Intel Broadwell Thread

[NTMBK]

Makes me wonder if BW-E will improve at all over HW-E.

Depends on how aggressively you're willing to overclock. If your motherboard and cooling can handle a 250W CPU, it will probably be pretty damn fast.

 

[frozentundra123456]

That is the problem though, I doubt it will overclock better than Haswell, perhaps not as well, based on BW-C. So all we are left with is a measly 5% ipc gain and a 1000.00 plus 10 core monster that will be a very niche product.

 

[ShintaiDK]

That is the problem though, I doubt it will overclock better than Haswell, perhaps not as well, based on BW-C. So all we are left with is a measly 5% ipc gain and a 1000.00 plus 10 core monster that will be a very niche product.

The custom 5.1Ghz 4/8 SKU does give some hopes tho.

 

[NTMBK]

The custom 5.1Ghz 4/8 SKU does give some hopes tho.

If you have a high-frequency-trader's budget, sure

 

[guskline]

Moar cores!!!!:thumbsup::thumbsup::thumbsup:

 

[jpiniero]

Aren't the 1U/E parts fabbed on different dies than the EP ones?

 

[witeken]

Okay, the numbers are in.

Broadwell-EP has a 1.85x higher transistor density.

Calculation: (662 mm² / 454 mm²) x (7.2B x'tors / 5.7B x'tors) = 1.84187

x'tor source: http://www.nextplatform.com/2016/03/31/broadwell-brings-xeon-e5s-balanced-performance-bump/

(Very close (within 4%) to the theoretical 1.92x (0.52x), but quite a bit below BDY-Y's 2.2x.)

 

[ShintaiDK]

Aren't the 1U/E parts fabbed on different dies than the EP ones?

Not the LGA2011 ones and for E no.

 

[jpiniero]

Not the LGA2011 ones and for E no.

OK, for some reason I thought the 1P/E, EP, and EX were all different dies.

 

[JDG1980]

The custom 5.1Ghz 4/8 SKU does give some hopes tho.

It will be interesting to see whether this part ever gets a HEDT release, or if it remains an ultra-expensive, custom-order part only. A guaranteed 5.1 GHz would be very nice for gamers, especially those who still play single-threaded and lightly-threaded titles that are CPU-bound (for instance, I believe Starcraft 2 falls into this category). We don't know what yields are like at these levels.

 

[jpiniero]

It will be interesting to see whether this part ever gets a HEDT release, or if it remains an ultra-expensive, custom-order part only. .

It's definitely an custom order offlabel part only. I believe the price of the Westmere 4.4 Ghz Xeon's was over 2 grand, and that was a cut dual core. This QC Broadwell one is likely more than that. But the NSA and Hedge funds will pay for it.

 

[maddogmcgee]

When is x99 broadwell supposed to come out anyway? Just bought a new PSU and am trying to decide if a wait is worth it. Worst case I might get a discount on haswell I guess...

 

[IntelUser2000]

That is the problem though, I doubt it will overclock better than Haswell, perhaps not as well, based on BW-C. So all we are left with is a measly 5% ipc gain and a 1000.00 plus 10 core monster that will be a very niche product.

We don't know how it'll overclock. Because Broadwell C is basically mobile with a desktop socket. There were rumors about that as well.

Or else why would Skylake improve on it so much?

(Very close (within 4%) to the theoretical 1.92x (0.52x), but quite a bit below BDY-Y's 2.2x.)

I don't think transistor density data is useful at all. Outside of engineers that designed them, we don't really know to the full extent what they do. 90nm Prescott had 125 million transistors which was well over 2x over 0.13u Northwood, but the speculation is most of them were for something not important to us(like better debugging circuits or such).

Broadwell Y has greater reduction because there's much less I/O and logic circuits that don't scale as well. Graphics and cache does which Y has plenty of.

So the ONLY real judgment of how a process fares is on the product itself. I'd say for PC their 22nm and 14nm sucks.

 

[ShintaiDK]

So the ONLY real judgment of how a process fares is on the product itself. I'd say for PC their 22nm and 14nm sucks.

And you defined they sucked how? Both pushed stock clocks past 4Ghz while giving huge performance/watt on lower/mobile parts.

 

[frozentundra123456]

And you defined they sucked how? Both pushed stock clocks past 4Ghz while giving huge performance/watt on lower/mobile parts.

That seems more of a marketing decision,to push stock clocks higher because performance gains were so small otherwise. Sandy Bridge on 32nm easily reaches equal or faster clocks than either 22 or 14 nm. Saying they "suck" is rather meaningless, but we have had 2die shrinks and 2 new architectures since SB with only about 25 to 30 percent performance gain, so I dont think one could argue that they were great from an absolute performance perspective. And 14nm was definitely a troubled node all along from a yield point of view.

 

[ShintaiDK]

6950X on downloadcenter.
https://downloadcenter.intel.com/pr...sor-Extreme-Edition-25M-Cache-up-to-3-50-GHz-

 

[frozentundra123456]

6950X on downloadcenter.
https://downloadcenter.intel.com/pr...sor-Extreme-Edition-25M-Cache-up-to-3-50-GHz-

Why no windows 10 drivers?

 

[LTC8K6]

Why no windows 10 drivers?

https://downloadcenter.intel.com/do...-on-3rd-generation-Intel-Processors-or-above-

Listed right there.

 

[witeken]

https://downloadcenter.intel.com/pr...sor-Extreme-Edition-25M-Cache-up-to-3-50-GHz-

Core i7 6950X

 

[LTC8K6]

https://downloadcenter.intel.com/pr...sor-Extreme-Edition-25M-Cache-up-to-3-50-GHz-

Core i7 6950X

That link is three posts above yours...

 

[Nothingness]

I have what is most likely a silly question :$

What's the point of a $1500 10-core 3.5 GHz 6950X when one can get a $940 10-core 3.4 GHz Xeon E5-2640 v4? Overclock? What do I miss?

Or is base frequency an issue? My E5-2650 v2 has no issue sustaining turbo...

 

[witeken]

That link is three posts above yours...

oops

 

[ShintaiDK]

I have what is most likely a silly question :$

What's the point of a $1500 10-core 3.5 GHz 6950X when one can get a $940 10-core 3.4 GHz Xeon E5-2640 v4? Overclock? What do I miss?

Or is base frequency an issue? My E5-2650 v2 has no issue sustaining turbo...

Where did you see 1500$ as an official price?

And yes, overclock.

 

[Aristotelian]

Where did you see 1500$ as an official price?

And yes, overclock.

There have been rumours about a price increase for a while (i.e., on WCCFTech), and we discussed this earlier in the thread. I'm hoping that the 6950X is USD 999 but we will see.

I, for one, am happy to pay more for an overclockable chip.

 

[LTC8K6]

If the E5-2640 v4 90W Xeon will work in the boards, and sustain it's 3.4ghz turbo speed, and it's cheaper, it would seem to be a no-brainer over the 6950X for enthusiasts.