Has Broadwell shrunk the performance gap between ULT i3 and i5?

With Ivy and Haswell the gap between i3 and i5 was huge because Turbo Boost easily accounted for a third of total performance, but with Broadwell's all-around higher base clocks and less impressive i5 turboing, are i3 and i5 now more similar? Or has Intel thrown another wrench into their labyrinth of segmentation strategies that once again cripples the i3 and "restores the balance"?

 

From my person experience, all of the ULV models wether it be i3, i5, or i7.. Are very similar in terms of the silicon.
Usually they have the same amount of cores, just one may enable Hyper-threading, and a slightly higher clock speed. In addition, you may get some special purpose enhancements, like VT-d in the i5 ULV etc.

There for the last few generations has no been much more than a 20% Gap between the i5 offerings and the i7 ones.

 

Who cares. ULV chips are all crap regardless and real-world sustained performance is very close between all of them due to the severely locked down TDP. You're just wasting money going for a more expensive i5 or i7 ULV with higher nominal Turbo Boost that it'll never be able to sustain under any kind of load.

 

I disagree. There are ultrabook laptops where they allow ULVs to run in full turbo mode indefinitely. CPU only tell half of the story, the other half is how the manufacturer designed the laptop and what restrictions they placed on the CPU (ideally none).

I recall a while ago there was a comparison of two different tablet/laptop combo device, one had an i3 and the other had a higher clocked i5.

The computer with the i5 had higher CPU performance, for about 8 seconds, and then it throttled to 800 mhz for much of the benchmarking.

The different brand computer with the i3 ran at full turbo for the entire two hours of benchmarking and thus pulled far ahead of the throttled i5.

 

When OEMs choose to further gimp an already locked down chip, that's called pouring salt on the wound and rubbing it in with a red-hot branding iron

 

Such as..?

 

Agreed - it's just a matter of picking the model that has the features you need and/or want. Intels' ARK site is great to find that info. I find the Core i5's are worth a little extra over the i3's because the Turbo Boost feature adds to responsiveness. If you're doing something that for a short period of time that requires a lot of CPU, an i5 can Turbo Boost its frequency and accomplish it faster than an i3. E.g. applying a filter in Photoshop or encoding an MP3 file.

That's a rather crude statement and TBH - I'd expect a little better from someone as technically savvy as yourself. Your statement is equivalent to saying a family sedan is a poor race car because its brakes and engine will overheat after five laps. There's a saying that goes: buy the right tools for the job. Using a low-wattage Intel processor like the ones being discussed in this thread for applications where they'll be running at 100% CPU for long periods of time is just a poor decision. Low-wattage processors can run at 100% for extended periods of time, sure, but as you aptly noted their sustained performance is inconsistent due to TDP limitations. They're the wrong tool. You can call anything 'crap' by putting it in the wrong scenario.

OP - back to your question - the practical difference between the Core i3 and i5 would depend on the task at hand. The i5 should be faster than the i3 in every situation; the key however is whether the difference will be noticeable e.g. matters. The difference is probably going to be unnoticeable in general usage like office productivity, but the performance gap will be there if you use the occasional application where the CPU will be stressed for brief periods of time (e.g. Photoshop). If you're trying to decide whether an i5 is worthwhile over an i3, I'd research benchmarks that are representative of the tasks you'll be running. If the i5 is 20% faster than the i3 for your applications, then it could be reasoned that the i5 is worth at most 20% more. Make it practical for yourself, though. If it takes 30 seconds for the i3 to do something (e.g. encode an MP3), the i5 would do it in 24 seconds assuming it's 20% faster. So if that six seconds means something to you ...then the 20% extra is money well spent.

Now, put that in perspective with the Broadwell i3's and i5's; the i3 is undoubtedly faster than its last gen. But it should theoretically occupy the same price point as the last gen and you're still faced with the above scenario, deciding whether the i5's performance difference is worthwhile. Using my previous example, the i3 doing something in 30 seconds while the i5 does it in 24 - maybe the Broadwell i3 does that scenario in 27 seconds. So whereas you might have opted for a previous gen i5 because you didn't want to wait the 30 seconds it took with the previous-gen i3, now you're only waiting 27 with the new i3 ...hopefully that train of thought makes sense.

Bottom line - buy the right tools for the job. Don't fuss over numbers - look at the results for your usage scenario and decide if the CPU is fast enough based on that. "I want a computer that can encode my videos in less than an hour" - go find benchmarks and see what meets your requirements.

 

Touché. I'll admit that ULV processors do have their merits, such as being able to fit in form factors unsuitable for higher TDP chips and having decent single-threaded performance due to Turbo Boost.

 

Unfortunately, with Intel's marketing machine obfuscating the true nature of their processors as much as possible and their "labyrinth of segmentation strategies" as the OP put it, things are not easy for PC buyers. I know of several people who recently bought new laptops which are actually a performance downgrade from their previous ones, and they didn't know it at the time because all they saw was the i7 on the label and also expected a brand new computer to be better than an older one. Either Intel is excessively idiotic or they are being malicious, but it shouldn't be this hard for people to figure out what it is exactly that they are buying.

I agree with octiceps that these ULV chips are all junk. The power saving features of the normal voltage chips means that under the same usage, they can consume the same power as the low voltage chips, yet have a lot more potential to get work done when necessary. With the ability of PC manufacturers to control TDP limits and thermal throttling, they can force a normal voltage processor to work perfectly fine in a small form factor, poor cooling laptop. Intel figured out how to get people to pay a premium for getting much less processor by pushing this ULV junk. The only real use I see for these processors is to put into x86 tablets.

I advise against using Intel ARK. It is less accurate and less thorough than Wikipedia or cpu-world.com.

 

As a test, I once transcoded a video on a Surface Pro (first gen) while watching CPU-Z and TMonitor, and the i5-3317U stayed between 2.3 Ghz and 2.5 Ghz for pretty much the entire ~40 minute duration at 100% CPU usage. While I would never want to do this for real (a dual core CPU running at 2.3-2.5 Ghz is still quite slow compared to a desktop i5/i7), I nonetheless came away from that experiment very impressed with the potency of Turbo Boost on i5/i7 ULV/ULT CPUs, especially when their i3 cousins like the i3-3217U and the i3-4100U completely lack it and are therefore permanently stuck at 1.8 Ghz.

So I'll have to respecfully disagree with octiceps, though I acknowledge that many laptops with ULV/ULT CPUs are poorly designed and thermally limited and that his/her experiences with those laptops may have influenced his/her views on these CPUs.

Of course, with Broadwell ULT base clocks are up across the board while the number of Turbo Boost bins on i5/i7 CPUs has been reduced accordingly so things may have changed - which is why I asked this question in the first place.

Perhaps I should add a caveat that I'm only interested in laptops that are well designed, well built, and do not suffer from cooling or other performance problems? I thought this would be an automatic expectation here on NBR given that people recommending (or more frequently, recommending against) laptops on the WNBSIB section often do so on this basis, but perhaps it's not...