Intel Core i9 ‘Gulftown’ Previewed

Zero82z said:
The problem is that the software itself isn't taking advantage of the resources, not that the resources aren't helping the software. And multiple cores, HT, and all that other stuff is still incredibly useful for people who do a lot of multitasking. Modern OSes are quite good at balancing threads across different cores.
Click to expand...

I do beleive paralellism is the future. However, it is not easy to implement. I think we -currently- are in a transition period where more and more programmer learn ot use multithreading more efficiently and where the tools we use are starting to give us easy ways of using paralele architectures.

If you just look at Microsoft's .NET, the 4.0 version of the framework is in the making and they are adding alot of stuff like "Parallel.for" and "Parallel.foreach" which basically threads every iteration of the for-loop and for-each-loop. Thus making it easy for .net programmers to quickly throw in some parallelism here and there.

My guess is, within 5 to 10 years, all software will be made with paralellism in mind.
 
nlz242 said:
I think its more a matter of what is multithreadable.
Im currently reading a book on threaded programming (i am a programmer) and some tasks arent easy to multithread. Games for example, are pretty hard to multithread and get decent gains... I have heard that Valve, with their latest engine (used in L4D) somewhat succeed at getting a nice implementation going... havent really tested it myself tho, so i cant say.

Have a look at : http://www.bit-tech.net/gaming/pc/2006/11/02/Multi_core_in_the_Source_Engin/1
Click to expand...

That sums it up yea. The only apps that ever bother to use multithread are easy implementations with predictable operations like compression, encoding, and encryption. Multithread could be used in games more, surely not everything it could be used for in games is that hard to implement. Specifically I don't know why developers still haven't used it much when it comes to destructable environments. I want to see *everything* in games be 100% destructable. I'm sick of shooting rockets or nades at walls and only leaving a stain on the wall lol The ghostbusters CPU based stuff I have seen here shows only a small amount of what could be done and that was chained to console limitations!

Think about future games where you can't simply bast away like a maniac because you will bring down the building if you do. Or say you blow up something in the dirt, well shouldn't it leave a huge hole in the ground? Crash a car into the side of a building to get in? There is so much you could do with this stuff. Is it really so hard to start up a new thread on another core when the event happens? It doesn't seem like it would even take that much processing and it would only persist for a short time while the event is taking place.

I simply doubt that programmers will ever bother to use multithread because it is too complex.
 
The fact that these are dual-socket enabled chips with 12MB of L3 suggests that these are Xeons. Parallelism is more important in the server market than at home.
 
niconx said:
Can't wait to see another review on hyperthreading. We know it didnt do anything for p4 I wonder if it makes a different when you have 12 instances of it lol, Lets see 0*12=0. :rolleyes: Still maybe things have changed.
Click to expand...

um. hyperthreading makes a difference, lets say you are rendering and remove one virtual core from the rendering programs affinity you can do something else and its not bogged down, yet that core can still be doing some of the work.

even on the p4 you could do that.
 
niconx said:
I simply doubt that programmers will ever bother to use multithread because it is too complex.
Click to expand...
Well, you're definitely wrong there. The trend with current CPUs is going towards more and more cores and not as much toward higher frequencies and better single-threaded performance. Programmers will have to start multithreading their programs in order to take advantage of multi-thread CPUs, or else their software simply won't perform very well. The fact that it's difficult means that it'll take a while, but it certainly will happen.
 
niconx said:
I simply doubt that programmers will ever bother to use multithread because it is too complex.
Click to expand...

Well, i think they will HAVE to. Processor aren't evolving as fast as they used to do.
I remember when i changed my Pentium 133 for a AMD K6-2 350, it was a huge step.
My Athlon 1GHz to a Athlon XP 2500+, huge step.

Changing my Core 2 E6600 to a i7 920? no so much of a step.
And i anticipate it'll only get worse. So, i think, programmers won't have a choice to multithread, because at some point it is all they will get (new cores).
 
nlz242 said:
Well, i think they will HAVE to. Processor aren't evolving as fast as they used to do.
I remember when i changed my Pentium 133 for a AMD K6-2 350, it was a huge step.
My Athlon 1GHz to a Athlon XP 2500+, huge step.

Changing my Core 2 E6600 to a i7 920? no so much of a step.
And i anticipate it'll only get worse. So, i think, programmers won't have a choice to multithread, because at some point it is all they will get (new cores).
Click to expand...

Very right that processors aren't making the IPC advancements they used to. Widspread usage of multithreading in complex tasks is something I have have to see to believe. Just because programmers have no option does not mean it is a feasible option. I am sure multithreading complex tasks introduces certain stability and security problems. If every programmer needs to have a PhD level of education and experience in order to write this software then there isn't going to be much of this software.
 
Zero82z said:
The problem is that the software itself isn't taking advantage of the resources, not that the resources aren't helping the software. And multiple cores, HT, and all that other stuff is still incredibly useful for people who do a lot of multitasking. Modern OSes are quite good at balancing threads across different cores.
Click to expand...

The core of what runs the software is the OS. Windows (until 7) did not use the cores or hyper threading that well. Now with 7, since it is written for multi-core, multi-threaded apps to maximize the cycle counts. Mac seems to use them as efficiently as Windows 7 and I'm haven't read anything about the Linux environment (even though I am starting to use it more) to see why anything beyond the i7 brand is a performance boost in raw performance numbers.

jynxycat said:
The additional cores themselves may or may not help in the end, but it's safe to say that the new line of Intel chips will have quite a large performance boost.
Click to expand...
Yes but raw numbers and benchmarks aside, they're going to be minimal compared to the C2D/C2Q to i7 chipset.

You can paint your arguments on my viewpoint either as someone trying to justify their i7 purchase but I can say with all the photo editing I do, video and gaming, there is no way I'm hitting the ceiling on this chipset yet especially with it running @ 3.8Ghz.
 
SymptomsReagent said:
The core of what runs the software is the OS. Windows (until 7) did not use the cores or hyper threading that well. Now with 7, since it is written for multi-core, multi-threaded apps to maximize the cycle counts.
Click to expand...
You're totally wrong. First of all, Windows has had proper HT support since XP, and even Vista is very good at balancing threads across different cores. 7 is a little better, but most improvements in that area are fairly minor.

Also, there's a big difference between programs that take advantage of multiple threads, and operating systems that handle single-threaded applications on a multi-thread CPU. A program's ability to multithread is independent of the OS, so even if Windows did have shitty thread balancing, a properly-multithreaded application will still be able to take full advantage of a multi-core CPU.
 
its not only the os that has to be compatible with the extra cores. the software has to be also capable.
 
would love to try Supreme commander on one of these. only game I know of that supports up to 8 cores.
 
Soggy said:
would love to try Supreme commander on one of these. only game I know of that supports up to 8 cores.
Click to expand...

it seems somewhat hard to make that game run fast without slowing down when playing large games. it doesnt even use 100% cpu.. the game itself just slows dooooowwwwnnnnn.... poor really... i love the game
 
SonicTron said:
one of these machines has the power to operate an entire high school's computing needs all at once :eek:
Click to expand...

Considering the quality of education in public high schools. I am fairly certain that an 8086 would satisfy the level of computation needed for all the math classes to be given on a given day. :D
 
SymptomsReagent said:
Diminishing returns. We've hit the point where cores, hyper-threading and god knows what else isn't helping the software.

Until Mac, Windows and Linux use all of them efficiently, waste of money for home users.
Click to expand...

Obviously you do not use any sort of CG type software. Companies like Pixar will capitalize on cpu's like this one. Hell, at my work we could use something like that. I have a Quad Core at work and with something like this, my 45 minute renders could be (potentially) 6x faster
 
Of course, that makes you have to work that much harder. No more "I'm rendering!" excuse while you're at the water cooler ;)

Unfortunately we're still relying on the same programming/cpu principles developed decades ago, and as such it will take us some time to truly realize multi-core cpu potential.
 
MisterSparkle said:
Of course, that makes you have to work that much harder. No more "I'm rendering!" excuse while you're at the water cooler ;)
Click to expand...
That's when you step up the size of the textures, AA, filters, and rendering passes. ;)Winning back the watercooler breaks will be the excuse to make Ray Tracing popular.
 
Maybe the i7 price will finally go down when these i9 are out in the market
 
niconx said:
That sums it up yea. The only apps that ever bother to use multithread are easy implementations with predictable operations like compression, encoding, and encryption. Multithread could be used in games more, surely not everything it could be used for in games is that hard to implement. Specifically I don't know why developers still haven't used it much when it comes to destructable environments. I want to see *everything* in games be 100% destructable. I'm sick of shooting rockets or nades at walls and only leaving a stain on the wall lol The ghostbusters CPU based stuff I have seen here shows only a small amount of what could be done and that was chained to console limitations!

Think about future games where you can't simply bast away like a maniac because you will bring down the building if you do. Or say you blow up something in the dirt, well shouldn't it leave a huge hole in the ground? Crash a car into the side of a building to get in? There is so much you could do with this stuff. Is it really so hard to start up a new thread on another core when the event happens? It doesn't seem like it would even take that much processing and it would only persist for a short time while the event is taking place.

I simply doubt that programmers will ever bother to use multithread because it is too complex.
Click to expand...

Bollocks, do read up:
http://www.pcgameshardware.com/aid,687620/ArmA-2-tested-Benchmarks-with-18-CPUs/Practice/

And stop calling this "i9", this is the "tick" of i7:
http://www.overclock3d.net/news.php?/cpu_mainboard/intel_release_32nm_westmere_information/1
 
Speaking as a programmer who is already implementing multi-process stuff, there's nothing too horrible stopping more applications making use of it. It's really pretty similar to putting 2 or more people to work at the same task. There's a whole load of extra work sitting down at the start and deciding who will do which part, and then during the task there is extra work communicating where everything is up to.

I have often thought games should be a fairly ideal candidate - possibly running opponent AIs in separate processes, and physics calculations in another process and so on. I'd have thought a lot of this could even just work one frame behind the main rendering which potentially would remove a lot of the concerns about locking between the processes (and automatically give a realistic 'reaction time' if applied to opponent AIs). But I'm not a games programmer and problems tend to look small when viewed from a distance.

There's a lot of concurrency issues leading to requiring locking, and therefore the possibility of deadlocks and livelocks etc. My position has always been that locking is too hard. I can't guarantee I'll have a team of people that can handle it well. So a major part of the design is often in assigning states to system elements, and assigning ownership of various states to different processes. Then a given process can just look around for elements in states that belong to it, without worrying about locking for most tasks.
 
Video card drivers are already multi core aware, with more cores they can further reduce latency getting instructions to the GPU.
With concurrent GPUs, the PCI-E bus may become a limiting factor on what performance increases more CPU cores can bring, hopefully that will see some improvements too.
Memory bandwidth will need to increase and/or use more channels to get the best out of them.
Maybe this is why i7 boards are triple channel as a prelude for what is to come.
 
Zero82z said:
The current LGA1366 lineup will probably be phased out once the 6-core CPUs are released.
Click to expand...

Meaning mainstream segment will be left with the i5 CPU without those triple channel memory thing? That sucks:(
 
i dont know about half of you guys that already own an i7 rig, but im not thinking too much about the new stuff.. what i got works great so....
 
Zero82z said:
You're totally wrong. First of all, Windows has had proper HT support since XP, and even Vista is very good at balancing threads across different cores. 7 is a little better, but most improvements in that area are fairly minor.

Also, there's a big difference between programs that take advantage of multiple threads, and operating systems that handle single-threaded applications on a multi-thread CPU. A program's ability to multithread is independent of the OS, so even if Windows did have shitty thread balancing, a properly-multithreaded application will still be able to take full advantage of a multi-core CPU.
Click to expand...

I never said that XP nor Vista didn't have multi-core or HT support, I said that it was not nearly as efficient as it is in 7. Big difference there. The way 7 handles the cycles is much, much better than how Vista and XP handled Multi-core and HT. It's just a proven fact given the recent RC1 download and I'm sure will again be proven with the OE release this October.

Again--- Not discounting the software that runs independent of Windows/Mac or Linux but my point rather is if you're not having a combination of the two using the available resources, these processors that will post speed improvements don't get utilized to their maximum potential unless the software is capable of using it. Thus again my argument of "Why past i7 for the consumer?" if it's not going to see any real world benefit?

This goes in grain with the whole argument of "Well apparently you don't like hardware improvements..." etc etc, and I am all for making things smaller, faster and more powerful, but in the realm of benefiting someone or something.
 
I never said that XP nor Vista didn't have multi-core or HT support, I said that it was not nearly as efficient as it is in 7. Big difference there. The way 7 handles the cycles is much, much better than how Vista and XP handled Multi-core and HT.
Click to expand...
Still incorrect. It is not much, much better. Only slightly, at best.
It's just a proven fact given the recent RC1 download and I'm sure will again be proven with the OE release this October.
Click to expand...
Proven? Based on what? The only thing that was announced by Microsoft in regards to SMP in Windows 7 was that HT support is a little bit improved.
 
rflcptr said:
Still incorrect. It is not much, much better. Only slightly, at best.

Proven? Based on what? The only thing that was announced by Microsoft in regards to SMP in Windows 7 was that HT support is a little bit improved.
Click to expand...

Incorrect based on what... your opinion? I can't do the reading for you but go out and Bing or Google Multi-core/HT Win7 vs. xp and see the stories posted on them. Specially looking at the optimizations for Windows 7 given the new processors Intel has put out. My view point of the matter is software created in 2009 is probably a bit better than software released back in 2001.

However, without running off a rabbit trail on this, PM me rflcptr if you want to carry on this conversation.
 
SymptomsReagent said:
Incorrect based on what... your opinion? I can't do the reading for you but go out and Bing or Google Multi-core/HT Win7 vs. xp and see the stories posted on them.

However, without running off a rabbit trail on this, PM me rflcptr if you want to carry on this conversation.
Click to expand...
No. Show us the official evidence of the "much, much better" implementation in Windows 7 versus the rest.

I'm not going to Bing or Google half-assed forum posts like yours.
 
Incorrect based on what... your opinion?
Click to expand...
Vista already was aware of and worked accordingly with Hyperthreaded machines. For example, it will schedule threads on an idle physical processor rather than choosing an idle logical processor on a physical processor that already contains a busy logical.
 
SymptomsReagent said:
I never said that XP nor Vista didn't have multi-core or HT support, I said that it was not nearly as efficient as it is in 7. Big difference there. The way 7 handles the cycles is much, much better than how Vista and XP handled Multi-core and HT. It's just a proven fact given the recent RC1 download and I'm sure will again be proven with the OE release this October.
Click to expand...
And you're still wrong, because I've been using Vista for about two years now and it has done a great job of balancing threads across my multi-core systems (especially my main rig which is constantly under heavy load and running many different applications). I have used quite a few different Windows 7 builds as well, including both the public beta and RC, and its handling of multi-core CPUs is not much different from Vista.

For someone who doesn't actually require a lot of power, faster CPUs like i7 and beyond obviously won't make a difference. But for people who do perform a lot of tasks that eat up CPU time, these new chips will be very beneficial.
 
Unless this is a huge change for Intel, it seems more likely that these are new Xeon 5XXX Series procs than Core i9s, by virtue of it being a dual socket system. Unless they're bringing back the Skulltrail-type setup.
 
Fantastic, right when I'm about to build an i7 rig. :rolleyes:

Oh well, what would be the point of hardware if I didn't have one more thing to lust after. :)
 
You must log in or register to reply here.
Back
Top