Intel Broadwell Thread

[RussianSensation]

I think AMD will have a very hard time gaining share in the enthusiast segment. Intel is in the driver's seat here. By the time Zen comes, Intel will be on the eve of launching Skylake-E, which might increase the top core count to 12, meaning that instead of gunning for Intel's $500+ CPUs with 8 core Zen, it will be stuck fighting defective 12 core chips masquerading as 8 core chips

Here is the scary part. Even if AMD puts an 8-core Zen out on time and hits 40% IPC increase, the CPU would need to have at least as good overclocking headroom as Intel's competing part, otherwise the IPC 'comeback' will get neutralized by Intel's overclocking headroom. Bulldozer/Vishera could hit high clocks but it's MUCH harder to overclock well with a short/efficient pipeline. OTOH, Intel is already hitting 4.4-4.5Ghz on the 5960X and this should grow with SKL-E.

AMD would need to undercut Intel on both the CPU side and motherboard/chipset platform side but with new AMD management, with Fiji they showed that they are not interested in playing any price/performance games even if their product is slower by a lot. That means AMD is not going to be in a position to play price/performance price wars with Intel either.

Basically they are just falling into the 101 business case strategy failures - no first mover advantage, no financial ability to compete on price, lagging behind technologically (IPC) on both per core and number of core performance offerings and most importantly, all of this assumes their chips don't skyrocket in perf/watt with overclocking. That's a big assumption in itself since we don't know how well GloFo's FinFET compares to Intel's proven node. By the time GloFo is on 1st gen high performance FinFET, Intel will be on an extremely proven 14nm and soon 10nm. Just another scenario where AMD cannot win.

AMD just needs to pray that Intel continues to have a lackluster $100-350 mainstream segment with cripple i3s, weak sauce i5 and glorified i5 - aka i7 quad. I honestly don't see Zen even touching the 6-core SKL-E HDET platform when looking at OC vs. OC. I'd be surprised if they could.

Their best strategy would be to have big discounts on AMD CPU + mobo + GPU combos but in the last 10 years they have shown that they don't want to do this.

 

[RussianSensation]

But you are only asking for 4 Ghz (not max clocks) and the EVO is rated at 180W TDP (with the single stock fan).

Curious to know where did this limitation of the 212 Evo come from in the first place?

It should be pretty easy to find mid-range priced coolers that would beat the 212 Evo for similar price or for barely more $.

http://www.nansgaminggear.net/ProductDetails.asp?ProductCode=TS-140-BW-RevA

http://www.performance-pcs.com/spire-gemini-iii-universal-cpu-cooler.html

http://www.performance-pcs.com/spire-thermax-eclipse-iii-universal-cpu-cooler.html

http://www.newegg.com/Product/Produ...5&cm_re=cryorig_h7-_-9SIA4UF2DZ6565-_-Product

 

[Arachnotronic]

Basically they are just falling into the 101 business case strategy failures - no first mover advantage, no financial ability to compete on price, lagging behind technologically (IPC) on both per core and number of core performance offerings and most importantly, all of this assumes their chips don't skyrocket in perf/watt with overclocking. That's a big assumption in itself since we don't know how well GloFo's FinFET compares to Intel's proven node. By the time GloFo is on 1st gen high performance FinFET, Intel will be on an extremely proven 14nm and soon 10nm. Just another scenario where AMD cannot win.

Basically this.

 

[Arachnotronic]

Can you believe some people tried to convince us Intel wouldn't react with more agressive Broadwell-E/Skylake-E SKUs to protect their margins? Curiously they are missing from this thread right now.

Yep. Those claims were a product of wishful thinking from the perspective of people who want to see AMD win/Intel fail.

 

[DrMrLordX]

Can you believe some people tried to convince us Intel wouldn't react with more agressive Broadwell-E/Skylake-E SKUs to protect their margins? Curiously they are missing from this thread right now.

. . . do you really want off-topic thread crapping? It's kind of nice that people can (mostly) stick to discussing Broadwell in a Broadwell thread. For now.

Zen looks to be an interesting product. This thread is not the proper place to discuss it at length.

 

[tenks]

Rumours have it:

6950X = $999
6900K = $550
6850K = $440
6800K = $390
http://wccftech.com/intels-flagship...s-detailed-upto-10-cores-25mb-cache-140w-tdp/

If these prices are true, RIP glorified i5 - aka i7 6700K. 6900K for $550 would be an incredible pre-emptive strike against Zen. And now imagine in 2017 SKL-E 6/8/10/12 core offerings. I personally think the above prediction is too optimistic. I mean $550 6900K basically makes 6850K completely pointless. Intel could probably easily sell 6900K for $699-799. It also implies an almost linear relationship between price and the number of cores ($390 / 6 = $65 per core on the 6800K vs. $550 / 8 = $68.75 per core on the 6900K).

The crazy part is 6950K is 10C/20T with 3.5Ghz boost, 6900K is 8C/16T with 3.7Ghz boost - both at 140W TDP - while 6800K is only 6C/12T with 3.6Ghz boost in the same 140W TDP. That means there should be a lot of overclocking headroom on the 6800K model since its 140W TDP seems more of an arbitrary rating.

If these prices and specs for BW-E come true and Intel adds even more cores with SLK-E, it's going to be incredibly hard for Zen to hit its profitability targets. I bet AMD would never have anticipated Intel moving 8-core to a $550-600 price levels. AMD better pray this doesn't happen. With Intel becoming so aggressive, the mainstream platform's i7 quads are going to be a hard sell imo.

Those prices are just pure wccftech speculation. It is by no means anymore accurate than any of our guesses on this board.

For $110 more you can go from 6-> 8 cores. But then for $450 more you can go from 8->10 cores...Cost $110 for 2 more cores in the middle of the range but then $450 for 2 more cores on the higher end?...not very even. Doesn't seem right.

I think this surprise for Broadwell-E is Intel stepping up stuff in time for Skylake-E..Who knows exactly when they will hit (2h 2017) for real, I'm sure Broadwell-E is going to compete with Zen for a short period of time and Intel wanted to be ready. Also from a marketing perspective, I'm sure they wanted to have a "10c" around and ready for Zen.

 

[ShintaiDK]

Those wccftech prices are wrong. Why do people even read that junksite? It can only attract imbeciles.

The current X model cost 1059$. Its not going to be cheaper than that.

 

[tenks]

Welp...looks like the 2nd delay rumor was true...no more Q1...q2 sadness

 

[cbn]

Their best strategy would be to have big discounts on AMD CPU + mobo + GPU combos but in the last 10 years they have shown that they don't want to do this.

It would be great to see them discount APU + A88X motherboard.

One reason is the A88X has eight SATA ports, so it lends its self to be a cheap NAS.

For example with this SATA controller card (reported to work straight out of the box on
this compatibility list), a person can hit the 12 drive limit for unRAID Server Plus.

 

[Sweepr]

Core i5 comparison: Core i5 4690 vs Core i5-5675C vs Core i5-6500

10% faster than Haswell @ gaming (at lower clocks), take a look at the minimum FPS.

www.computerbase.de/2015-10/intel-core-i5-6500-5675c-4690-test

Really underrated chip. If only it was widely avalable wordwide at the recommended price I think it would be a nice upgrade option for Haswell Pentium/Core i3 owners.

 

[JoeRambo]

Core i5 comparison: Core i5 4690 vs Core i5-5675C vs Core i5-6500

10% faster than Haswell @ gaming (at lower clocks), take a look at the minimum FPS.

www.computerbase.de/2015-10/intel-core-i5-6500-5675c-4690-test

Really underrated chip. If only it was widely avalable wordwide at the recommended price I think it would be a nice upgrade option for Haswell Pentium/Core i3 owners.

Indeed. That L4 really shines when faced with game like load that does not fit in caches and has not that predictable access patterns. Skylake with L4 is gonna be even more epic.

 

[Sweepr]

Some Broadwell-EP action. 2x Xeon E5-2698 v4 (20C/40T @ 2.1GHz):

http://world.taobao.com/item/35605880767.htm

In comparison, scores from current the Xeon E5 v3 lineup:

~3.75x the MT performance of Core i7-5960X.

According to the latest leak the fastest model (Xeon E5-2699 v4) will sport 22C/44T.

 

[dark zero]

How much hundred thousand dollars it will cost?

 

[Sweepr]

Broadwell-E launching near Computex 2016 (May/June) according to Bench-life.

Still no word on pricing, but the four previously leaked SKUs are still there, including the 10C/20T monster.

 

[Arachnotronic]

Broadwell-E launching near Computex 2016 (May/June) according to Bench-life.

Still no word on pricing, but the four previously leaked SKUs are still there, including the 10C/20T monster.

The big downside to a May/June launch (which seems reasonable enough) is that this probably means SKL-E is a May/June 2017 affair. BDW-E being more compelling than thought should help mitigate that though.

 

[frozentundra123456]

Well, lets wait and see how "compelling" BW-E is outside of the niche 10 core monster. Basically, how will it overclock. If it has less maximum overclock than HW, like the desktop chips, performance could be a wash (or even a slight regression) except for someone that can use the top of the line 10 cores.

 

[AtenRa]

Number one benefit over HW-E will be power consumption due to 14nm. Performance on the same core count should be the ~same.

 

[Sweepr]

Number one benefit over HW-E will be power consumption due to 14nm. Performance on the same core count should be the ~same.

Not at stock. 100-200MHz clock bump + better IPC (~4-5%) could result in ~10% better MT performance for 6C-8C models. Overclocking potential remains to be seen. Also I expect these models to be priced lower than same core count Haswell-E, Intel will basically set Summit Ridge's price ceiling here.

Core i7-6950X is looking like a killer EE model, 25% more cores with no clock regression and slight IPC bump.

 

[ShintaiDK]

Number one benefit over HW-E will be power consumption due to 14nm. Performance on the same core count should be the ~same.

Be careful you dont overestimate the 14nm benefit at those frequencies. 10 cores, 3Ghz at 14nm is 140W, just as 8 cores was on 22nm.

And its no surprise, 4Ghz on Skylake is 91W. 4Ghz on Haswell was 88W. 3.4/4Ghz vs 3.2/4Ghz in the 65W segment for 14nm Skylake vs 22nm Haswell. Broadwell is at 3.3/3.8Ghz, but it cheats due to GT4 and EDRAM.

It wont be any different with 14LPP.

 

[Fjodor2001]

Welp...looks like the 2nd delay rumor was true...no more Q1...q2 sadness

Yes, as usual the IST members were wrong. Despite that they vigorously tried to defend their stance to the bitter end, ignoring all signs of the second Broadwell-E delay.

 

[tenks]

Yes, as usual the IST members were wrong. Despite that they vigorously tried to defend their stance to the bitter end, ignoring all signs of the second Broadwell-E delay.

IST members?


You're a blind, bitter AMD fanboy that brings nothing to the conversation. I know you're sad about Zen's non-2016 launch date and being wrong, but you just literally posted in an intel thread to say "see, neener neener, I told you so".


Intel's minor delays bring you happiness. AMD's constant delays put you in denial. Sad life is sad.

 

[Arachnotronic]

Yes, as usual the IST members were wrong. Despite that they vigorously tried to defend their stance to the bitter end, ignoring all signs of the second Broadwell-E delay.

You realize that this delay is probably due to the fact that 14nm yields are still a challenge meaning that the initial volumes of Broadwell-EP will go to data center customers who will pay big bucks for those chips, right?

In the meantime Intel will continue to sell Haswell-E chips as they still offer excellent performance.

 

[Phynaz]

Yes, as usual the IST members were wrong. Despite that they vigorously tried to defend their stance to the bitter end, ignoring all signs of the second Broadwell-E delay.

Feel better now? Got it all out of your system?

 

[Fjodor2001]

Wow! The members of the IST brigade decides that attack is the best defense. Better to attack the messenger for telling the truth instead of admitting you were wrong.

Hahaahha .... you are hilarious...

 

[Sweepr]

Wow! The members of the IST brigade decides that attack is the best defense. Better to attack the messenger for telling the truth instead of admitting you were wrong.

Hahaahha .... you are hilarious...

They were wrong by one quarter, while you are wrong by ~1 year in your desktop Skylake launch prediction. Hilarious isn't it?

Also, I'm pretty sure those users will be running Broadwell-E long before you ever touch a Zen CPU. There will be plenty of marketing slides to fuel your jokes though, like this one:

You're just bitter because in 2016 you'll be sitting on an expensive and slow 4 core Intel CPU, while others will be using a cheaper and faster 8 core AMD CPU.

Please keep us entertained.