coming out. One could just as easily pick up a good OCing Haswell-E today as pick up a poor OCing Broadwell-E in two months, in that case with buying Haswell the performance difference is nil, and the expenditure less. You don't have a crystal ball any more than I do, chances are Broadwell-E will be a very minor improvement, but not enough to wait for if someone is ready. Again, it's not as if an updated CPU can't be dropped in the X99 platform, the CPUs we talk about here are socketed for a reason.

While that may be true in general, that's the most generic argument you could possibly come up with.. "new products are always coming out". Well in this case, we're talking about HEDT intel chips, and they dont always come out. The last one to come out was in 2014 and the next one to come out is less than 2 months away, that's 2+ years apart..And you're recommending the older one because "new product always come out" and somehow there is a possibility it MIGHT not OC as well, despite being on 14nm vs 22nm...solid argument there.

"Hey guys, Intel HEDT chips only come out once every 18-24 months. A new one is coming out in less than 2 months. I definitely recommend the one that came out 22 months ago" Right.

I dont need a crystal ball to tell me, based on logic, the 14nm chip will have higher OC potential than a 22nm chip. Broadwell-E is not Broadwell-C and there has been substantial extra work done on the core compared to past HEDT iterations. There is a reason why Intel was able to add a 4th sku to the lineup, a 10 core chip within the same TDP. With all things being equal, prices/core count/clock, Broadwell-E is the better buy.

 

[crashtech]

...I dont need a crystal ball to tell me, based on logic, the 14nm chip will have higher OC potential than a 22nm chip...

Hey, we're all just having some speculative fun here, but some might find your opinion to be a bit bold, since there really isn't anything out there except ES chips and extrapolations on which to base it.

 

[ehume]

. . .

I dont need a crystal ball to tell me, based on logic, the 14nm chip will have higher OC potential than a 22nm chip. Broadwell-E is not Broadwell-C and there has been substantial extra work done on the core compared to past HEDT iterations. There is a reason why Intel was able to add a 4th sku to the lineup, a 10 core chip within the same TDP. With all things being equal, prices/core count/clock, Broadwell-E is the better buy.

But transistors are closer together. Logic would say that it is possible that Broadwell will OC less than Haswell/DC. I think that we will need to see data on how well the 14nm chips OC compared with their 22nm predecessors.

Data beats theory every time.

 

[tenks]

Hey, we're all just having some speculative fun here, but some might find your opinion to be a bit bold, since there really isn't anything out there except ES chips and extrapolations on which to base it.

And all the information and benchmarks provided by the EP chips that are already out and people who've had their hands on the ES's since mid last year like FUGGER over @ XtremeSystems.

 

[crashtech]

And all the information and benchmarks provided by the EP chips that are already out and people who've had their hands on the ES's since mid last year like FUGGER over @ XtremeSystems.

I have appreciated the opportunity to have this back and forth with you. I don't think you are wrong at all. Again though, at worst, someone building a Haswell-E now (good deals can be found on 5820Ks in particular) is at worst 5% behind, but probably would have saved much more than 5%, and will get to enjoy the benefits of a fast system now. But honestly that wasn't really even my advice, Worthington didn't give enough details to really say anything meaningful. All I meant to point out is that with HEDT, you really can't go wrong. Even those those with OCed 3930Ks still have formidable systems that can handle just about anything. So I took exception to the "horrible advice" thing, I don't think 5% is horrible by any stretch. It's a calculated risk, a price/performance calculation, and lastly, how bad the guy needs to upgrade, which we don't know.

 

[ehume]

After all, SOMEbody is buying these almost-obsolete CPU's.

 

[tenks]

I have appreciated the opportunity to have this back and forth with you. I don't think you are wrong at all. Again though, at worst, someone building a Haswell-E now (good deals can be found on 5820Ks in particular) is at worst 5% behind, but probably would have saved much more than 5%, and will get to enjoy the benefits of a fast system now. But honestly that wasn't really even my advice, Worthington didn't give enough details to really say anything meaningful. All I meant to point out is that with HEDT, you really can't go wrong. Even those those with OCed 3930Ks still have formidable systems that can handle just about anything. So I took exception to the "horrible advice" thing, I don't think 5% is horrible by any stretch. It's a calculated risk, a price/performance calculation, and lastly, how bad the guy needs to upgrade, which we don't know.

Well the reason why I chimed in was because I too, am in need of upgrade, as soon as possible, so I took what you said as if it was advice towards me. I'm sorry if my responses seemed aggressive or blunt. I've been heavily invested in this ever since I decided not to buy the 5820k after planning to for so lonh and wait for BDW-E. Given the choice of buying a 5820k now for the same price as a 6800k in less than 2 months, I'd likely wait for the 6800k, especially since I've already waited this long. The price difference will be $50-$60 tops and to me, its worth paying extra knowing I'd have the newest HEDT processor and on the newest process. I tend to make my systems last 4-5 years so I'd want to buy the newest chip I could. So you are very right in the sense that each situation is different, but it is very rare that a new Intel chip is slower than the one it replaced. SO that being said, it just made sense for me to wait.

As far as buying a 5820k now and then possibly upgrading to BDW-E later, you'd lose money because you'd be paying near full price for the 5820k now. After BDW-E releases, the 5820k would lose some resale value because you can get a 6800k for the same price.

 

[crashtech]

Well the reason why I chimed in was because I too, am in need of upgrade, as soon as possible, so I took what you said as if it was advice towards me. I'm sorry if my responses seemed aggressive or blunt. I've been heavily invested in this ever since I decided not to buy the 5820k after planning to for so lonh and wait for BDW-E. Given the choice of buying a 5820k now for the same price as a 6800k in less than 2 months, I'd likely wait for the 6800k, especially since I've already waited this long. The price difference will be $50-$60 tops and to me, its worth paying extra knowing I'd have the newest HEDT processor and on the newest process. I tend to make my systems last 4-5 years so I'd want to buy the newest chip I could. So you are very right in the sense that each situation is different, but it is very rare that a new Intel chip is slower than the one it replaced. SO that being said, it just made sense for me to wait.

As far as buying a 5820k now and then possibly upgrading to BDW-E later, you'd lose money because you'd be paying near full price for the 5820k now. After BDW-E releases, the 5820k would lose some resale value because you can get a 6800k for the same price.

When you get your Broadwell-E, I hope you will do me the honor of posting your Cinebench 11.5 result to the thread linked in my sig. 11.5 is long in the tooth, but we have a pretty extensive list of CPUs in it now, and it's interesting to see how they stack up.

I hope you are right about Broadwell's price structure, I was not actually expecting price parity like you say.

 

[IntelUser2000]

But transistors are closer together. Logic would say that it is possible that Broadwell will OC less than Haswell/DC. I think that we will need to see data on how well the 14nm chips OC compared with their 22nm predecessors.

The "transistors are closer together so it clocks less" theory was only a modern thing. It never happened pre-32nm process.

The fact that overclocking records are achieved with extreme pipelined CPUs like Prescott Celerons and Bulldozer and only at 8-9GHz using LN2 suggests there are hard physical limits that won't ever be breached.

Also, 22nm and 14nm was the time Intel was killing its process on PC to hopefully be more competitive with mobile. With LGA115x chips on Haswell they also decided they'll go from solder to thermal paste, a crappy one at that. Broadwell C was rumored to be derived hastily from mobile, and not only that it was when 14nm had terrible yields.

Conclusion: Wait till it comes out and then decide.

 

[Aristotelian]

Fallout 4 is the major example of huge performance increases on Skylake with fast memory. I know the benchmarks have been posted in these forums but I am at work and dont have time to look them up. If I recall correctly, it can be a 30 or 40 percent gain in FPS with fast memory over DDD4 2133. But that is kind of a corner case, and most other games show a smaller gain.

It would be very interesting to see Haswell-E/Broadwell-E vs Skylake at max overclocks and different memory speeds in the same test, but I dont think it has been done.

Thanks, but this is why I phrased my question as "3000 or above DDR 4" because although Skylake can run memory speeds higher than Haswell-E, I could not find benchmarks showing a significant difference between say, 3000 and 4000 DDR4 in games (other applications do seem to benefit).

I have also read that this is due to 'memory controller' differences between Skylake and Haswell-E, but is it certain that this hasn't been changed with Broadwell-E?

I plan to buy the 10 core Broadwell-E for gaming and for many other applications that benefit from the use of additional cores. But I'm not surprised to see what Adam wrote - it's not like the 5960X was bad, from a gaming perspective, compared with a 6700K - although from a per dollar (gaming) performance perspective it was not the best value.

 

[guskline]

If I wanted to build an enthusiast rig now, I would probably wait because the release of Broadwell E is to be soon. You will have Haswell E chips still in the pipeline with less of a price so you can decide.

I had a 3930k rig and liked it and when I decided I wanted a newer E chip (and it was my 64th birthday present to myself!) I decided to go whole hog on a 5960x. Expensive? Yes but I always wanted the top end E chip at the time and I was able to swing it.

Some posters on this forum, like AdamK47, always upgrade to the highest end E chip Intel releases. I respect AdamK47. He is consistent and always builds a solid VERY high end system. His reasons make sense to him and I respect that.

If I was building a system almost exclusively for gaming with a single High end card where performance for the dollar was critical, the 6600k/6700k chips and mbs are hard to beat.

If I value core count and have other uses, the E series comes into the picture.

As it pertains to this thread, Broadwell E will soon officially be the standard bearer for the E series. I'm excited to see the true performance increases from the Haswell E series.

 

[csbin]

While that may be true in general, that's the most generic argument you could possibly come up with.. "new products are always coming out". Well in this case, we're talking about HEDT intel chips, and they dont always come out. The last one to come out was in 2014 and the next one to come out is less than 2 months away, that's 2+ years apart..And you're recommending the older one because "new product always come out" and somehow there is a possibility it MIGHT not OC as well, despite being on 14nm vs 22nm...solid argument there.

"Hey guys, Intel HEDT chips only come out once every 18-24 months. A new one is coming out in less than 2 months. I definitely recommend the one that came out 22 months ago" Right.

I dont need a crystal ball to tell me, based on logic, the 14nm chip will have higher OC potential than a 22nm chip. Broadwell-E is not Broadwell-C and there has been substantial extra work done on the core compared to past HEDT iterations. There is a reason why Intel was able to add a 4th sku to the lineup, a 10 core chip within the same TDP. With all things being equal, prices/core count/clock, Broadwell-E is the better buy.

14nm chip will have higher OC potential than a 22nm chip?



http://www.anandtech.com/show/8426/...view-core-i7-5960x-i7-5930k-i7-5820k-tested/3

http://www.anandtech.com/show/9482/intel-broadwell-pt2-overclocking-ipc/2

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/6

 

[tenks]

14nm chip will have higher OC potential than a 22nm chip?



http://www.anandtech.com/show/8426/...view-core-i7-5960x-i7-5930k-i7-5820k-tested/3

http://www.anandtech.com/show/9482/intel-broadwell-pt2-overclocking-ipc/2

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/6

you can show me all the graphs you want, doesnt change anything. When you shrink a node, there is more potential headroom. Whether the company can make that work for them is another thing entirely.

 

[zentan]

14nm chip will have higher OC potential than a 22nm chip?



http://www.anandtech.com/show/8426/...view-core-i7-5960x-i7-5930k-i7-5820k-tested/3

http://www.anandtech.com/show/9482/intel-broadwell-pt2-overclocking-ipc/2

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/6

Not really sure we can draw much from Broadwell-C's lacking of high overclocks in general.It could be very much due to it being limited to its overall platform(65W and effects from edram being on the package).

i7 6700k OC on average is better than the one you referenced w/ that board at AT. Can do ~4.7/4.8 at the top-end. AT have done OC with many other boards and they do seem to do better.

Lot more in the links below
http://www.anandtech.com/show/9854/asus-maximus-viii-impact-z170-rog-review-itx
http://www.anandtech.com/show/10236/the-msi-z170a-sli-plus-review
http://www.anandtech.com/show/10206/the-asus-maximus-viii-extreme-review
http://www.anandtech.com/show/9729/the-asus-z170-a-motherboard-review
http://www.anandtech.com/show/9817/the-asrock-z170-extreme7-review-triple-m2-x4-in-raid


Avg OC on air(from Hwbot)

i7 3770K 4547MHz 22nm
i7 4770K 4475MHz 22nm
i7 4790K 4654MHz 22nm
i7 6700K 4604MHz 14nm


i7 4930K(6C/12T) 4501MHz
i7 4960X(6C/12T) 4511MHz
i7 5820K(6C/12T) 4322MHz
i7 5930K(6C/12T) 4754MHz
i7 5960X(8C/16T) 4428MHz

The avg on air clocks are fairly close between 4790k and 6700k.
14nm still has to show better results perhaps,may be with kabylake successor,we do not know but will know.

Also notice the OC on 6C/12T SKUs. 5930k does seem to do quite well.
Even if Broadwell-E doesn't OC that well it could still be quite close with better efficiency and perhaps better price points for 8C/16T i7 6900K.
How good overclocking on Broadwell-E as compared compared to Haswell-E be?we will know when it comes.
It's better to wait like someone already mentioned above.

 

[Dresdenboy]

you can show me all the graphs you want, doesnt change anything. When you shrink a node, there is more potential headroom. Whether the company can make that work for them is another thing entirely.

There is a "14 nm" process with different possible metal layer densities (called "SoC", "performance", etc.) and other varying metrics. Then there are different logic design styles and pipeline FO4 delays. Having the same basic number of nanometers doesn't mean, chips have to behave the same in OC'ing.