Forget Intel Haswell, Broadwell on the Way

The U CPU vs dual core M CPU performance is definitely converging. The M series have the advantage of a faster "base" clock speed, but turbo can run close between the two. The M CPU also gets the better GPU. But in general, unless you need to do frequent heavy loads, the U CPU will suffice perfectly fine. Check out the difference between the i7-4500U vs i5-4330M here: ARK | Compare Intel® Products

 

it should be noted, that it is highly dependent on what you do. If you have any program that is heavy single threaded like Opera....you want to get the fastest one you can. Do research into your habits and what you use and how you use it. Also just completely ignore the highest "turbo speed" it is a pure shame. Whatever is the highest turbo speed on the lowest core is the number you want to look at. If it claims to be 3.6GHz 1 core and 3GHz on 2 cores the best you will get is 3-3.1GHz

 

Not sure if I understand what you are saying. For single threaded applications my 4700MQ spends most of it's time at the highest turbo.

 

Intel on Track with “Broadwell”, Production This Quarter

 

with my 920xm and 3720qm i have never seen it use single thread boost speed. I have even tried to force it with no luck. I ca set it at 4.0/3.8/2.6/2.6 and it'll run at 2.6 give or take

Anyways i throw in a little extra goodie showing my 3720qm at a magical 43x lol

Anyways my 3720qm works a little different than the 920xm. if i put 27 27 16 16 single thread would run at like 16-19 but on the 3720qm it appears to jus trun at 3.4Ghz no matter how you mess with it.

EDIT: i tried 40 26 26 26 and it was running at 34 :/

None the less i say single turbo is a pure lie.

 

It's also worth taking into consideration that the quad-core i7s have a much higher thermal envelope, which is not the case with the U-series dual-cores. With the quads, you may not see maximum 1-core Turbo speeds, but you DO see maximum 4-core Turbo speeds. As long as the systems in which they're installed have halfway decent cooling, the 45-47W TDP quad-cores will continue to run at speeds higher than their nominal base speed during most workloads (I'm not talking about stress testing with Prime 95 here, just heavy real-world usage).

The U or Y processors have a much more constricted Turbo setting, and they have a tendency to downclock to avoid exceeding TDP. So in heavy real-world usage, a U series processor cannot handle Turbo speeds for very long without exceeding TDP and needing to downclock back to nominal speed or below.

Anandtech did a comparison on this and came up with the results I just said. So, while turbo speeds may be hard to maintain in low-voltage processors, it doesn't seem to be an issue in the quad-cores.

 

Better yet it took half the world to realize that ULV throttle in games even though i started it like a year before or sooner than anandtech figured it out -_- If you plan on putting heavy load CPU and GPU load on a ULV dont bother upgrading. I haven't touched a recent ULV but my lowest level bottom of the barrel SB ULV uses like 16 watts for CPU and 16 watts for GPU so if you play a game expect 50% CPU and 50% GPU....well more like 60-70% because lower freqs require lower voltage but another topic though

The only advantage a high end ULV give you if 1-5s burst if even that before they are forced into a lower TDP. The day intel lets you play i with a ULV with its actual TDP it will be nice. If you can get a non ULV dual core you are better off. Just pop up XTU or TS and throttle it when you need battery life.

broadwells idle TDP* will be interesting. You might just rather get a quad core and throttle it and call it a day.

 

Forget haswell because broadwell and Rockwell is the way to go????? Are you nuts??? Why wait for Rockwell when Maxwell is going to be out in 2069 assuming you are still alive??? The hell with Maxwell, might as well wait for Bowel or Powell by the next century.

 

-_- .........??????......?????<----------this is me totally lost

 

Ah I see. There are going to be background tasks running now and again meaning you'll be using more than 1 core so for that time your multi is going to drop and will be more noticeable with a bigger gap such as 40,26.

I run 34,34,35,36 so not such a big gap.

Linpack single thread limited to 34x

45.98GFlops at 10000

Linpack single thread unlimited

48.16GFlops at 10000

Since Linpack scales quite well the effective gain from 34x would be 48.16 / 45.98 x 34 = 35.6x. So not too bad IMO.

Linpack would be a great comparison test for Haswell / Broadwell. Not only processing efficiency but power handling too.

 

The thread title is indeed a joke on "there's always something new coming," but the point of the thread is more about comparisons and speculation regarding (mostly) revealed products. Obviously, in practice everyone buys the best system available at the time. Your post showed a lack of understanding the humor in the hyperbole.

 

Well, there's no issue in waiting if you don't plan on using the new stuff for a while. I'm only going to PC gaming for about a week during the spring break for the entire spring semester as an engineering college student, so getting a new gaming laptop now would be a waste, even though my 4-years old light gaming laptop is clearly insufficient to maintain 30+ FPS.

And by the time the spring finals are over, hopefully there's something new.

 

So broadwel won't support DDR4 for the mainstream, which is a bit of a let down but what are the real advantages of having DDR4? I am talking about performance aspect and not voltage. I know the advantage for mobile is the lower voltage but is there a tangible gain for DDR4 in the desktop market?...besides the integrated GPU. I know integrated GPUs will have a large boost with double the speed for games but for desktops with full blown GPUs does it really matter?

Also a bit of a bummer that the new PCIe standard won't be part of it :/

 

For some workflows.

It'll matter as SSDs and CPUs get faster.

 

how would SSDs play into ram speeds mattering?

 

When the storage medium starts to approach the speed of the RAM - the RAM will become bottlenecking the speed of the system as a whole.

 

yea but that isn't happen in any near future........ Unless you have a ramdisk PCIe card or multi PCIe cards in raid 0 -_- Never will see SSDs reach ram speed unless you are talking about some wacky scenario where you have like 8 PCIe 3.0 16x things in raid.

 

It doesn't have to be anything crazy like that. The theoretical speeds are not where the limits are; they are always, effectively much lower.

 

According to numbers by Storage Review, 1600 Mhz RAM is approximately 14x faster in sequential read, 12x faster in sequential write, 10x faster in 4k read (at low queue depth), and 4x faster in 4k write (at low queue depth) compared to the fastest consumer SATA SSDs. Seems to me that SSDs still have a ways to go before they catch RAM speeds. Now, I admit that upcoming switching to SATA Express and more widespread use of M.2 sockets is going to help SSDs close the gap, but I'd guess it's still going to be a few years at least before SSDs are competitive with RAM in most types of storage performance.

 

i was googling around a little bit for numbers and didn't find anything concrete just a bunch of FPS BS. I should install sisoft sandra and test. If i recall my DDR 3700 or whatever was like 8500MBps (dual channel) when i was a kid and no SSD touches that and that was 15 years ago :/

Well i found one place was saying 1600 is ~7000-9000 MBps i assume single channel.

Now sure why i remember my Ram being 8500 or so MBps dual channel :/ Granted accessing my crap memory and 15(maybe more like 12) years ago at that lol

 

I checked Wikipedia about Broadwell and it still doesn't support faster DDR3 than 1600MHz. Is this normal guys? I would expect to work these processors at least with 1866MHz, rather 2133MHz modules.

 

mainstream wont have DDR 4 until skylake. The only real bummer in my eyes is the fact it'll hurt IGPU badly

 

Hopefully the models with an L4 cache will expand somewhat with Broadwell to compensate.

 

For me DDR3 is ok, but why is only 1600MHz supported? I d like a Core i7 ULV with 2133MHz Ram

 

yea higher freq would have atleast been a little better :/

 

Haswell supports 1866

Beamed from my G2 Tricorder

 

Than why do they write only 1600MHz? or 1866MHz need modded bios?

 

Beamed??? is a tricorder a phaser now? Shouldn't it be "Scanned by my G2 Tricorder"?

 

No you scan with the tricorder to get information, you beam to send. Get your Vulcan ears straight!

Beamed from my G2 Tricorder

 

How about transmitted....better than beamed -_-

 

I like beamed, I like it a lot.

 

if we are going to get all trekkie we need to at least be accurate -_- What trekkie does what HT is doing :/

 

Transmit doesn't sound Trekkie enough, beamed does. We're talking semantics over a theoretical technology. My Tapatalk sig stays!

 

brings...

 

and i bring All the soul energy of the universe and blow you beyond kingdom come

 

What the hell. Thanks a lot Intel
(and you who buy iPad`s instead of PCs)
(and Microsoft for making Windows 8 which is hated by many)

Intel postpones Broadwell availability to 4Q14

 

bullocks....hogwash...garbage....balderdash....you name it

No one bought haswell because there was no point -_- Only reason to buy haswell was IRIS and that was like 3 models with the real thing so no damn point. Awesome...milking the market even further.

 

Well yeah. We got what? 5-6% better clock vs clock boost with Haswell?

I am mostly dissappointed because worst case we wont get new notebook models until motherboards that support Broadwell is available. Which is pretty close to launch.
Ugh,

Greedy Intel post poning to sell out their Haswell`s :/

 

with all the delays over the last 4-5 years we are at least 1 gen maybe 2 gen behind. SB was delayed IB was IIRC and wasn't Haswell? I know we are at least 1 full year maybe 2 years behind

 

OK, fine. Then produce less Broadwells and just hurry up and get to Skylake and DDR4. I don't care about Broadwell's improvements, from what we've heard, I'm more pissed about Skylake.

 

i am more interested in broadwell due to it is going to be an actual good boost in CPU due to the die shrink. The lack of DDR4 is really troublesome because i was looking towards an IRIS SOC for an ultrabook but with the lack of DDR4 it does kill it a bit but it will still be a great CPU compared to what is out...only advantage to skylake is DDR4 IIRC that is noteworthy for mobile

 

Have we seen anything to state that Broadwell is going to actually increase CPU performance in a manful way and break from most die-shrinks?

 

In laptops where tdp and heat are the issue die shrinks always win

 

Without a change in materials smaller die means smaller mass which means thermals are going to be worse. With that in mind IMO the only path to substantially increase performance will be more cores.

I bought Haswell laptop as an upgrade from C2D. Some replace their laptops on a yearly basis, I hope to use it for a least 5 years so the new instructions are good to have for when software eventually catches up to provide some decent performance in the years to come.

For example.

I think you'll have a hard time getting such throughput on earlier chips.

 

Will it support 16G DIMMS?
http://forum.notebookreview.com/har...odimms-when-will-they-here-6.html#post9562642

 

Smaller process = less energy consumed = higher performance at same power or same performance at lower power.

 

thank you meaker...been busy haven't been able to post that

 

For the same amount of transistors maybe but usually a shrink will mean more transistors and this is where the performance increase comes from.

Now back to the heat problem.

The Haswell paradox: The best CPU in the world… unless you’re a PC enthusiast | ExtremeTech

Or if you would prefer a quote from Intel some time ago.
AP-577 An Introduction to Plastic Pin Grid Array (PPGA) Packaging

 

Still irrelevant when your dealing with tdp in the 10-55 watt range

Even. In the desktop world they can still easily cool it...maybe in the 5-10 Nm range we may face an issue

 

I tend to agree with Dufus here. Not irrelevant at all.

As the cpu area shrinks and the specific point in that smaller surface area generates more heat per square mm; the importance to properly cool these 'low power' chips are just as challenging as when they were 100W + TDP's.