Intel Broadwell Thread

[ehume]

It seems like prices from amazon and other retailers were just a premium
http://www.tweaktown.com/articles/7...-lga-1150-cpu-performance-overview/index.html

Thanks for the link. It also discusses OCing the Broadwell. The top end looks to be 4.0-4.2GHz, down a bit from Haswell.

 

[dark zero]

Seems that OC a Broadwell isn't as nice as expected... If the GPU could be OC'd to Gtx 750 levels....

 

[zir_blazer]

Inmature BIOSes and Linux support. Half a year of delay and they STILL need to tinker with it?

It would be a miracle of Skylake launchs in a timely manner with actual stock. Since they're launching the K series models first with the mainstream models one or two months later, I guess you can get the picture.

 

[DrMrLordX]

Weird. That almost reminds me of the weird problems Tom Pabst had trying to do Linux kernel compilation on a P3 1.13 GHz. Almost.

 

[lowmotion]

iGPU cant be overclocked like haswell. Thats very sad

Expect final bios at the end of the month.

 

[steve wilson]

Inmature BIOSes and Linux support. Half a year of delay and they STILL need to tinker with it?

It would be a miracle of Skylake launchs in a timely manner with actual stock. Since they're launching the K series models first with the mainstream models one or two months later, I guess you can get the picture.

Do they normally release the K models first?

 

[DrMrLordX]

iGPU cant be overclocked like haswell. Thats very sad

Eh? That TweakTown review showed that it could be overclocked via multiplier or by using a bclk strap. Ditto for the eDRAM. Or are you more concerned by the clockspeed limits on both?

 

[jpiniero]

I guess you won't really see Broadwell-C (or Broadwell-H for that matter) until Windows 10 gets released.

 

[zir_blazer]

Do they normally release the K models first?

No as far that I recall. K models were initially introduced with Nehalem-based Lynnfields, but they were an afterthought, not part of the initial lineup. The Core i7 875K arrived like 9 months later than the other models. With the Nehalem shrink, Westmere, the same thing happened, there was a Core i5 655K Clarkdale which arrived months after the mainstream models.
Sandy Bridge, Ivy Bridge, and Haswell had K models at launch. Desktop Broadwell has only two K... urgh, I say, C series models. And now, Skylake will launch first with K models, then at least a month later the mainstream parts.

However, since Broadwell arrived so ridiculous late, was launched with absolutely no stock (Paperlaunch), and the guys with the samples complain about inmature BIOSes (First the guy that couldn't overclock via Multiplier, now inability to boot up or Kernel panics)... Can someone actually say that Skylake, that was pushed forward, wouldn't have the same issues or worse? At this point I would think that the reason why they are launching the high priced parts first is because they may still not be able to deliver enough volume to fill the market.

 

[ehume]

They gave us the i7 875K Lynnfield (45nm) instead of a Westmere (32nm), despite the 6-core folks getting a Westmere. My daughter has my old 4-core 875K. It was like getting Devil's Canyon instead of Broadwell.

 

[Auric]

iGPU cant be overclocked like haswell. Thats very sad

Expect final bios at the end of the month.

Eh? That TweakTown review showed that it could be overclocked via multiplier or by using a bclk strap. Ditto for the eDRAM. Or are you more concerned by the clockspeed limits on both?

http://www.tweaktown.com/articles/7...ga-1150-cpu-performance-overview/index11.html

I would think that most people who buy the 5775C will buy it for the integrated Iris Pro graphics, and while this part of the article doesn't really cover iGP overclocking, I will show you the eDRAM multiplier.

...

The bottom line is that you probably won't be getting the same overclocks as Haswell. That being said, you can overclock to 4.0-4.2GHz stable with a modest memory overclock, and you can overclock the iGP. You need to be careful if you use the BLCK divider, you will need reduce the multipliers, especially the cache ratio.

Not everyone wants to overclock the CPU, and doing so is likely limiting versus just overlocking the GPU. Haswell GT2 can easily OC from 1150 up to 1600 MHz (39%) on ASRock H97 and others. So hopefully that option will be available if not already in the current Broadwell microcode BIOS.

 

[lowmotion]

If you want to use Broadwell by using iGPU you would like to set CPU to ~3.0 Ghz and GPU to max. Level.

At the moment there is no max. Level. Sure bclk strap is an option, but a bad one.

Lots of games are playable at 1920x1080. Next time i will try Battlefield 4.

 

[DrMrLordX]

Well right, the iGPU is the limiting factor, not the CPU, at least when it comes to using the iGPU for games.

It's just that the tweaktown article made it sound like iGPU overclocking worked. It neglected to say how well, or what trouble they had to go through in order to accomplish such a feat. Overclocking the iGPU by way of bclk straps is awkward since you need to lower other multipliers to compensate (assuming you aren't overclocking everything).

 

[StrangerGuy]

Weird. That almost reminds me of the weird problems Tom Pabst had trying to do Linux kernel compilation on a P3 1.13 GHz. Almost.

Broadwell is a chip that has no good reason to exist.

 

[Idontcare]

Broadwell is a chip that has no good reason to exist.

From the consumer's perspective, you could say that about any "tick" in Intel's tick-tock model.

Tick's are about risk-management in ramping new process nodes to HVM, you need a "known good" silicon vehicle for those crucial 2-3 quarters of HVM ramp so your engineers can spend their time debugging process issues in isolation of silicon design issues.

What we lack, knowledge-wise, is the history of how tick's have helped Intel have smoother (or less delayed) node ramps to date, 14nm included.

Had Broadwell not existed, we can easily conclude 14nm skylake would probably have been all the more delayed as Intel's process engineers would be expending resources trying to differentiate between process induced issues and design induced issues. The very reason the tick-tock model was implemented.

 

[ehume]

I still want to see AT do an OC article on the Broadwell.

BTW - I wish we had a Thanks button. IDC's point about ticks (not the arthropod variety) was thoughtful.

 

[Sweepr]

Another review:

www.thinkcomputers.org/intel-core-i7-5775c-broadwell-processor-review

Once again, much more efficient than 84-88W Haswell.

 

[frozentundra123456]

I am assuming he meant desktop broadwell, for which I also have a hard time finding a suitable fit except for a SFF device where the user wants best igpu *and* cpu performance and doesnt care about price. On the desktop, it actually has the same problem as AMD's apus, in that a discrete card is still a much better solution except in a very power or space constrained device. Desktop iris pro is an even worse value because of the price, but that is partially compensated for by much better cpu performance.

But otherwise, as IDC said, the statement makes no sense.

 

[frozentundra123456]

Another review:

www.thinkcomputers.org/intel-core-i7-5775c-broadwell-processor-review

Once again, much more efficient than 84-88W Haswell.

Overclocked is faster than Haswell, but still uses less power. Nice, but why dont they just cut the TDP and put it in a laptop? Downclock the cpu somewhat to save power, and leave as much headroom as possible for the igp. Seems like this level of igpu performance in a 600.00 to 700.00 laptop would find a good market.

Seems about 30% more efficient in the first pass test. (12% faster and 17% less power). Clearly puts to rest those arguments of BW being less efficient than Haswell posited by some (at least one) posters.

Edit: problem is, ultimate performance is still about the same or even lower if you overclocked Haswell. On the desktop, I would still rather see a performance increase even if it used correspondingly more power.

 

[ShintaiDK]

Nice, but why dont they just cut the TDP and put it in a laptop?

Its already in laptops.

http://ark.intel.com/products/family/84979/5th-Generation-Intel-Core-i7-Processors#@Mobile

 

[frozentundra123456]

Well, the chip may be "launched" but do you have any information on a laptop you can actually buy that has the chip in it?

 

[ShintaiDK]

Well, the chip may be "launched" but do you have any information on a laptop you can actually buy that has the chip in it?

May not be the one you seek, but:
http://www.newegg.com/Product/Product.aspx?Item=N82E16834152807

 

[frozentundra123456]

May not be the one you seek, but:
http://www.newegg.com/Product/Product.aspx?Item=N82E16834152807

Thanks for the link, but that is just like a desktop. The iris pro is wasted. The point was a cheaper one using the iris pro for light gaming.

 

[DrMrLordX]

For those looking for good i7-5775C OC data, Fugger delivers:

http://www.xtremesystems.org/forums/showthread.php?291735-Broadwell-5775C-on-Phase

Here we see an i7-5775C hitting 4.8 GHz with an AiO cooler. The voltage is kinda ugly. Broadwell might be able to take those kind of volts, or maybe not. My vote would be for not.

Also, for those of you who thought the HKEPC OC was fake . . . do you also think Fugger is a fake?

 

[StrangerGuy]

For those looking for good i7-5775C OC data, Fugger delivers:

http://www.xtremesystems.org/forums/showthread.php?291735-Broadwell-5775C-on-Phase

Here we see an i7-5775C hitting 4.8 GHz with an AiO cooler. The voltage is kinda ugly. Broadwell might be able to take those kind of volts, or maybe not. My vote would be for not.

Also, for those of you who thought the HKEPC OC was fake . . . do you also think Fugger is a fake?

I very suspicious of stellar OCing results even if it was real because who knows what kind of cherrypicking was done behind the scenes to get those clocks.

If the OC ceiling is now even lower than a stock 4790K the entire OCing scene will pretty much be dead if Intel maintains stock clocks from Haswell.