Haswell Core Count

[Nemesis 1]

Whats the problem here? If you want more cores buy the high end intel stuff . If you want more cores for more threads buy an amd 8 core its priced at the lower end but it has 8 cores. Argueing about haswell core count is stupid. ya want more than 4 cores buy the more expensive intel 8 cores or the cheaper amd to run all these threads some here need . Stop cring about intels performance cpus there meant for the masses . Ya want a workstation build SB-e or buy AMD . Not one person here cares what ya buy . not one

 

[Phynaz]

760 45nm to 2500K 32nm (more performance from IPC and frequency increase + less power consumption)
2500k 32nm to 3570K 22nm (more performance from IPC and frequency increase + less power consumption)

IB 22nm to Haswell 22nm (if Haswell has 10% IPC and almost same power consumption i dont see any reason for anyone to upgrade from IB at the time).

I love how you speculate about performance.

When you talk about piledriver you write things like "may be 30% faster". Write about Haswell and it's "if it's 10% faster".

If you want to be taken seriously you should tone down the bias.

 

[BenchPress]

I love how you speculate about performance.

When you talk about piledriver you write things like "may be 30% faster". Write about Haswell and it's "if it's 10% faster".

If you want to be taken seriously you should tone down the bias.

I don't think this is bias. Intel already has a very strong micro-architecture when it comes to IPC. It's very doubtful that they'll be able to increase it significantly without harming other metrics. So 10% is the best I can imagine. If you think it will be higher, please explain what sort of technology they might use to achieve that.

AMD on the other hand has a lot of opportunity for catching up. There's a lot that can be improved and a 30% increase in IPC really is within the realm of possible.

But they'd better implement AVX2 and TSX sooner rather than later.

 

[frozentundra123456]

Whats the problem here? If you want more cores buy the high end intel stuff . If you want more cores for more threads buy an amd 8 core its priced at the lower end but it has 8 cores. Argueing about haswell core count is stupid. ya want more than 4 cores buy the more expensive intel 8 cores or the cheaper amd to run all these threads some here need . Stop cring about intels performance cpus there meant for the masses . Ya want a workstation build SB-e or buy AMD . Not one person here cares what ya buy . not one

The "problem" is that the AMD 8 core is really not 8 true cores, performs poorly, and uses a lot of power. Intel on the other hand, has an efficient architecture, but has made no effort in bringing 6 core CPUs to the mainstream. Quad cores came into the mainstream very soon after dual cores, and have come down in price significantly. On the other hand, we have had mainstream quads for what, 5 years plus now, and no real progress to hexcore in the mainstream segment. I am also not convinced Haswell will bring large IPC gains except in very specific cases. If that is true we will have gone from SB to IVB to haswell with very little progress in any area except the iGPU, which is not that significant on the desktop. And for your information, I an not "crying" about anything. In fact I probably will not upgrade for a while. I just would like to see more progress in CPU performance. It seems we have really stagnated, and now even intel is more interested in the GPU than increasing CPU performance. That is great for mobile, but not for desktops.

 

[Dinkydau]

It's a big disappointment to hear that the "mainstream" haswell CPUs will only have 4 cores. I don't see why intel can't use both advantages, avx2 AND more cores for mainstream. Applications not using all the cores is not a valid argument because you aren't going to run single-threaded applications on a quad-core either. Actually if you aren't doing anything special even a quad-core is massive overkill, but those things sell. That's because there's demand. People want more CPU power. There are more than enough multithreaded applications that mainstream users could use extra speed for, in whatever form that may be, more efficiency per core or just more cores. I'm sure there would be more than enough market for mainstream 6-core and 8-core hasswell CPUs.

 

[BenchPress]

I am also not convinced Haswell will bring large IPC gains except in very specific cases. If that is true...

It is false. You should not have high hopes that Haswell will significantly increase Instructions Per Clock under any circumstance. Doing so would cost performance/Watt and this is not desirable. Instead, Haswell increases the amount of work done per instruction. AVX2 offers twice the SIMD throughput and the addition of gather and vector-vector shift makes it highly suitable for auto-vectorization which can make some 32-bit scalar code run eight times faster.

...we will have gone from SB to IVB to haswell with very little progress in any area except the iGPU, which is not that significant on the desktop. And for your information, I an not "crying" about anything. In fact I probably will not upgrade for a while. I just would like to see more progress in CPU performance. It seems we have really stagnated...

Of course there's no real progress between Sandy Bridge and Ivy Bridge. The latter is just the 'tick', the process shrink, of the former. Haswell however adds AVX2 which included 256-bit integer SIMD instructions, fused multiply-add, and gather. Just to be clear here: it would have been a great chip with just one of these, but Intel added all three! And last but not least we're getting TSX, which also consists of two pieces of revolutionary technology.

TSX is a necessity to further increase the number of cores for single-process performance scaling. Without it, threads just convoy around locks and very little work gets done.

So Haswell will stick with quad-core a little longer, but with each core doing a lot more work a lot more efficiently it's anything but stagnating!

 

[CPUarchitect]

It's a big disappointment to hear that the "mainstream" haswell CPUs will only have 4 cores. I don't see why intel can't use both advantages, avx2 AND more cores for mainstream. Applications not using all the cores is not a valid argument because you aren't going to run single-threaded applications on a quad-core either. Actually if you aren't doing anything special even a quad-core is massive overkill, but those things sell. That's because there's demand. People want more CPU power. There are more than enough multithreaded applications that mainstream users could use extra speed for, in whatever form that may be, more efficiency per core or just more cores. I'm sure there would be more than enough market for mainstream 6-core and 8-core hasswell CPUs.

Yes, there is plenty of demand for more CPU power. But the mistake you're making is that this automatically means more cores is the best answer. It's absolutely not.

AVX2 offers a lot more performance for a very small increase in power consumption. Likewise, adding more cores is a waste when synchronizing threads makes them burn a lot of power while spin looping (read: doing nothing). Therefore Intel is increasing the efficiency of the existing cores with TSX.

 

[Edrick]

People need to let go of the idea that GPUs are magically faster than CPUs. They merely have wide vector units, FMA, and gather. With AVX2 you get the exact same things on the CPU, and it already has a vastly superior memory hierarchy and much more advanced scheduling.

Haswell just can not get here fast enough.

 

[exar333]

I don't think this is bias. Intel already has a very strong micro-architecture when it comes to IPC. It's very doubtful that they'll be able to increase it significantly without harming other metrics. So 10% is the best I can imagine. If you think it will be higher, please explain what sort of technology they might use to achieve that.

AMD on the other hand has a lot of opportunity for catching up. There's a lot that can be improved and a 30% increase in IPC really is within the realm of possible.

But they'd better implement AVX2 and TSX sooner rather than later.

One of the silliest and most naive posts I have read in a long time. Don't you think AMD would have increased BD performance by '30%' somewhere in the 5+ years they have been working on BD arch? The improvements going into Haswell are 3-4x more significant than anything we will see from BD-PD. PD will have 10% more clocks, 5% more IPC, and better efficiency. This is from AMD themselves.

Hey, I love AMD and want them to succeed, but your comments are 100% wishful thinking. AMD may be more competitive in 2-3 years, but PD is no 'Conroe'. Never was, AMD has even said so.

 

[Dinkydau]

Yes, there is plenty of demand for more CPU power. But the mistake you're making is that this automatically means more cores is the best answer. It's absolutely not.

AVX2 offers a lot more performance for a very small increase in power consumption. Likewise, adding more cores is a waste when synchronizing threads makes them burn a lot of power while spin looping (read: doing nothing). Therefore Intel is increasing the efficiency of the existing cores with TSX.

I agree that more efficiency is actually a better improvement, but without extra cores intel is missing an opportunity to offer an even better product. Intel could have done both: more cores, AND better efficiency per core.

 

[frozentundra123456]

I agree that more efficiency is actually a better improvement, but without extra cores intel is missing an opportunity to offer an even better product. Intel could have done both: more cores, AND better efficiency per core.

Excellent point.

 

[2is]

I agree that more efficiency is actually a better improvement, but without extra cores intel is missing an opportunity to offer an even better product. Intel could have done both: more cores, AND better efficiency per core.

But how is that at all beneficial to them? Even with 4 cores they will have the highest end chip on the market. If you truely "need" more power than what you have, you'll either spring for the expensive stuff or make due with their quads. You certainly won't move to AMD"s 8 half-cores if you actually need power. Why release an 8 core and cannibalize their own product? Maybe when AMD's 8 core processors can consistently beat Intel's dual cores and compete with their quads, there will be more incentive for Intel to bring 6-8 core i7's into the mainstream market. Right now it makes zero business sense to do so.

 

[BenchPress]

One of the silliest and most naive posts I have read in a long time. Don't you think AMD would have increased BD performance by '30%' somewhere in the 5+ years they have been working on BD arch?

Don't call my post silly when you give the most ridiculous argument ever. Hindsight is 20/20. AMD was betting on software to be heavily multi-threaded by now. And by the time they realized that was a mistake, they had already radically changed their micro-architecture. They have limited engineering resources and they had no choice but to finish this complex project. Note that it takes Intel two years between every tock, the project timelines overlap, and they make less radical changes.

Only after Bulldozer was finished, AMD could start concentrating on IPC again. What you're suggesting is plain stupid. You can't build a pickup truck and halfway through decide you want a sports car instead...

The improvements going into Haswell are 3-4x more significant than anything we will see from BD-PD.

What IPC improvements are going into Haswell that you know of?

My post was more about the general potential AMD has for improving IPC, not so much about specific designs. That said, Haswell will be facing Steamroller, not Piledriver. And I'm not claiming the latter will be 30% faster, only that this would be within the realm of possible.

It's easier to catch up than improve an almost perfect design.

 

[BenchPress]

I agree that more efficiency is actually a better improvement, but without extra cores intel is missing an opportunity to offer an even better product. Intel could have done both: more cores, AND better efficiency per core.

There will be Haswell-based CPUs with more cores. You'll just pay a premium. How much that will be depends entirely on AMD. If Steamroller has significantly improved IPC, and clocks high, then Intel would be forced to sell 6-8 core chips at affordable prices.

 

[2is]

Don't call my post silly when you give the most ridiculous argument ever. Hindsight is 20/20. AMD was betting on software to be heavily multi-threaded by now. And by the time they realized that was a mistake, they had already radically changed their micro-architecture. They have limited engineering resources and they had no choice but to finish this complex project. Note that it takes Intel two years between every tock, the project timelines overlap, and they make less radical changes.

Only after Bulldozer was finished, AMD could start concentrating on IPC again. What you're suggesting is plain stupid. You can't build a pickup truck and halfway through decide you want a sports car instead...

What IPC improvements are going into Haswell that you know of?

My post was more about the general potential AMD has for improving IPC, not so much about specific designs. That said, Haswell will be facing Steamroller, not Piledriver. And I'm not claiming the latter will be 30% faster, only that this would be within the realm of possible.

It's easier to catch up than improve an almost perfect design.

AMD was not betting on that at all. AMD found the cheapest way they could market an "8 core" processor and hope people buy into the marketing BS. And to a certain extent it works. How many people actually know that BD's core count is more like 8 half cores which don't even equate to 4 full ones, they are quite literally 8 crippled cores. If intel couldn't pull off a paradigm shift with Itanium, AMD certainly isn't going to with BD. Nor do I think they really thought they would.

 

[AtenRa]

I love how you speculate about performance.

When you talk about piledriver you write things like "may be 30% faster". Write about Haswell and it's "if it's 10% faster".

If you want to be taken seriously you should tone down the bias.

You will find that when im talking about PD im saying up to 30% Performance increase not IPC. I dont expect more than 10% IPC increase for PD as well.

Do you honestly believe that Haswell will have more than average 10% IPC increase in legacy code??

 

[AtenRa]

Yes, there is plenty of demand for more CPU power. But the mistake you're making is that this automatically means more cores is the best answer. It's absolutely not.

AVX2 offers a lot more performance for a very small increase in power consumption. Likewise, adding more cores is a waste when synchronizing threads makes them burn a lot of power while spin looping (read: doing nothing). Therefore Intel is increasing the efficiency of the existing cores with TSX.

More cores is the best answer for more performance for day one. By the time Haswell will be released, more apps will be multhithreaded to take advantage of the increased core count than apps supporting AVX/AVX2.

Technically it might be better and more efficient to increase the work done per core but AVX/AVX2 is an ISA. It needs to be implemented in to the application and it will take years for AVX2 applications to be available in desktop.

Im not saying we dont need AVX/AVX2 but we will not be able to use that performance from day one.

 

[Lonbjerg]

Im not saying we dont need AVX/AVX2 but we will not be able to use that performance from day one.

False.

 

[AtenRa]

@Lonbjerg

You believe there will be AVX2 ready applications the first day of release of Haswell ??

 

[Lonbjerg]

@Lonbjerg

You believe there will be AVX2 ready applications the first day of release of Haswell ??

There already is:
http://software.intel.com/en-us/forums/showthread.php?t=103133

:whiste:

 

[Th3Loonatic]

You guys talk as if it were that easy designing a CPU lol. The engineers must probably be sweating blood trying to design a chip like Haswell. And here you guys are complaining that you want more cores for cheap.

 

[AtenRa]

There already is:
http://software.intel.com/en-us/forums/showthread.php?t=103133

:whiste:

One ?? When there are more than 10-20 wake me up

 

[Lonbjerg]

One ?? When there are more than 10-20 wake me up

Fallacy: Moving the goalposts.

You got slapped by reality...get over it, I don't care about your smoke&mirrors.

 

[AtenRa]

Fallacy: Moving the goalposts.

You got slapped by reality...get over it, I don't care about your smoke&mirrors.

There was one AVX application since SB and BD was released, x264 HD. After than long and we still only have that one.

I hope we will have more than one at Haswell released day but i doubt it.

 

[bronxzv]

One ?? When there are more than 10-20 wake me up

Sure, our case with Kribi 3D is arguably anectodical but note that Intel's IPP and MKL have already AVX2 paths 1 year ahead of launch so I'll say 1000s of applications will support AVX2 from day 1

Also as mentioned by BenchPress all codes using intermediate languages like OpenCL will use AVX2 at launch