Intel Broadwell Thread

[IntelUser2000]

http://www.notebookcheck.net/Asus-Transformer-Book-T300-Chi-Convertible-Review.138153.0.html

what about the T300 Chi

T300 Chi performs pretty well, much better than all the other production Core M devices, including the Macbook, but it sucks elsewhere. I am becoming more convinced its entirely Core M's fault.

At 4.5W, it can't perform anywhere near Intel hyped back with reference designs. T300 performs much better, but needs 7-9W to do so.

http://www.ultrabookreview.com/6868-transformer-book-chi-review/
http://www.ultrabookreview.com/wp-content/uploads/2015/03/perf-cinebench.jpg
http://www.ultrabookreview.com/wp-content/uploads/2015/03/throttling-games.jpg

when running the OpenGL part of the Cinebench R15 test, HWInfo shows the CPU Package Power reaching values of around 12W, while when running the multi-core CPU test the Package Power stabilizes around 7.5-7.6 W for the entire duration of this benchmark. Only when performing the Single Core CPU pass the Package power drops below the designed TDP of 4.5 W.

That's pretty horrible. It looks like Abwx was right on the mark where it needs exotic metal plate cooling system to achieve anywhere near what Intel hyped back at IDF. Battery life is horrible too. I bet Intel management is laughing their asses off to folks that buy a $1300 system that performs like a one that's half the price.

Another thing. People in the Apple Forums are pretty disappointed that the Macbook loses to an iPad Air 2. It might not be Broadwell generation, but I really see Apple taking Intel out of few(if not all) lines in the near future.

 

[krumme]

I think the 4.5w tdp is set as goal by top management. Reminds me of the mhz race.
Damn stupid and a shame for a fine arch. 14nm have some battery problems but its kind of natural especially for controlling a new node. No need to go below 8-9w. Its clearly out of this ooo big arch window.
Lets see what skylake and cannondale can do but it looks to get harder and harder. 4.5w is more pr bs than reality.

 

[Qwertilot]

Well give it a die shrink or two - and them not expanding out the base TDP of core - and it might come into the natural range.

Until then it looks like a fair few are configuring to a higher TDP (as I'm fairly sure it allows), and seemingly making quite decent machines with it. Especially the ones priced more round 6/700.

It, of course, really is devices that matter in the end here

 

[IntelUser2000]

I think its more than management issues. They are having real problems.
-Intel claims that Core M is 2x the performance of iPad Air in their own benchmarketing applications. Why is it fairing much worse everywhere else?
-Perhaps CPUs can be optimized for such Tablet benchmarks and usage models, then why does Atom suck against ARM competition there too?

Even if in the PC usage model ARM chips suck much comparatively to Intel chips, it doesn't matter because mobile benchmarks seem pretty relevant to mobile usage.

Intel seems full of promises but also full of unfulfilled ones and excuses. Either CPU advancement is nearing its end or that other manufacturers are far more competent than them.

 

[StrangerGuy]

I think its more than management issues. They are having real problems.
-Intel claims that Core M is 2x the performance of iPad Air in their own benchmarketing applications. Why is it fairing much worse everywhere else?
-Perhaps CPUs can be optimized for such Tablet benchmarks and usage models, then why does Atom suck against ARM competition there too?

Even if in the PC usage model ARM chips suck much comparatively to Intel chips, it doesn't matter because mobile benchmarks seem pretty relevant to mobile usage.

Intel seems full of promises but also full of unfulfilled ones and excuses. Either CPU advancement is nearing its end or that other manufacturers are far more competent than them.

Intel's secrecy and delays about 14nm sounds like a trainwreck in progress to me. Heck, even the 14nm delays started all the way back to late 2012 with quick Google.

 

[jj109]

It's just a perception issue on your end.

The A8X throttles by 30% in long term CPU-only load.
Core M does not at 4.5W

Core M CPU single thread is 60% faster than A8X CPU single thread, climbing to >80% after extended load.

Core M GPU is 2x as fast as the A8X's GPU in Ice Storm Unlimited

Core M will also happily eat >10W in Prime95 and Furmark until it hits thermal limits (more than the entire Apple tablet!), so there's that going for the Air.

 

[ShintaiDK]

Core M pretty much showed how far the best ARM chip is behind and made a big joke out of it. A8X looks like a slow powerhungry Atom compared.

 

[krumme]

It's just a perception issue on your end.

The A8X throttles by 30% in long term CPU-only load.
Core M does not at 4.5W

Core M CPU single thread is 60% faster than A8X CPU single thread, climbing to >80% after extended load.

Core M GPU is 2x as fast as the A8X's GPU in Ice Storm Unlimited

Core M will also happily eat >10W in Prime95 and Furmark until it hits thermal limits (more than the entire Apple tablet!), so there's that going for the Air.

So what?

When you start to do some actual work that normal users do eg. real world gaming the platform starts to tank.

Its a freaking expensive solution for very little benefit. Positioning a products so you end up competing directly with far less expensive solutions with no perceptible user benefit is a dead end and the sales numbers will show even when they force it down oem and comsumers throat. Completely idiotic as if 9w tdp slick business machines was not far more valuable. They erode their performance brand.

 

[IntelUser2000]

So reading some user review of the T300 Chi says that the battery life figure was quoted in Asus Store as such:

Core M 5Y71
2560x1440 display
5 hours battery life

Core M 5Y10
1920x1080 display
8 hours battery life

Users say they get 6-7.5 hours consistent on the FHD version. That's not bad. We don't know the effect of having 5Y71, but its clear the display has an effect.

I am wonder why only one manufacturer managed to get a superior performing and better form factor device for few hundred dollars cheaper?

 

[CHADBOGA]

But, but...MorphCore! WCCFTech told us Skylake would be big and disruptive

After Skylake comes out, I wonder how long it will take for some to twist the story & replace WCCFTech, with Intel. :hmm:

 

[Arachnotronic]

After Skylake comes out, I wonder how long it will take for some to twist the story & replace WCCFTech, with Intel. :hmm:

Not long, unfortunately.

 

[Hulk]

CPUmark99 Yes I know this benchmark is very outdated, only tests integer performance for legacy code and does not take multicores into account. These are actually the reasons I like to use it to compare IPC of different generations of CPUs.

I also like to convert the score into how many MHz a particular CPU needs to generate 1 CPUmark99.

Here are some "MHz/CPUmark99" scores. Notice how Broadwell improves on Haswell. I've been collecting scores for years. First number is MHz/CPUmark99, second number is percentage improvement from previous CPU.

I find the obvious wrong turn of Northwood and Prescott (Netburst) interesting as well as the nearly doubling of IPC upon moving from Prescott to Conroe.

Of course this doesn't take clockspeed or cores into account which is why the apparently small increase from Nehalem to Sandybridge was actually much larger due to the clockspeed difference between them.

Anyway I thought you guys might find this interesting. Broadwell actually brings a decent improvement from Haswell. The larger OOS, 50% larger TLB, additional TLB misshandler, improved branch predictor and other improvements seem to add up.

Sorry about the terrible formatting I don't know how to keep the columns straight when posting.

1989 - 486 25.00
1995 - Pentium P54C 16.40 52.4%
1997 - Pentium P55C (MMX) 14.90 10.1%
1995 - Pentium Pro 10.50 41.9%
1998 - Celeron Mendocino 12.50 -16.0%
1998 - Pentium II (Dechutes) 13.20 -5.3%
1999 - Pentium III (Katmai) 13.00 1.5%
1999 - Pentium III (Coppermine) 11.20 16.1%
2000 - Pentium 4 Willamette 17.00 -34.1%
2001 - Pentium III (Tulatin) 11.00 54.5%
2002 - Pentium 4 Northwood 15.80 -30.4%
2004 - Pentium 4 Prescott 20.70 -23.7%
2006 - Conroe 7.10 191.5%
2007 - Penryn 6.90 2.9%
2008 - Nehalem 6.50 6.2%
2010 - Westmere 6.60 -1.5%
2011 - Sandy Bridge 6.40 3.1%
2012 - Ivy Bridge 6.30 1.6%
2013 - Haswell 5.80 8.6%
2014 - Broadwell 5.35 8.3%

 

[SAAA]

Wow! The Pentium Pro looks so good in that test, wish we had infinite time and money to compare all the chips at the same multi-GHz speeds we have now...
Also what the hell were they thinking with Prescott? xD

 

[Hulk]

Wow! The Pentium Pro looks so good in that test, wish we had infinite time and money to compare all the chips at the same multi-GHz speeds we have now...
Also what the hell were they thinking with Prescott? xD

They were thinking (dreaming) 10GHz

But they weren't thinking about the consequences of P=IV

If you remember, through the 90's processors were sold almost entirely on MHz. It was an easy sell, easy to understand for consumers, and easy for the marketing people to understand and push. Faster is better. Perhaps the marketing people pushed the engineers into Netburst and they all went over the cliff hand-in-hand. Luckily someone grabbed a "Conroe" branch before they hit the bottom!

BTW, back in 2005 before the sh$t hit the fan for AMD the Athlon X2 was getting a 7.7 in this benchmark. In reality only about 8% behind Conroe in integer performance IPC-wise. But Conroe had better floating point performance and clockspeed among other things which made it unarguably the better part.

 

[DrMrLordX]

BTW, back in 1995

1995? Try 2005. Heh.

 

[ShintaiDK]

1989 - 486 25.00
1995 - Pentium P54C 16.40 52.4%
1997 - Pentium P55C (MMX) 14.90 10.1%
1995 - Pentium Pro 10.50 41.9%
1998 - Celeron Mendocino 12.50 -16.0%
1998 - Pentium II (Dechutes) 13.20 -5.3%
1999 - Pentium III (Katmai) 13.00 1.5%
1999 - Pentium III (Coppermine) 11.20 16.1%
2000 - Pentium 4 Willamette 17.00 -34.1%
2001 - Pentium III (Tulatin) 11.00 54.5%
2002 - Pentium 4 Northwood 15.80 -30.4%
2004 - Pentium 4 Prescott 20.70 -23.7%
2006 - Conroe 7.10 191.5%
2007 - Penryn 6.90 2.9%
2008 - Nehalem 6.50 6.2%
2010 - Westmere 6.60 -1.5%
2011 - Sandy Bridge 6.40 3.1%
2012 - Ivy Bridge 6.30 1.6%
2013 - Haswell 5.80 8.6%
2014 - Broadwell 5.35 8.3%

So pretty much destroys the notion that everything was better IPC wise in the past using legacy code.

 

[Hulk]

1995? Try 2005. Heh.

I was thinking 2005 and typed 1995. Fixed. Thanks.

 

[witeken]

1989 - 486 25.00
1995 - Pentium P54C 16.40 52.4%
1997 - Pentium P55C (MMX) 14.90 10.1%
1995 - Pentium Pro 10.50 41.9%
1998 - Celeron Mendocino 12.50 -16.0%
1998 - Pentium II (Dechutes) 13.20 -5.3%
1999 - Pentium III (Katmai) 13.00 1.5%
1999 - Pentium III (Coppermine) 11.20 16.1%
2000 - Pentium 4 Willamette 17.00 -34.1%
2001 - Pentium III (Tulatin) 11.00 54.5%
2002 - Pentium 4 Northwood 15.80 -30.4%
2004 - Pentium 4 Prescott 20.70 -23.7%
2006 - Conroe 7.10 191.5%
2007 - Penryn 6.90 2.9%
2008 - Nehalem 6.50 6.2%
2010 - Westmere 6.60 -1.5%
2011 - Sandy Bridge 6.40 3.1%
2012 - Ivy Bridge 6.30 1.6%
2013 - Haswell 5.80 8.6%
2014 - Broadwell 5.35 8.3%

Can't you compile the data to a nice graph or so?

 

[Hulk]

Yes. How do I link to an image file?

I created and uploaded a chart but don't know how to link it to show up here?

http://www.hyperactivemusic.com/cpumark99chart.jpg

http://www.hyperactivemusic.com/cpumark99.jpg

http://www.hyperactivemusic.com/cpumark99percent.jpg


I remember that the old celerons with the full speed integrated L2 cache did better than the PII Klamath on this test because it fits in the L2. This is the main reason the PIII did better than the PII, fully integrated full speed L2 cache. P4's simply sh$t the bed. It's kind of funny how after Willamette and Northwood Intel doubled down with Prescott and increased the pipe to 31 stages! Stalls and mispredictions really killed it.

Because this benchmark excludes multiple cores, trips to main memory, and even clockspeed the way I've computed it, as well as new instructions it should be the toughest test for new cores. How much more efficiently can legacy code be executed without the improvements I mentioned above. You have to give Intel some credit for continuing to eek out 8% improvement over the last two generations. Of course I'm looking at the glass half full. Others will look at it half empty and see IPC improvements stalled since Conroe even all those little improvements have added up to a 30% IPC increase from Conroe to Broadwell.

 

[ninaholic37]

Yes. How do I link to an image file?

I created and uploaded a chart but don't know how to link it to show up here?

http://www.hyperactivemusic.com/cpumark99chart.jpg
http://www.hyperactivemusic.com/cpumark99.jpg
http://www.hyperactivemusic.com/cpumark99percent.jpg

You forgot a significant milestone in your graphs:

http://forums.anandtech.com/showthread.php?t=2357190&page=2

2004 Pentium M - 8.2

Nice to see that Broadwell is still improving. Makes my Dosbox speeds faster. :thumbsup:

 

[Hulk]

You forgot a significant milestone in your graphs:

http://forums.anandtech.com/showthread.php?t=2357190&page=2

2004 Pentium M - 8.2

Nice to see that Broadwell is still improving. Makes my Dosbox speeds faster. :thumbsup:

Yes you're right. The Pentium M was monster and the father of Conroe.

I wonder if that Pentium M you linked to was Banias or Dothan?

 

[alcoholbob]

Was able to score a new 2015 MBA for $750 from a buddy who works at a local retail store. It's got a HD6000 so hopefully it can run Diablo 3. Although I wonder if this is too much IGP for this laptop. According to core temps this cpu idles at 85C and load temps are 97C just from installing Office 365. Lmao I wonder if it can survive 15 minutes of gaming without overheating.

 

[Dufus]

Sounds like something is very wrong there, perhaps the heatsink is not in proper contact with the CPU.

 

[alcoholbob]

Sounds like something is very wrong there, perhaps the heatsink is not in proper contact with the CPU.

It's a fanless cpu that boosts to 2.7ghz and has the Iris Pro IGP chip. I mean Apple stuff usually runs hot doesn't it? Their ACD monitors were furnaces because they always refused to install fans or vents to make them completely silent.

 

[Brunnis]

It's a fanless cpu that boosts to 2.7ghz and has the Iris Pro IGP chip. I mean Apple stuff usually runs hot doesn't it? Their ACD monitors were furnaces because they always refused to install fans or vents to make them completely silent.

No, it doesn't have Iris graphics. That's the 6100 and 6200. It's also not fanless. I've owned two MBAs (Sandy Bridge i7 and Haswell i7) and those idle temps you're getting are crazy... Load temps will be high, like most ultrabooks, but it will of course not overheat.