Intel "Haswell" Speculation thread

Discussion of all forms of processors, from AMD to Intel to VIA.

Moderators: Flying Fox, morphine

Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 3:53 am

Is anyone else as excited about this new architecture as I am? Even if the performance only goes up by a small amount, it will introduce and support pretty much all the standards on the road map for the foreseeable future. I can't think of a better reason to upgrade, as it will likely last you many many years before retiring.

To fuel the discussion somewhat, Intel apparently aims to overtake AMD on the integrated graphics front:
http://www.dailytech.com/Intel+Says+Has ... e22002.htm

Intel has already released details on new instructions that will be included with Haswell:
http://software.intel.com/en-us/blogs/2 ... available/

The following is taken from the Haswell Wikipedia page:

    Features carried over from Ivy Bridge

  • A 22 nm manufacturing process.
  • 3D tri-gate transistors (Ivy Bridge processors and onwards).
  • A 14-stage pipeline (since the Core microarchitecture).
  • Mainstream up to quad-core.
  • Native support for dual channel DDR3.
  • 32KB data + 32KB instruction L1 cache per core.
  • 256KB L2 data cache per core and up to 32MB L3 cache shared by all cores.

    Confirmed new features

  • Advanced Vector Extensions 2 (AVX2) instruction set, also called Haswell New Instructions (includes gather, bit manipulation, and FMA3 support).
  • New sockets — LGA 1150 for desktops and rPGA947 & BGA1364 for the mobile market.
  • Intel Transactional Synchronization Extensions (TSX).
  • Graphics support in hardware for Direct3D 11.1 and OpenGL 3.2.
  • DDR4 for the enterprise/server variant (Haswell-EX).

    Expected features

  • 32 nm PCH.
  • A new cache design.
  • Support for Thunderbolt technology.
  • There will be three versions of the integrated GPU: GT1, GT2, and GT3. According to vr-zone, the fastest version (GT3) will have 20 execution units (EU). Another source, SemiAccurate, however says that the GT3 will have 40 EUs with an accompanying 64MB cache on an interposer. An additional source, AnandTech, agrees that GT3 will have 40EUs, but makes no mention of an interposer. Haswell's predecessor, Ivy Bridge, has a maximum of 16 EUs.
  • New advanced power-saving system.
  • Base clock (BClk) increase to 133 MHz.
  • 128 bytes cache line.
  • Execution trace cache will be included L2 caching design.
  • Fully integrated voltage regulator, thereby moving another component from the motherboard onto the CPU.
  • 37, 47, 57W TDP mobile processors.
  • 35, 45, 55, 65, 77 and ~100W+ (high-end) TDP desktop processors.
  • 15W TDP processors for the Ultrabook platform (multi-chip package like Westmere).
And here are a couple slides that Intel has shown regarding Haswell:

Image

Image

What are you all expecting from Haswell? Do you think we'll see significant IPC gains or will it be more towards IGP improvement than anything else?
I'm not sure exactly what AVX2/TSX will bring. What I'm hoping for the most is a better thermal interface than Ivy Bridge to help with overclocking.

Unfortunately for AMD, I fear this could be the final nail in the coffin, which would be unfortunate for all of us past 2013. They are so far behind...
Last edited by DeadOfKnight on Sun Jul 29, 2012 11:22 am, edited 1 time in total.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 4:32 am

New socket?
DDR4 for the enterprise / server but not desktop?
Definitely sounds like something to wait for a 6 - 12 month refresh on if you ask me. H
AbRASiON
Gerbil Team Leader
 
Posts: 274
Joined: Tue Mar 18, 2003 8:06 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 4:37 am

It looks like more evolutionary tweaks.

IHMO, there's little reason to wait for it unless you want to have built-in Thunderbird support and/or want a beefer integrated GPU.
Ivy Bridge i5-3570K@4.0Ghz, Gigabyte Z77X-UD3H, 2x4GiB of PC-12800, EVGA 660Ti, Corsair CX-600 and Fractal Refined R4 (W). Kentsfield Q6600@3Ghz, HD 4850 2x2GiB PC2-6400, Gigabyte EP45-DS4P, OCZ Modstream 700W, and PC-7B.
Krogoth
Maximum Gerbil
Silver subscriber
 
 
Posts: 4402
Joined: Tue Apr 15, 2003 3:20 pm
Location: somewhere on Core Prime

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 5:07 am

AbRASiON wrote:New socket?
DDR4 for the enterprise / server but not desktop?
Definitely sounds like something to wait for a 6 - 12 month refresh on if you ask me. H

New LGA1150 socket for Haswell and then a new socket for Broadwell again, supposedly.

This is old news with Intel; we're lucky SB/IB shared the same socket. LGA1156 didn't last as long.

DDR4 might be expensive for early adopters and I doubt it would matter much for desktop users.

Krogoth wrote:It looks like more evolutionary tweaks.

IHMO, there's little reason to wait for it unless you want to have built-in Thunderbird support and/or want a beefer integrated GPU.


There's little reason to wait ever, if you need the upgrade now.

This has always been true, but I think if you did wait then your chip might last you more than an extra year with these new technologies built in.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 5:25 am

Yes my Q9450 will finally retire when Haswell releases the desktop variant of their quad core 3770k variant in Haswell terms.

I am very excited.
Core i7 4770K | eVGA GTX770 SC ACX | 16GB DDR3 2133mhz | Asus Z87-PLUS | Corsair HX650 | Fractal Define R4 | Samsung 840 Pro 256GB | Windows 8 x64
yogibbear
Gerbil Elite
Gold subscriber
 
 
Posts: 668
Joined: Fri Feb 08, 2008 11:30 am

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:10 am

Unless it nukes all the Intel atom CPUs then my wife won't realize we need a new laptop...
Star Brood
Gerbil First Class
Silver subscriber
 
 
Posts: 181
Joined: Wed Mar 21, 2012 4:57 am

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:14 am

Star Brood wrote:Unless it nukes all the Intel atom CPUs then my wife won't realize we need a new laptop...

Isn't there an Intel 22nm "ValleyView" SoC coming next year? That should be cheap enough and make enough of a difference.

In fact, for those who are interested, that will probably be more of an exciting update than Haswell.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:48 am

I find it pretty hard to get excited about what are essentially minor tweaks. IMO the last real revolutions in x86 were the introduction of 64-bit instructions and integrated DRAM controllers. (OK, hardware virtualization support was pretty huge too, but that's not something everyone uses.)
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37632
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:52 am

I'd like to see Intel release variants with the better graphics across a broader range of desktop CPUs. It bothers me that the best integrated GPUs are available on the 'enthusiast' CPUs which are least likely to use them. A 2c/4t CPU with the best graphics would make a great compact system.
MadManOriginal
Graphmaster Gerbil
 
Posts: 1411
Joined: Wed Jan 30, 2002 7:00 pm
Location: In my head...

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:53 am

just brew it! wrote:I find it pretty hard to get excited about what are essentially minor tweaks. IMO the last real revolutions in x86 were the introduction of 64-bit instructions and integrated DRAM controllers. (OK, hardware virtualization support was pretty huge too, but that's not something everyone uses.)


I tend to agree on the CPU side, although AVX2. FMA3, and the transactional memory support in Haswell will be nice (not revolutionary) features when they get software support. Where Haswell will be great (if Intel lives up to its hype) is in the GPU + mobile power envelope. If everything works out according to the rumors, an Ultrabook with Haswell at 15 watts will probably have a GPU that's competitive with the highest-range Trinity notebooks at 35 watts (while easily beating Trinity at CPU). That's a pretty nice hunk of GPU for a small power envelope that will make Ultrabooks more interesting. Once again, if all goes well, it should mean Ultrabooks get a bump in battery life too. Larger notebooks should get a nice performance boost and bump in battery life due to Haswell's improved sleep states and power management even though the peak power consumption will be about the same as today.
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 560
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 6:58 am

just brew it! wrote:I find it pretty hard to get excited about what are essentially minor tweaks. IMO the last real revolutions in x86 were the introduction of 64-bit instructions and integrated DRAM controllers. (OK, hardware virtualization support was pretty huge too, but that's not something everyone uses.)

I don't consider it a minor tweak that, if the slides are true, they'll be bumping the high end LGA1150 part's TDP back up to 95W...that could mean more performance in a meaningful way.

I also don't consider beefed-up integrated graphics a minor tweak, although admittedly that feature isn't very exciting for those of us who require discrete graphics (most members of this forum).

Again, I don't really know what the implications of TSX and AVX2 are, but I assume that it could be exciting later on down the road when developers begin to utilize the new instructions.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 7:07 am

chuckula wrote:I tend to agree on the CPU side, although AVX2. FMA3, and the transactional memory support in Haswell will be nice (not revolutionary) features when they get software support. Where Haswell will be great (if Intel lives up to its hype) is in the GPU + mobile power envelope. If everything works out according to the rumors, an Ultrabook with Haswell at 15 watts will probably have a GPU that's competitive with the highest-range Trinity notebooks at 35 watts (while easily beating Trinity at CPU). That's a pretty nice hunk of GPU for a small power envelope that will make Ultrabooks more interesting. Once again, if all goes well, it should mean Ultrabooks get a bump in battery life too. Larger notebooks should get a nice performance boost and bump in battery life due to Haswell's improved sleep states and power management even though the peak power consumption will be about the same as today.

I agree, where Haswell will really shine is in low power computing. We'll actually get to see what their hyped-up 22nm process is really capable of.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 7:14 am

MadManOriginal wrote:I'd like to see Intel release variants with the better graphics across a broader range of desktop CPUs. It bothers me that the best integrated GPUs are available on the 'enthusiast' CPUs which are least likely to use them. A 2c/4t CPU with the best graphics would make a great compact system.

Yeah, that makes a lot of sense. I know I won't be taking full advantage of the integrated graphics.

What baffles me about this is that Intel obviously understands this if they're pairing better graphics with mobile chips.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 7:19 am

I'm just curious if Intel will use this generation to bring true quad-core CPUs into the budget/Core i3 price range. So far only AMD has done it, while Intel has kept with dual-core + HT. I have a Core i3-2120 right now, which is damn fast no doubt, but it still loses in (some) multi-threading (not by much, admittedly) to my Q9550 that I was forced to abandon.
Under Construction Forever~~~
Kurotetsu
Gerbil Elite
 
Posts: 525
Joined: Sun Dec 09, 2007 12:13 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 7:23 am

Kurotetsu wrote:I'm just curious if Intel will use this generation to bring true quad-core CPUs into the budget/Core i3 price range. So far only AMD has done it, while Intel has kept with dual-core + HT. I have a Core i3-2120 right now, which is damn fast no doubt, but it still loses in (some) multi-threading (not by much, admittedly) to my Q9550 that I was forced to abandon.

Well it's confirmed that they will be releasing dual-core variants of Haswell, no word on pricing though.

Even so, they should be getting a boost in IPC and even clock speed on their matured 22nm process.

This is another reason I choose to upgrade on a "Tock" as you tend to see bigger performance gains come with a new architecture on a matured process.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 8:18 am

I'm really looking forward to seeing the performance of the bottom of the range Haswell quads. If the numbers are right I reckon I'll retire my ageing desktop Core 2 Duo with some of the Haswell hotness.

I suspect I'll wait until Broadwell to replace the Core 2 Duo notebook I have.

Having said that, both systems are massively overpowered for he type of stuff I'm doing these days, making justification for the upgrades somewhat more difficult.
Ubuntu 12.04 AMD64: E8200 // P35 // HD 4850 // 4GB
OS X 10.8.x: iMac12,2, MacBook 5,2
pedro
Gerbil First Class
 
Posts: 176
Joined: Fri May 11, 2007 6:13 am

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 8:25 am

pedro wrote:Having said that, both systems are massively overpowered for he type of stuff I'm doing these days, making justification for the upgrades somewhat more difficult.

Well then I expect they'll be even more massively overpowered and you won't have to justify it again until they die on you.

I think the I/O, IGP, and efficiency are probably enough to justify it this time around though. From Core 2, that is.

Anything older and I'd probably have upgraded with Sandy, anything newer and I can see users wanting to hold off.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 10:51 am

Does anyone have any technical insight on what Haswell new instructions TSX, AVX2, and FMA3 mean and how they will work?
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 10:57 am

There was quite a bit of good info regarding transactional memory over at Real World Tech recently.
Ubuntu 12.04 AMD64: E8200 // P35 // HD 4850 // 4GB
OS X 10.8.x: iMac12,2, MacBook 5,2
pedro
Gerbil First Class
 
Posts: 176
Joined: Fri May 11, 2007 6:13 am

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 11:15 am

pedro wrote:There was quite a bit of good info regarding transactional memory over at Real World Tech recently.

Yeah, I do remember skimming over this bit awhile ago.

http://www.realworldtech.com/haswell-tm/

Some of that stuff is hard to follow, not sure what it all means for the end user.
I was reading somewhere that AVX2 would be pretty significant but it was equally hard to follow.
Again, I have no idea what these will mean for consumers. I'm sure it means a lot to the devs.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 7:26 pm

DeadOfKnight wrote:Does anyone have any technical insight on what Haswell new instructions TSX, AVX2, and FMA3 mean and how they will work?
These terms encompass sets of instructions intended to address particular weaknesses/points of emphasis in x86 or to open new areas of development:

    TSX -- Transactional Memory support. The previously-linked RWT article is probably the best introduction to the details of Intel's implementation, but you may need to do some background reading on the general concept to make much sense of it. At the highest level, this offers a potential mechanism to get more real throughput out of multi-threaded code by eliminating unnecessary serialization (where the code running in one thread thinks it needs to wait on changes being made another thread, when in a particular case it actually doesn't). There are actually two separate mechanisms defined by Intel: one which has limited potential but can be implemented without major changes to existing code, and another which is more exciting but will require significant re-architecting of most codebases.

    AVX2 -- further extensions to the AVX set of instructions introduced with Sandy Bridge. This is the latest generation of SIMD instructions (in the lineage that stretches back through SSE to MMX), further extending the 256bit operations and among other things adding a GATHER instruction (to collect operands from a series of non-contiguous memory locations, which potentially eliminates a lot of gyrations involving multiple loads and moves when working with common data structures).

    FMA -- Fused Multiply-Add. This is a particular operation that is very common in floating point code (including physics, 3D, and image/video/audio processing); the 3 in FMA3 refers to the 3-operand version (ie two numbers multiplied together and added to a third) without losing any precision in the intermediate steps. CPUs without an FMA instruction have to perform several other instructions to get the same result, which obviously takes more cycles/time; this has been one long-standing criticism of x86's SIMD implementation versus competitors such as Power's Altivec. Of some concern is the divergence of AMD and Intel with respect to FMA implementations; without going into the history and details of their various FMA3 and FMA4 specs, we should just note that a potential incompatibility exists which might (further) limit the adoption of the instruction. However, Intel clearly has the momentum wrt future marketshare and (especially) developer-relation resources, while AMD has shown a history of conforming to Intel implementations (eventually) when necessary, so FMA3 at least should see adoption over time.
AVX2 and FMA (which really should be considered just part of AVX, if not for the circumstances of its birth) are clear enhancements for floating point code, and of course that makes them potentially interesting for immediate real-world applications like gaming. TSX is significantly more speculative and long-term (much of the work to date with transactional memory has remained in academia, because the software-only implementations were too slow to be of practical utility), and its most likely first applications will be in specialized contexts like HPC where the hardware is known and the entire codebase belongs to one set of developers. Thereafter it will show up in server applications like MySQL that employ many interdependent threads. Its migration to the actual OS or consumer-level applications is probably quite distant, and will await lessons learned by both Intel and software architects experimenting with this first iteration (there are some obvious limitations in this initial implementation, and I would expect many of those to be mitigated before it sees employment in, for example, the Windows kernel). Of course the nature of open source, and the level of academic interest in the subject, means that there will be plenty of experimental forks of existing codebases to take advantage of Intel's work -- which is kind of exciting, in that we don't really know how those will turn out. The total gains generally will be modest -- especially with the more limited "easy" version Intel offers -- but since this will be the first widespread deployment of a heretofore experimental concept, interesting results might turn up.

However, it's worth emphasizing that -- like any new instructions -- all of these require changes in existing code; there's no "free" upgrade here like we often see with other CPU features. And we haven't even seen much AVX code yet (AFAICT) in non-specialized contexts, despite all the shipments of AVX-enabled SB/IB, so patience is required. The vast majority of developers don't work at the level of assembly language anyway, so at the very least they wait on compiler support; but in most cases the dependency is on underlying code they don't own -- libraries, game engines, etc -- which have to be modified to varying degrees (FMA a relatively easy in-place replacement, TSX a potential throw-it-out-and-start-over). Meanwhile these ISA extensions only exist in new processors, which will constitute a tiny (though ever-growing) fraction of the market, so the return on investment for developers doing those changes is initially very small. The new code has to be put behind switches so that it falls back to the boring instructions we're already using when it is run on pre-Haswell CPUs (one exception to this is Intel's "easy" version of TSX, which is cleverly designed to work correctly even on older processors, but as I said the potential gains from that version won't be huge.)

So while these are all arguably exciting features, they're not going to do anything for you the day Haswell arrives... and in fact there's pretty much an inverse relationship between how exciting the feature is and how immediately we'll see its benefits. FMA and "easy" TSX are relatively cheap but not super exciting, AVX2 is more exciting but will take longer, and "real" TSX could be pseudo-revolutionary but probably will only be of any real note sometime after Broadwell ships.
UberGerbil
Gerbil Khan
 
Posts: 9974
Joined: Thu Jun 19, 2003 3:11 pm

Re: Intel "Haswell" Speculation thread

Postposted on Tue Jul 24, 2012 8:59 pm

Ubergerbil: Excellent rundown of the features. One thing that I would add is that AVX2 adds a whole bunch of integer 256-bit instructions in addition to the floating point instructions from the original AVX. This will make AVX very useful for a wide range of integer number crunching applications including lots of cryptographic functions. Some video compression systems like x264 also heavily use integer code and could get a good bump from AVX2 as well.
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 560
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: Intel "Haswell" Speculation thread

Postposted on Wed Jul 25, 2012 1:20 am

Yes, thank you for that, it was very insightful. So basically what I gather is that Haswell is a very future-looking architecture, but it will take several years for the software to catch up. When it does, this chip will grow with you.

That's definitely saying something. It may not seem worth the upgrade right now just for that as future chips will support it as well, but this chip should last you a good long time before it is rendered obsolete. AVX2 is an exciting feature that is almost "too new", but it has to start somewhere.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Wed Jul 25, 2012 2:32 am

HPC and parallel computing researcher here. UberGerbil already described AVX2 and FMA and those are the things I am most looking forward to. Not sure how Intel has implemented it, but FMA3 potentially can double the FP throughput over Sandy Bridge/Ivy Bridge. Now SB/IB core had dual 256-bit units so I am assuming each of them was widened to handle FMA3 to give twice the throughput, but the arrangement might very well be different. We won't know until IDF or later i suppose.

Gather support is also very nice to have. Makes the SIMD units more like a GPU SM, except that AVX2 vector lanes aren't predicated whereas GPU vector lanes are. In that regard, the vector instruction set of Many Integrated Core is better IMO (wider but more flexible vector units) but AVX2 is still a nice improvement over AVX.

About the GPU improvements, I am mostly interested to see GPU compute performance enhancements if any. I really have no idea about what will be in GT3. The linux kernel commits seemed to indicate the EU ISA is largely unchanged.

I have a whole bunch of OpenCL code. HD 4000 OpenCL performance has been very disappointing compared to Llano on all my OpenCL codes, with Llano often being 2x faster. Might be different for other people, I dunnow. My own codes do fairly well on both AMD and Nvidia GPUs but poorly on Intel despite lots of tuning effort. Not sure if its due to driver issues or hardware as Intel's OpenCL documentation and tools are not very mature.
Anyway, would be nice if Haswell GPU did better on OpenCL.
Last edited by codedivine on Sat Aug 04, 2012 8:43 pm, edited 1 time in total.
codedivine
Gerbil Elite
Silver subscriber
 
 
Posts: 702
Joined: Sat Jan 24, 2009 8:13 am

Re: Intel "Haswell" Speculation thread

Postposted on Wed Jul 25, 2012 3:12 am

chuckula wrote:Ubergerbil: Excellent rundown of the features. One thing that I would add is that AVX2 adds a whole bunch of integer 256-bit instructions in addition to the floating point instructions from the original AVX. This will make AVX very useful for a wide range of integer number crunching applications including lots of cryptographic functions. Some video compression systems like x264 also heavily use integer code and could get a good bump from AVX2 as well.
Good catch. I thought I'd mentioned Integer but I guess that got lost while I was editing the Gather bit. And you're right: integer (and bitwise) operations have wide applicability in many areas, from AI to crypto to database indices. It's also kind of funny in that it brings things full circle: Intel's very first SIMD implementation (MMX) was entirely integer.
DeadOfKnight wrote:So basically what I gather is that Haswell is a very future-looking architecture, but it will take several years for the software to catch up. When it does, this chip will grow with you.
Well, yes, but all ISA extensions are inherently future-looking. The transactional memory feature makes this potentially more so than most, but then again that could turn out to be something of a dead end, too. And it's quite possible that none of this will be widely supported in software until long after Haswell is old news and TR is full of people looking forward to replacing their Broadwell systems.
codedivine wrote:Gather support is also very nice to have. Makes the SIMD units more like a GPU SM, except that AVX2 vector lanes aren't predicated whereas GPU vector lanes are. In that regard, the vector instruction set of Many Integrated Core is better IMO (wider but more flexible vector units) but AVX2 is still a nice improvement over AVX.
Well, the Knights* products look a lot like a test run of features that will eventually work their way into some future mainline x86 products if they prove useful. It doesn't make a lot of sense to have two divergent (or even parallel) x86 ISAs over the long term. But for now they can widen the functional units internally to support things like FMA without committing to wider registers and other software-visible functionality. (That still leaves the issue of throughput: in a tight loop of FMA code, can they keep everything fed?)

Your observations on Intel's GPU compute are interesting, as that's an area where I have no direct knowledge. One wonders if Intel's heart is really in it, as their corporate DNA is all about pushing code onto the CPU, not offloading it to coprocessors, and the spur to implement OpenCL at all may be coming largely from one large malic customer. Of course they also have a long history of less-than-adequate graphics drivers (for whatever reason) so it's always possible to ascribe it to Hanlon's Razor.
UberGerbil
Gerbil Khan
 
Posts: 9974
Joined: Thu Jun 19, 2003 3:11 pm

Re: Intel "Haswell" Speculation thread

Postposted on Wed Jul 25, 2012 3:28 am

Better drivers are definitely one thing to hope for. If Intel is really so proud of the GPU coming with Haswell and confident that it can compete directly with AMD's integrated graphics, let's see if they will love and care for it as much as AMD does.

No matter how much the graphics improves, until they do this AMD will still be the better option for casual gamers utilizing integrated graphics. Whether it be a small 13" notebook or an HTPC, people will still want game optimized driver support.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Wed Jul 25, 2012 7:29 pm

HD3000 has pretty damn good gaming support, and I have an HD4000 laptop that I can try to test out. Drivers for gaming really isn't Intel's biggest issue, but OpenCL support is something that everyone can benefit from in the long run. They might as well catch up with AMD and Nvidia on this point, and they might as well start working more towards a Fusion-like setup with shared resources. It makes the most sense in the mobile computing space, but it can be useful everywhere.
Canon 6D||[24-105/4L IS USM|100/2.8L Macro IS USM|70-300/4-5.6 IS USM|40/2.8 STM|50/1.4 USM|85/1.8 USM|Samyang/Bower 14/2.8 Full-Manual Rectilinear Wide-angle|
Canon EOS-M|11-22/4-5.6 IS STM|22/2 STM|EF-M 18-55/3.5-5.6 IS STM|
For sale!|24/2.8 IS USM
|
Airmantharp
Maximum Gerbil
 
Posts: 4986
Joined: Fri Oct 15, 2004 10:41 pm

Re: Intel "Haswell" Speculation thread

Postposted on Thu Jul 26, 2012 12:42 am

Airmantharp wrote:HD3000 has pretty damn good gaming support.

Yes it does, but from what I've read the support is coming mainly from the video game developers, not Intel.

It would be better if we saw that kind of support coming from both sides like we do with AMD and NVIDIA.

We all know Intel can afford it.
Intel Core i7-875K, Asus P7P55D-E Pro, Win 7 Home Premium
MSI GTX 560 Ti OC, Mushkin 2x2GB DDR3-1333, Corsair TX650
Cooler Master Hyper 212+, Logitech Z-2300, ASUS Xonar DX
Samsung Spinpoint F3 1TB, Dell Ultrasharp U2410, Antec P183
DeadOfKnight
Gerbil Elite
Gold subscriber
 
 
Posts: 635
Joined: Tue May 18, 2010 1:20 pm

Re: Intel "Haswell" Speculation thread

Postposted on Thu Jul 26, 2012 3:40 pm

codedivine wrote: I have a whole bunch of OpenCL code. HD 4000 OpenCL performance has been very disappointing compared to Llano on all my OpenCL codes, with Llano often being 2x faster. Might be different for other people, I dunnow. My own codes do fairly well on both AMD and Nvidia GPUs but poorly on Intel despite lots of tuning effort. Not sure if its due to driver issues or hardware as Intel's OpenCL documentation and tools are not very mature.
Anyway, would be nice if Haswell GPU did better on OpenCL.



This is a little freaky. I just got an email from an Intel employee asking for my OpenCL code samples citing this forum post :o
I am replying back to them. So own up, who amongst you works for Intel and how did you get my email address :D ?
Last edited by codedivine on Sat Aug 04, 2012 8:43 pm, edited 1 time in total.
codedivine
Gerbil Elite
Silver subscriber
 
 
Posts: 702
Joined: Sat Jan 24, 2009 8:13 am

Re: Intel "Haswell" Speculation thread

Postposted on Thu Jul 26, 2012 8:11 pm

codedivine wrote:This is a little freaky. I just got an email from an Intel employee asking for my OpenCL code samples citing this forum post :o
I am replying back to them. So own up, who amongst you works for Intel and how did you get my email address :D ?


Oh Really? I'm going to have to have a talk with the drones developers about that. Obviously our surveillance equipment needs an upgrade if they feel the need to actually *ask* you for the code that isn't working! (just kidding obviously, but it is interesting that at least somebody at Intel cares about making OpenCL actually work).
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 560
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Next

Return to Processors

Who is online

Users browsing this forum: No registered users and 0 guests