Intel Broadwell Thread

[Nothingness]

It seems the memory score is much higher: http://browser.primatelabs.com/geekbench3/compare/1846459?baseline=1834798
Faster RAM?

Also the multi-core score is higher which might be a combination of faster memory and BIOS tweaking for turbo control.

Single thread integer and FP scores are almost the same as the Dell.

 

[Enigmoid]

So Haswell is clocking to 1.6Ghz @.66vcore while Broadwell is only clocking to 1.5Ghz on .76 vcore, am I reading this correctly? I'm no expert but 1.5Ghz sounds lower than 1.6Ghz.

The Haswell system is 1/1.1 ghz on the core clocks.

 

[dahorns]

It seems the memory score is much higher: http://browser.primatelabs.com/geekbench3/compare/1846459?baseline=1834798
Faster RAM?

Also the multi-core score is higher which might be a combination of faster memory and BIOS tweaking for turbo control.

Single thread integer and FP scores are almost the same as the Dell.

Yeah, but then compare it to the 5200u scores:

http://browser.primatelabs.com/geekbench3/compare/1846653?baseline=1846459

Something just seems off.

 

[oobydoobydoo]

The Haswell system is 1/1.1 ghz on the core clocks.

Ok, so I was reading it wrong. It looks like Broadwell is averaging 1.7Ghz while Haswell is only 1/1.1Ghz like you said. For a 30W load, it looks like broadwell is able to maintain higher clocks than Haswell.


Uncore is higher on broadwell but that is to be expected as it is clocking higher.

 

[Nothingness]

Yeah, but then compare it to the 5200u scores:

http://browser.primatelabs.com/geekbench3/compare/1846653?baseline=1846459

Something just seems off.

5Y71 turbo and L3 cache are bigger than 5200U: http://ark.intel.com/compare/85212,84672

 

[Sweepr]

Holy crap, a Geekbench score for the Asus T300 chi with the 5y71 just posted:

Single-Core: 2789
Multi-Core: 5412

http://browser.primatelabs.com/geekbench3/1846459

Way above anything we've seen before and higher than most of the u-series processors. I assume something screwy happened with this run.

Perhaps it's a better implementation of Core M than previous designs? We will know soon, I really like what ASUS has been doing lately.

 

[dahorns]

5Y71 turbo and L3 cache are bigger than 5200U: http://ark.intel.com/compare/85212,84672

Well, there ya go, I did not realize that.

 

[Sweepr]

Another Broadwell-U notebook review:

NotebookCheck's Dell Latitude 14 E7450 Ultrabook Review

This time idle battery life is about on par with similar Haswell notebooks while WiFi, H.264 and load battery life is up.

It is not easy to answer the question whether Broadwell manages longer battery runtimes than Haswell when you compare differently equipped devices from several manufacturers. However, the E7450 is on par with the rivals both in the idle scenario of the Battery Eater Reader's Test (energy-saving mode, minimum brightness, WLAN off) as well as the CPU-demanding Classic Test (high-performance mode, maximum brightness, WLAN on), so there does not seem to be any noteworthy improvements.

However, we can see noticeable efficiency improvements in more practical scenarios. Our WLAN and H.264 tests (around 150 cd/m²) determine more than 7.5 hours before the battery is empty – a full business day without a socket is actually possible. This usually required much larger batteries.

The E7450 does not need active cooling in many scenarios, not only while idling, but also in many everyday applications (office, browsing). If the fan should start to spin, it is usually in the lowest rpm setting, which is hardly noticeable at around 30 dB(A). Thanks to the SSD in our review unit, access noises from the hard drive are not a problem.

This does not change at first, when you start a very CPU or GPU demanding application – the notebook is often silent for one or two minutes even under maximum load. After that, the murmur of the fan starts to increase slowly and finally levels off at around 37 dB(A).

We have already mentioned that Broadwell does not have much higher clocks than comparable Haswell models, at least when we look at the Turbo Boost. But the new generation does have two big advantages: First, thanks to the 14-nm manufacturing process, the power consumption has been reduced, so the TDP does not limit the clock so often. Intel has also been able to increase the per-MHz performance by another 5% – not a huge improvement, but still respectable when you consider the high level of the predecessor.

As expected, the Core i5-5300U can maintain the full 2.7 GHz even under continuous load for both cores, and it can even exceed a Core i7-4600U or i7-4650U in longer benchmarks.

www.notebookcheck.net/Dell-Latitude-14-E7450-Ultrabook-Review.135484.0.html

 

[witeken]

Not too bad it seems.

 

[AtenRa]

Heh, HD5500 has slower/equal iGPU performance vs HD5000 even at 15W TDP. This is becoming a huge mess with the Intel iGPU naming.

Dell Latitude 14 E7450 Ultrabook Review Broadwell HD5500 15W TDP

Asus AsusPro Advanced BU401LA-CZ020G Ultrabook Review Haswell HD5000 15W TDP

 

[bullzz]

@AtenRa - isnt that expected? 24EU vs 40EU. are you falling for the "5500 is greater than 5000. so must perform better" idea. I am actually excited that it can match HD5000. really hope MBA picks up HD6000

 

[AtenRa]

@AtenRa - isnt that expected? 24EU vs 40EU. are you falling for the "5500 is greater than 5000. so must perform better" idea. I am actually excited that it can match HD5000. really hope MBA picks up HD6000

Im not talking about me but for unaware consumers who have come to believe all those years that higher number = faster.

Also to point out, 22nm HD5000 is heavily TDP constrained at 15W TDP so i was expecting a 14nm HD5500 to be a lot faster.

 

[oobydoobydoo]

So it looks like whatever efficiency gains Broadwell makes are quickly eaten up in maintaining higher clocks for longer periods of time, which is basically what Anand said. Interesting.



Still does not look, to me, like a compelling product. Why would I buy a 2015 Broadwell laptop when I can get a 2013 Haswell that will be cheap and have identical battery life, but benchmark slightly slower? I understand much of intel's market is business consumers who don't necessarily care how efficient Broadwell is but there is still a huge market for Macs and PCs with intel products, and won't people think the same thing? I struggle to see how intel will pitch this as an upgrade worth having at all, besides saying it's "new" or lying outright.

 

[dahorns]

So it looks like whatever efficiency gains Broadwell makes are quickly eaten up in maintaining higher clocks for longer periods of time, which is basically what Anand said. Interesting.



Still does not look, to me, like a compelling product. Why would I buy a 2015 Broadwell laptop when I can get a 2013 Haswell that will be cheap and have identical battery life, but benchmark slightly slower? I understand much of intel's market is business consumers who don't necessarily care how efficient Broadwell is but there is still a huge market for Macs and PCs with intel products, and won't people think the same thing? I struggle to see how intel will pitch this as an upgrade worth having at all, besides saying it's "new" or lying outright.

Because it offers much higher sustained performance, higher graphics performance, and slightly higher peak performance, while at the same time not negatively affecting battery life, even improving it in certain use cases?

 

[North01]

Holy crap, a Geekbench score for the Asus T300 chi with the 5y71 just posted:

Single-Core: 2789
Multi-Core: 5412

http://browser.primatelabs.com/geekbench3/1846459

Way above anything we've seen before and higher than most of the u-series processors. I assume something screwy happened with this run.

That score was also achieved on Geekbench 3.2.0 (32-bit).

The Dell Venue 11 Pro 7140 with 5Y71 scored lower in Geekbench 3.3.0 (32-bit) test:

Singe-Core: 2565
Multi-Core: 4549

http://browser.primatelabs.com/geekbench3/1812753

For reference, here is the Dell Venue 11 Pro 7140 with 5Y10 running on Geekbench 3.2.2 (64-bit):

Singe-Core: 2180
Multi-Core: 4506

http://browser.primatelabs.com/geekbench3/1319713

 

[mikk]

Heh, HD5500 has slower/equal iGPU performance vs HD5000 even at 15W TDP. This is becoming a huge mess with the Intel iGPU naming.

Dell Latitude 14 E7450 Ultrabook Review Broadwell HD5500 15W TDP

Asus AsusPro Advanced BU401LA-CZ020G Ultrabook Review Haswell HD5000 15W TDP

Comparing a singlechannel device with dualchannel is a bad idea for iGPU tests.

 

[dahorns]

That score was also achieved on Geekbench 3.2.0 (32-bit).

The Dell Venue 11 Pro 7140 with 5Y71 scored lower in Geekbench 3.3.0 (32-bit) test:

Singe-Core: 2565
Multi-Core: 4549

http://browser.primatelabs.com/geekbench3/1812753

For reference, here is the Dell Venue 11 Pro 7140 with 5Y10 running on Geekbench 3.2.2 (64-bit):

Singe-Core: 2180
Multi-Core: 4506

http://browser.primatelabs.com/geekbench3/1319713

Good point, but I don't see much evidence of significant score differences between 3.2 and 3.3.

Here is a Venue 7140 with 5y71 on 3.1.5

Single-Core: 2539
Multi-Core: 4406

http://browser.primatelabs.com/geekbench3/1739600

And here is it with the 5y10 on 3.2.2

Single-Core: 1939
Multi-Core: 4098
http://browser.primatelabs.com/geekbench3/1621607

The point is that there doesn't appear to be a trend of score deflation going from 3.2 to 3.3. Thus, there isn't much reason to think that the Asus score is inflated because it was running on the older version.

 

[AtenRa]

Comparing a singlechannel device with dualchannel is a bad idea for iGPU tests.

They run Broadwell with Dual Channel as well,

Have a look at 3D Mark 11, 3D Mark (2013) etc.

The maximum clocks of the HD 5500 are slightly lower compared to the predecessor, so the performance advantage in the various 3DMarks is quite small. Even with a second memory module, the increase compared to an HD 4400 (dual-channel memory as well) is limited to only 5 to 20%. Overall, even a low-end dedicated GPU like the GeForce 820M clearly beats the Broadwell GPU.

 

[Phynaz]

Comparing a singlechannel device with dualchannel is a bad idea for iGPU tests.

Not when you have an agenda to push :awe:

Some people need to realize that when comparing products you doing exactly that. You aren't comparing one component that makes up the product.

 

[coercitiv]

Not when you have an agenda to push

Some people need to realize that when comparing products you doing exactly that. You aren't comparing one component that makes up the product.

That's true for one product, but once more and more products start telling the same story it gets increasingly harder for any agenda to be pushed.

 

[Enigmoid]

They run Broadwell with Dual Channel as well,

Have a look at 3D Mark 11, 3D Mark (2013) etc.

As I said before <40% is a failure with a new arch + node shrink. 40% is mediocre, and 50% is good.

 

[coercitiv]

Product review on ArsTechnica: The Broadwell NUC

 

[frozentundra123456]

As I said before <40% is a failure with a new arch + node shrink. 40% is mediocre, and 50% is good.

How long has it been since any intel or AMD cpu product showed a 40% increase from generation to generation? I dont think we have seen anything like that since PIV to Conroe. So far I agree Broadwell has been somewhat disappointing, but 40% is a huge increase on a mature product like cpus. Even Intel themselves only said "up to" 30% increase in performance per watt, not absolute performance.

 

[Sweepr]

Product review on ArsTechnica: The Broadwell NUC

Finally some Broadwell 2C+GT3 numbers. This particular Core i5 model should be present in Apple's Macbook Air (2015) base configuration. Other bechmarks from the same link:

This may not be true for thin-and-light laptops with less room to dissipate heat, but despite having a "base" clock speed of 1.6GHz, the i5-5250 in the NUC can sustain a Turbo Boost clock speed around 2.4GHz indefinitely. We ran the Prime95 CPU torture test for more than 20 minutes, and the clock speed hung out somewhere between 2.37 and 2.45GHz for the entire time. Even the Haswell NUC would ramp down to about 2.0GHz under the same kind of sustained load.

Once again Broadwell-U's turbo is rock solid, which means it can sustain ''peak'' CPU performance better than Haswell-U.

Here, Broadwell is definitely a better performer than Haswell. In most of our benchmarks, the HD 6000 is between 20 and 30 percent faster than HD 5000. It often doubles the performance of Ivy Bridge's HD 4000.

Intel's 14nm Tick offers 2x graphics performance of the 22nm Tick, not too shabby. Looking forward to HD6100/HD6200 results. I wish they tested iGPU performance for extended periods of time, I guess Broadwell also maintains its iGPU performance better than Haswell.

 

[Enigmoid]

How long has it been since any intel or AMD cpu product showed a 40% increase from generation to generation? I dont think we have seen anything like that since PIV to Conroe. So far I agree Broadwell has been somewhat disappointing, but 40% is a huge increase on a mature product like cpus. Even Intel themselves only said "up to" 30% increase in performance per watt, not absolute performance.

I was talking about GPU performance.