Haswell model specs leaked

[riva2model64]

HD4000 was already pretty fast for an Intel integrated graphics chip, being able to run Skyrim on Medium. Granted, not as fast as Trinity, but still shockingly fast.

A Laptop equipped with an HD 4600 would be decent indeed.

 

[Idontcare]

The IPC increases between those generations look too low to me from actual reviews. Those numbers you posted might come from Intel but I doubt they cover as many programs as professional reviewers do. The general consensus on our forums for years has been about a 20% increase in IPC for Nehalem (although I see it's more now in the latest apps) and 14-16% for SB over Nehalem. Reviews back this up.

Sandy Bridge --> Nehalem is not 7%, but 15-16%
Core i5 760 2.8ghz = 100%
Core i5 2500K 2.8ghz = 115%
or
Core i7 860/930 2.8ghz + SMT = 100%
Core i7 2600K 2.8ghz + SMT = 116%
http://www.computerbase.de/artikel/prozessoren/2011/test-intel-sandy-bridge/46/

Secondary source confirms that it's around that range:
i5 2500K 2.8ghz is faster than i5 760 2.8ghz by 14%
http://ixbtlabs.com/articles3/cpu/sandybridge-core-vs-lynnfield-p2.html

Nehalem/Lynnfield was actually a much larger increase in IPC over Kentsfield/Yorkfield than SB is over Nehalem.

"5 - 10% increase in general application performance at the same clock speeds as Penryn. Where Nehalem really succeeds however is in anything involving video encoding or 3D rendering, the performance gains there are easily in the 20 - 40%"
http://www.anandtech.com/show/2658/20

Using the same source at Computerbase:

Core 2 Quad Q9550 2.8ghz = 100%
Core i5 760 2.8ghz = 122%
Core i7 930/870 2.8ghz = 127%

So about 22-27% without/with HT for Nehalem. This is somewhat skews since Kentsfield/Yorkfield perform very poorly in encoding and rendering and neither has HT, which can swing the average greatly in multi-threaded apps.

If Haswell IPC increases just 10%, it would be the worst increase in IPC from a new Intel architecture in 5 years. Although if they fix the solder issue, 10% IPC and 5.0ghz+ easy overclocks for Haswell on air without delidding, it wouldn't be that bad, but obviously as others have noted the pace of CPU performance at Intel has slowed down tremendously. If they don't fix the solder at all, it'll be a yawn as a 10% increase in IPC alone is nothing to write home about for existing SB/IVB users.

I'll even go as to say that for a lot of people people rocking i7 860/920 @ 3.9-4.0ghz+, there is still very little reason to upgrade the CPU for games outside of MMOs (WOW) that are insanely CPU dependent or if you happen to have 2 high-end GPUs where CPU bottleneck shows up. Otherwise, you are just wasting $. GPU bottlenecks will only be exacerbated as next generation consoles launch and games made for DX11 from the ground-up are made. Far Cry 3 is going to be nothing compared to next gen engines and by that point cards like GTX680 will be choking to death and it won't matter at all if you have a 4ghz Nehalem or a 10ghz Haswell.

http://www.guru3d.com/articles_pages/far_cry_3_graphics_performance_review_benchmark,7.html

The main draw for Haswell will probably be improvements in very specialized programs (Monte Carlo simulation due to Random Number Generator) or some other programs that heavily benefit from new instructions. For the general public gaming or using office apps, it will be a worthless speed upgrade given the cash outlay for a new mobo+CPU. I can see enthusiasts upgrade as usual as they enjoy playing with new parts and OCing, but like SB/IVB were over Lynnfield/Nehalem, I don't expect any earth shattering performance increases over Core i7 2600K OC, maybe not even over i7 920 @ 4.0ghz. Even the Z87 chipset isn't looking to bring any new cool features. Z77 also has Thunderbolt if you want that and PCI Express/DDR4 are still years away. LGA2011 also looks lame as IVB-E is the only thing it has going by Q3 2013. All the excitement of PC upgrades now rests in the GPU and SSD space.

Unless I'm mistaken, the performance numbers you are citing and linking are based on multithreaded applications...which is not what the Intel graph is about (single-threaded IPC).

Am I missing something?

 

[Maragark]

HD4000 was already pretty fast for an Intel integrated graphics chip, being able to run Skyrim on Medium. Granted, not as fast as Trinity, but still shockingly fast.

That's actually Llano beating the i7 there, not Trinity.

 

[A5]

I'll even go as to say that for a lot of people people rocking i7 860/920 @ 3.9-4.0ghz+, there is still very little reason to upgrade the CPU for games outside of MMOs (WOW) that are insanely CPU dependent or if you happen to have 2 high-end GPUs where CPU bottleneck shows up. Otherwise, you are just wasting $. GPU bottlenecks will only be exacerbated as next generation consoles launch and games made for DX11 from the ground-up are made. Far Cry 3 is going to be nothing compared to next gen engines and by that point cards like GTX680 will be choking to death and it won't matter at all if you have a 4ghz Nehalem or a 10ghz Haswell.

A 20% IPC increase from Nehalem to Haswell is probably enough to upgrade, even if you only hit the same clocks. Especially if you're like me and got a crappy OC on their 920

 

[WhoBeDaPlaya]

A 20% IPC increase from Nehalem to Haswell is probably enough to upgrade, even if you only hit the same clocks. Especially if you're like me and got a crappy OC on their 920

Seriously? Just 3GHz on a 2.66GHz stock chip?!

 

[tweakboy]

Remember Dukes of Haswell, I mean hazards.........

their going to come out with 4 core 8 thread at low tdp in late 2013. Prior to that their releasing the mobile versions and the 2core 4 thread ones,,,,,,

I wonder how Ivy Bridge E will fair. That is my next cpu after the one I just got , give her couple years,, once I need more cores then....

 

[ShintaiDK]

Remember Dukes of Haswell, I mean hazards.........

their going to come out with 4 core 8 thread at low tdp in late 2013. Prior to that their releasing the mobile versions and the 2core 4 thread ones,,,,,,

I wonder how Ivy Bridge E will fair. That is my next cpu after the one I just got , give her couple years,, once I need more cores then....

You make no sense again...

 

[Shephard]

ok so it has better integrated graphics. don't mean anything for us gamers.

what about the actual core performance?

I don't see 3570k equivalent there. I assume it would be 4570k @ 3.4ghz.

how much faster than my 3570k I just bought.

 

[ShintaiDK]

ok so it has better integrated graphics. don't mean anything for us gamers.

what about the actual core performance?

I don't see 3570k equivalent there. I assume it would be 4570k @ 3.4ghz.

how much faster than my 3570k I just bought.

10-15% in unoptimized code. Can be above 100% in optimized code. I expect Linpack for example to more than double.

Everything is expanded to 256bit width, 2 extra issue ports, AVX2, TSX etc. And a lower platform power consumption as well as a bonus. Plus improved overclocking flexibility.

 

[Shephard]

well I hope my 3570k was a good choice there's no going back now.

 

[A5]

Seriously? Just 3GHz on a 2.66GHz stock chip?!

It used to do 3.6, but a BIOS update nuked any good OCs I got out of the board. I'm actually running at 2.7 GHz right now, because I can't POST with any bclk over 135.

 

[nsKb]

He said exactly that in slightly different words.

No, he said mounting a heatsink with TIM to a bare die trumps mounting the heatsink with TIM to a soldered on IHS. I asked for his numbers. Unless the TIM's thermal conductivity was pretty damn close(lets say within an order of mag) to the solders the soldered on IHS will win.

That would be a stupid thing to do unless A) Intel is looking for future lawsuits B) The heatsink is flexible as rubber

If it contracts and expands the fragile die would break the first day you get your PC.

I was not actually suggesting that Intel or any other manufacturer pursue this option.

 

[Idontcare]

No, he said mounting a heatsink with TIM to a bare die trumps mounting the heatsink with TIM to a soldered on IHS. I asked for his numbers. Unless the TIM's thermal conductivity was pretty damn close(lets say within an order of mag) to the solders the soldered on IHS will win.

Look at thermal conductivity of Indigo Xtreme, Liquid Ultra, Liquid Pro, IC Diamond, NT-H1, etc.

But I respectfully disagree with your general premise.

If you took your heatsink cooler and used TIM to mount it to the die, the contact area is very tiny (160mm^2 for i7-3770k) and as such the pressure on the TIM will be quite high, making the TIM thickness rather thin.

If you took your heatsink cooler and soldered it to the die, only to then proceed to use a saw and cut the heatsink block clean-through at some distance above the die and then re-attach the sawed off heatsink block with regular TIM the surface area will be much greater and the result thickness of TIM will be as well.

To get the TIM to be the same thickness between the IHS and the HSF as it ends up being between the die and the HSF in a bare-die mounted application, the mounting force itself will have to be increased by exactly the same amount as the ratio of surface areas involved.

The i7-3770k IHS is ~29mmx29mm (841mm^2) in surface area, making the delta between it and the die = 5.26x

That means whatever force I apply to the HSF to mount it to the die, squeezing thin the interfacial layer of pliable TIM, I would have to apply ~5 times as much mounting force to result in the TIM between the IHS and the same HSF be the same thickness. (and as we learned, the thickness of the TIM is what makes nearly all the difference)

In practice that just doesn't happen. We end up applying as much mounting force on the bare die mount as we do with the IHS mount, so the thermal conductivity of that heat transfer interface to our cooler is going to be quite superior.

I suspect the engineers who design mobile systems know this as well, which is why mobile chips don't have an IHS. It is all done by bare-die mounting, in an application where thermal conductivity is at a premium to that of the desktop.

 

[IntelUser2000]

I'm guessing that the 2W increase in Haswell is not due to existing integration, but to account for future, whether its the on package memory or southbridge. They did it with Nehalem/Westmere generation mobile chips. Even though the high end quad cores omitted the integration of iGPU, the increase in TDP compared to Penryn was still the same 10W to account for the Northbridge integration. The thing is, iGPU-less chipsets had few watts less TDP than enabled ones, so why the same increase?

This is a way of "cheating" around the TDP limit. In practice, Northbridge and Southbridge(itself even more than Northbridge) are harder than the CPU to reach to TDP levels. If you look at datasheets, you'd see that TDP is for when nearly every feature in the chips are being used.

By integrating the Northbridge and the Southbridge, and increasing the package TDP by simply adding the numbers up, you get more potential for power on the CPU cores, and thus higher clock. You do it before the actual integration on the "Tock" so the gain seems greater, before its there on the "Tick".

The system designers likely also have lower costs on the Tick since same cooling can be used. Intel could have chose to increase TDP in one generation, and lower it in the next, only to do it again. But they don't.

 

[Ben90]

I bet you could get some pretty sweet thermal conductivity numbers if you printed the die strait onto the heatsink.

 

[frozentundra123456]

I'm guessing that the 2W increase in Haswell is not due to existing integration, but to account for future, whether its the on package memory or southbridge. They did it with Nehalem/Westmere generation mobile chips. Even though the high end quad cores omitted the integration of iGPU, the increase in TDP compared to Penryn was still the same 10W to account for the Northbridge integration. The thing is, iGPU-less chipsets had few watts less TDP than enabled ones, so why the same increase?

This is a way of "cheating" around the TDP limit. In practice, Northbridge and Southbridge(itself even more than Northbridge) are harder than the CPU to reach to TDP levels. If you look at datasheets, you'd see that TDP is for when nearly every feature in the chips are being used.

By integrating the Northbridge and the Southbridge, and increasing the package TDP by simply adding the numbers up, you get more potential for power on the CPU cores, and thus higher clock. You do it before the actual integration on the "Tock" so the gain seems greater, before its there on the "Tick".

The system designers likely also have lower costs on the Tick since same cooling can be used. Intel could have chose to increase TDP in one generation, and lower it in the next, only to do it again. But they don't.

Sort of off topic, but not really I guess. Just yesterday, TH published leaked specs of the Haswell mobile chips. The TDP actually goes up a couple of watts there also on the quads.
I am confused. I though Haswell was supposed to be this great advance for tablets/ultrabooks with great power savings. Where is it??

 

[Idontcare]

I bet you could get some pretty sweet thermal conductivity numbers if you printed the die strait onto the heatsink.

Phonon-phonon coupling would truly be maximized with such an approach, especially so if they manage to do it via epitaxy methods.

You would then have thermal conductivity that is equivalent to the bulk itself, sans interface resistivity.

Sweet indeed :thumbsup:

 

[IntelUser2000]

Sort of off topic, but not really I guess. Just yesterday, TH published leaked specs of the Haswell mobile chips. The TDP actually goes up a couple of watts there also on the quads.
I am confused. I though Haswell was supposed to be this great advance for tablets/ultrabooks with great power savings. Where is it??

But they are quad core and standard voltage chips. Those aren't for Tablets/Ultrabooks On the traditional form factors, its probably better to get even an ounce of extra performance anyway.

I think their long term focus is to have Tablets take up the lower end of computing, Ultrabooks on the mainstream, and quad core chips pushed further for desktop replacement and ultimate gaming laptops. The gaming laptops don't need sacrifice in performance for "lower power".

 

[fixbsod]

Ok this is a slight aside but I did not want to make a new thread.

Whatever happened to the days when intel / AMD would introduce a CPU core and then continue its development with faster and improved specs? IE--when Pentium came out it was Pentium 60, then 66, 75, 90, 100, etc etc etc plus improvements. Nowadays its like ok here's Ivy Bridge -- ALL AT ONCE. Really? How did intel make the top and bottom end all at once? Is this the whole tick-tock schedule? Do they take what would be incremental upgrades and roll it all up up into new cores (like instead of 200 patches its a Service Pack?).

It still seems like SOME stuff trickles out -- ie 3970x just came out and 2700k came out way after 2600k but nothing beyong 3770k and most of these trickles are just better binned top ends rather than anything substantive.

 

[IntelUser2000]

Whatever happened to the days when intel / AMD would introduce a CPU core and then continue its development with faster and improved specs? IE--when Pentium came out it was Pentium 60, then 66, 75, 90, 100, etc etc etc plus improvements.

In one word, Volume.

If you want it explained, here it is:

Back then Intel shipped lot less CPUs then they do today. You want your investment to make sense, so you won't make another frequency steps just to sell maybe few thousand more. Volume allows lot more speed grade to make sense, because each can sell in hundreds of thousands, or even millions.

Performance was so dominant of a factor back then, that increasing power use by 2x or 3x in one generation didn't matter. So you had rapid increases.

Another is that how CPUs are differentiated are changed. At one point, it was purely about performance. But nowadays its lot of things, graphics, Hyperthreading, cores, virtualization, security, etc etc.

Summarized:

Pentium came in at 60MHz and quickly scaled clock speeds. If we apply how Intel uses Ivy Bridge but back with Pentium, here's how it might look at launch.

40MHz, 45MHz, 50MHz, 55MHz, 60MHz(top end)

That's how it works today. They just introduce more lines that succeed each ad every previous ones.

 

[Saylick]

Ok this is a slight aside but I did not want to make a new thread.

Whatever happened to the days when intel / AMD would introduce a CPU core and then continue its development with faster and improved specs? IE--when Pentium came out it was Pentium 60, then 66, 75, 90, 100, etc etc etc plus improvements. Nowadays its like ok here's Ivy Bridge -- ALL AT ONCE. Really? How did intel make the top and bottom end all at once? Is this the whole tick-tock schedule? Do they take what would be incremental upgrades and roll it all up up into new cores (like instead of 200 patches its a Service Pack?).

It still seems like SOME stuff trickles out -- ie 3970x just came out and 2700k came out way after 2600k but nothing beyong 3770k and most of these trickles are just better binned top ends rather than anything substantive.

This is just an opinion so someone with better understanding of the chip industry should pitch in:

I think it does have to do with Intel's tick-tock cadence, but I also have a hunch that the reason why you see this trend has to do with the state of the industry: you mention AMD/Intel in the past would release one chip followed by a succession of chips, each faster than its predecessor, and you mention that today, you see Intel releasing a whole line-up at once. Perhaps this is a result of Intel's dominance in the chip industry in that they simply do not feel the pressure from AMD to be releasing chips as quick as they used to. Therefore, they just sit back, do some (and by some I mean billions of dollars worth) of R&D, tapeout a chip, then work on producing enough supply for a successful launch. Naturally, the production process produces high and lower binned chips, which makes it easy for Intel to release multiple SKUs at launch. Had AMD remained competitive, I presume Intel would be more pressured to release their chips as soon as they could, as the competition would always try to one-up them. In other words, the days of one-upmanship are over and so is the pressure to be constantly rolling out with faster and faster clocked versions of the same chip. Of course, another factor to consider is that lithography and chip fabrication tech is extremely more difficult to improve in today's time than in the past, which would lead to less frequent updates to a chip.

 

[fixbsod]

Yeah, I got a feeling competition, or rather the lack thereof, has the most to do with intel's willingness to release an entire line up (Sandy Bridge, IB) all at once. Makes me wonder how much intel is sitting on its hands and we miss out just because of this.

Guess this trend to more mobile and smaller platforms will give intel a kick in the pants as this is an area they've been lagging in.

 

[Idontcare]

It still seems like SOME stuff trickles out -- ie 3970x just came out and 2700k came out way after 2600k but nothing beyong 3770k and most of these trickles are just better binned top ends rather than anything substantive.

A certain amount of it is simple business decisions to manage the rate of obsolescence, no question.

But a certain amount of it is also due to the realities of the production node inside a fab.

Back in the day, the days you are thinking of anyways, a node would be transferred from the development team to the production team well before the node had actually hit its parametric specs. The Idrive spec for example might have been 0.1 mA/um but the node in reality delivered maybe 0.09 mA/um...so the production team would be tasked with making due in producing chips with lower drive current (lower clockspeeds) and similtaneously they would endeavor to tweak the process and further optimize the Idrive values to boost the clockspeeds.

In laymens parlance we would refer to this as "node maturity".

That part of a node's maturation no longer really happens inside the production fab anymore, the nodes aren't allowed to transfer out of the hands of the development team until they get much closer to their spec'ed values. That means the clockspeeds coming out of the production fab from day one are pretty much spot-on with what they were designed to be.

No real opportunity is left on the table for improvements after the fact.

Now of course the production engineers do still work on tweaking the process, and we laymen still refer to that as "node maturation" but the tweaking being done is more for boosting functional yield than for boosting parametric yield. Getting more sellable chips per wafer, but not necessarily faster chips.

Nowadays the big opportunity for improving clockspeed distributions and power-consumption comes from new mask steppings. But even there things have appeared to slow down, but only because Intel spends far more time iterating the mask sets before production release these days.

In other words, unlike in years past, we consumers are no longer given the opportunity to be early-adopters and buy ES chips that are rushed to market and labeled as a production stepping.

Nowadays the ES chips really do remain ES, and the process node really does remain in development until it functions as intended. So what we get to buy as consumers is a truly polished product, both in design and in process maturity.

 

[frozentundra123456]

But they are quad core and standard voltage chips. Those aren't for Tablets/Ultrabooks On the traditional form factors, its probably better to get even an ounce of extra performance anyway.

I think their long term focus is to have Tablets take up the lower end of computing, Ultrabooks on the mainstream, and quad core chips pushed further for desktop replacement and ultimate gaming laptops. The gaming laptops don't need sacrifice in performance for "lower power".

True, they are not the ULV chips. And the clocks (turbo) were pretty high. If the performance is only about 10% better at the same TDP though, it still doesnt seem like a very big jump in performance per watt. I thought haswell was supposed to bring something like 40% improvement, in not absolute performance, but performance per watt. Maybe the power is going to the igp. Depending on how adequate the igp is that could be good or bad. It is good if the igp is really a big improvement. However, if the igp eats more power but still is kind of "almost but not quite", (which it what I expect) it seems a waste.

Weren't they supposed to get a 10w chip with the performance of a current 17 watt chip?
Guess we will have to see what the ULV chips bring, but I am wondering if Intel has sort of hit the wall with x86 and reducing power consumption. Weren't SB quads 45 watts also? That is 2 generations with little improvement except in the IGP.

 

[cbn]

Guess we will have to see what the ULV chips bring, but I am wondering if Intel has sort of hit the wall with x86 and reducing power consumption.

Hopefully someone with more knowledge than me will chime in, but I have been under the impression ULV is built on Intel's "SP" (Standard Power) process, rather than LP (Low Power) or ULP (Ultra Low Power).


http://www.eetimes.com/electronics-...ate-SoC---how-low-can-you-leak--?pageNumber=0

If, in fact, Core ULV is built on SP, that should leave Intel a good deal more room for lowering power consumption at some point in Core's future. (Notice the very large differences in leakage in the third chart for "SP", "LP" and "ULP" on the 22nm node.)

As an analogy think of how much TDP Intel was able to lower on 45nm Pine Trail atom by going with low leakage 45nm for Oaktrail. (TDP dropped from 5.5 watts for the single core 1.66 Ghz Pine Trail to 3 watts for the single core 1.5 Ghz Oak trail.