Next Generation Quad-Core Yorkfield is Now on Track

[malware]

INTEL DEVELOPER FORUM, San Francisco, Calif., Sept. 26, 2006 - Intel President and CEO Paul Otellini outlined the company's plan to accelerate its technology leadership. There is a upcoming Quad-Core Intel Core 2 Extreme QX6700 (2.66GHz/4MB x 2 L2/1066MHz FSB) available in November 16 this year. The new model will provide a 67% higher performance than Intel Core 2 Extreme X6800 (2.93GHz/4MB L2/1066MHz FSB). By the way, there is another Quad-Core model under development, according to the latest roadmap from Intel. The model is codenamed Yorkfield and is expected to be released in the second half of 2007. Similar to Conroe, Yorkfield will be a hybrid Quad-Core processor which shares the same L2 Cache. Yokrfield features 1333MHz FSB, DDR3-1333/DDR2-800 and dual PCI-Express 2.0 interface. Coupled with Bearlake X chipset, Yorkield perhaps will be the strongest CPU in 2007. Intel is now co-operating with software developers to optimize Quad-Core products.

View at TechPowerUp Main Site

 

[magibeg]

Intel seems to really be pushing itself as hard as it can go now. Its about damn time . AMD better hussle it up and break out something amazing before intel gets too much of a performance lead.

 

[cdawall]

lol intel in 1st its like an oxymoron poor amd gotta catch up now

 

[ktr]

amd is not focusing on performance, but money and a stronger connection in the market. look what is has acomplished in this month alone...connections with Sun, IBM, Cray and Fujitsu so that they can use the opteron socket, new computer manufactures like Apple and Dell, partnership with ATI, releasing more affordable cpus...etc. Intel...well...just released the c2d...and thats it. but the c2d are nice cpus, but the boards are pricy...

 

[Judas]

Amd will catch up its gonna take some time though, intel seem to be pushing out some nice processors these days

 

[cdawall]

either way you are not really getting a bad cpu its just the intels are a little better

 

[FLY3R]

This is point less, Half the softwear out there cant even use 2 cores, what are they going to do with 4. personaly the softwear needs to catch first.

 

[POGE]

This is point less, Half the softwear out there cant even use 2 cores, what are they going to do with 4. personaly the softwear needs to catch first.

Agreed.

 

[W2hCYK]

there is a performance increase with single core apps though.. the cores seem to be starting to be much more efficient, so either way, youre saving power on the electric bill, and gaining performance per clock..

I really want to see some programs come out for compatability with dual and quad cores. i cant imagine something like 3dmark07, but with full dual and quad core support. Or something like the Adobe series of programs, such as premiere. Using single core, my P4 @ 4GHz took about 45 mins to render a 10 minute video(sadly) and with my c2d on single core, it takes about 7 mins.. for a 30 min video.. :-D looool..

The biggest thing for me is to see Adobe Photoshop and Premiere to turn dual and quad, or higher, core capable. I cant imagine seeing a 1minute render time for a 30 minute video...

 

[Ketxxx]

sorry, but im not about to listen to something from hkepc, how many times have they been wrong again?

 

[JC316]

4 cores for the love of God why???

 

[magibeg]

You all realize we're going to look back with our 80 core systems and go "what were we thinking not wanting more cores?". This is the same idea as the not needing more then so many kb of ram. Just because we dont THINK we need it yet doesn't mean that we're not going to need it in the future. Think of buying a quad core as adding to the lifespan of your computer in a few years

 

[ktr]

by the time quad core has full support, quad core will be so much faster. if 4x4 are at 2.4ghz, chances are that in few years it will be 3.8ghz (bsing numbers). best of getting the technology that is hot now and had full support and upgrade when the next tech has full support. think of it, dual core cpu and new and has no support of their year of exsitance. buy next year the dual core will be so much faster (ie 3800x2 vs 6000+x2 or P4D vs C2D). your buying a technology where when it had support it will be a p.o.s.

 

[Ketxxx]

well said. hence why im sitting on a single core 3500+ still, not to mention its hardly a bad clocker with a possibility for 3.1GHz on a 3rd party cooler

 

[cdawall]

well quad cores would still be fun i dont care if it is useless ill just be able to download stuff/burn dvds/play gmaes/ watch tv on my vivo card all at the same time

 

[Tatty_Two]

I think its just a load of cr*p personally and yet another way to sucker the public into spending heaps of money for something they dont need, I am thinking we are gonna be seeing more and more of this bullshit over the next year or so, it seems to no longer be what you need rather than what THEY need you to have.

 

[Alec§taar]

4 cores for the love of God why???

and

I think its just a load of cr*p personally and yet another way to sucker the public into spending heaps of money for something they dont need, I am thinking we are gonna be seeing more and more of this bullshit over the next year or so, it seems to no longer be what you need rather than what THEY need you to have.

Well, not intending to "cross" you guys, but think about this, 1 single process in particular (known to be a "hog" supposedly, though I never thought so, because I use the 'trimmed' corporate model - NAV Corporate 10.x.x)

Here?

It's running 101 (had to check) threads currently, in 1 program (VPTray.exe) & 5 added services coordinating with it...

Today's OS, as I have stated here many times, are capable of "spinning" those threads out to 1-N cpu cores present.

(Windows can, currently/afaik, work up to 64 CPU's in DataCenter & 32 CPU's in Enterprise (edition I use of Windows Server 2003)).

Tomorrow's Windows? You can bank on it, that it will be able to leverage more... I have seen folks in the Linux world state online they have 'hacked' their OS to handle up to 1,024 cores!

That said? Well, maybe then, tools like NAV & such which bear multiple thread design wouldn't be though of a such "resource hogs", if they ran smoother, etc.

(And, this IS what multiple thread design & current OS thread processing via their process scheduler portions are about & can leverage for BETTER multitasking operations... just like having 2-N arms would help YOU do a job, faster, pretty much (for lack of a better analogy)).

This is just 1 process here too, mind you... I have 255 concurrent threads in operation now, not counting the SYSTEM process (which adds a TON more).

Do we "NEED" this? No, but then, I don't need computers to survive period (or, do I? It's my livelyhood, I may be an exception... most folks don't though!)

Anyhow, those are my thoughts on the current developments being made in multiprocessing hardware vs. multithreaded software & OS' for them.

This isn't so much for "outright speed" as o/c'ing & mhz gain you, but more for being able to DO MORE @ 1 STROKE/clock cycle (along with Branch Prediction/Speculative Execution etc.), via multiple CPU's handling threads, allowing for process overlap/concurrency.

Do you go faster?

E.G.-> Well, do GANNT charts work?? Yes, they do... & the SAME ideas are used in OS' process schedulers (to a HIGH degree @ least), See here:

A SIMPLE PROCESS SCHEDULER DESIGN:

http://www.emunix.emich.edu/~evett/OS/Assignments/Scheduler/SimpleScheduler.htm

It's ALL about process scheduling & concurrency (ms project uses that extensively & it has been proven for decades (GANNT's methods/algorithms, I picked up on them back in 1984 during my 1st degree in fact), to overall speedup projects tasks).

APK

P.S.=> In games, Quake 4 SMP 1.2-1.3 have been shown to get you gains, & AT ANY RESOLUTION & AA/AF usage as well: I can post those tests, but I am certain most of you have seen this already... gains diminish @ higher resolutions w/ AA+AF in "full force", but gains result, nevertheless... even playing in SUCH 'impractical' ways (extremely hi-res + high AA/AF in force).

(Though, up to & past 1600x1200 why anyone would use AA/AF is beyond me - the images do NOT 'distort/jaggy' nearly as much as they do @ lower resolutions (e.g.-> 640x480 - 800x600 etc.))... apk

 

[Alec§taar]

JC316: LOL! Love the signature bottom-line...

Ok, on track now.

This is point less, Half the softwear out there cant even use 2 cores, what are they going to do with 4. personaly the softwear needs to catch first.

Open your taskmgr.exe... I'd wager, a GOOD 90-100% of what you run, bears multiple threads... 2-N worth. The MAJORITY will be multithreaded I wager.

Once in it, go to its PROCESS TAB!

Then, use the VIEW menu, SELECT COLUMNS submenu, & checkoff THREADS as visible... then, tell us how many threads are present & how many apps of yours bear more than 1 thread.

Here? Presently, & usually, I don't have ANY single-threaded apps running & have a total of 255 threads running, that per-process, bear 2-110 threads each.

This MAY differ for you, since you may run diff. apps... it does for me on occasion too. When I run my FULL compliment of apps? I have 4 processes that bear 1 thread only.

Still, the majority of what I have? Multithreaded.

I KNOW an 80-core CPU would help that run smoother, & yes, overall faster (per how GANNT charting & process scheduling for concurrency work).

Provided OS process scheduler kernel subsystems are "prepped" for 80 cores that is... Windows currently, is not (max amt. is stated above, close to it though). There is/are UNIX/LINUX that I know can though.

I think you guys think in terms of "1 process @ a time" only (like a game etc.) but, the OS has to run ALL OF THEM, & this is where it takes effect, the most.

Now, I don't KNOW if you guys are aware of this, or not? Multithreaded code, which is the prevalent type, today?? RUNS SLOWER ON SINGLE CPU/CORE RIGS... period. Just a fact, it has more overheads there.

Running multi-core rigs assures you of being more "future proof" as well as overall faster with TODAY'S CODE DESIGNS!

Trust me: GUYS FAR SMARTER THAN I CAME UP WITH THIS & IT ALL WORKS (otherwise, why do it?)... guys like GANNT (whom I mention above & whose ideas do contribute to process scheduling designs to a high extent).

Try to entertain the notion/possibility, this is NOT 'pointless' guys... because it is proven & over a great deal of time & OS + Softwares for them.

APK

P.S.=> Then, anyhow: Once you have that data on your threadcount? Then... see my above statements in my last post. It is right man, & largely what this is ALL about & why it is occurring more than ever nowadays. Threads are proven to help speedup overall work processing on multi-CPU/SMP/MultiCore rigs... apk

 

[Ketxxx]

(otherwise, why do it?)...

Simple, just because they can.

 

[Alec§taar]

Simple, just because they can.

Hehe... well, don't know what to tell you guys, other than the above as proofs as to WHY this is used & done.

(If you guys are in academia, one day? Odds are, you'll be exposed to GANNT charting & process concurrency stuff (quite possibly, because you ARE "fellow nerds/geeks", scheduler designs), & see what I mean - in my day? You did the algorithms by hand, & proved it TO YOURSELF, it works!)

And then? "God'll send the rain..." (right into your brain!)

APK

 

[Tatty_Two]

Alec you are missing the point, I dont deny that many apps have multi threads but the performance difference between a dual core and a single core running at the same speed with the same cache on a non optimised (not specifically written for dual core)app is minimal, now minimal to me may not be minimal to you, if I am going to pay an extra £100 for a 4800 x2 over a San diego 4000+ in the UK I wanna see more than a 3-5% improvement. Now quad core will probably cost me 4 times as much for about a 20% improvement, its purely economics not performance that is the key to this debate, you are thinking performance, I am thinking economics!

let's put it another way, I have a car that goes 5MPH slower than yours but it cost me $5000 less, we live next door to each other, we both work in the same building, we leave for work at the same time driving our cars......WE GET TO WORK AT THE SAME TIME.......how is that when your car is faster than mine?????????

The speed limit is 50MPH!!!!!! make sense?

 

[Alec§taar]

Alec you are missing the point, I dont deny that many apps have multi threads but the performance difference between a dual core and a single core running at the same speed with the same cache on a non optimised (not specifically written for dual core)app is minimal,

WELL, I have a question: Is there a performance gain? Yes, you concede that much... & it largely depends on the application & coder's skill as well.

(There's tasks that DO NOT lend themselves to 'concurrency' & I can give you an easily understood mathematical example thereof next)

A = B + C
B = C + D
C = A + 1
D = A + D

Not a good one, but point is here: Myself, putting A-D's calculation on 1 thread each, won't help (@ ALL, too "linear" in nature), because A has to wait out the result of B for instance, & D has to wait out A's result & so on.

now minimal to me may not be minimal to you, if I am going to pay an extra £100 for a 4800 x2 over a San diego 4000+ in the UK I wanna see more than a 3-5% improvement.

Overall, due to process schedulers in today's modern OS', per GANNT's principles largely no less? There is & will be improvement. Small?

Perhaps... depends on the task @ hand, how well it is coded (avoiding 'race conditions' (chasing same data by 2 diff. threads & locking up)), & more...

Many benchmarks of actual programs can demo this to you in fact... & much more than 5% gains...

Quake 4 SMP 1.2-1.3, even in beta, illustrates UP TO 87% gains via J. Carmack & crew's redesign of that game in fact... probably HIGHER overall, in final build code (no debug symbols & routines usually left in final builds).

Now quad core will probably cost me 4 times as much for about a 20% improvement, its purely economics not performance that is the key to this debate, you are thinking performance, I am thinking economics!

I am thinking PURE PERFORMANCE - I am "into it", as are many of you as well... no questions about either point, really!

I say that, just judging from the topic of discussions here largely & all the tests/competitions + tips/tricks/techniques noted in both the Software + the Overclocking & Cooling forums alone.

let's put it another way, I have a car that goes 5MPH slower than yours but it cost me $5000 less, we live next door to each other, we both work in the same building, we leave for work at the same time driving our cars......WE GET TO WORK AT THE SAME TIME.......how is that when your car is faster than mine?????????

The speed limit is 50MPH!!!!!! make sense?

To quote Sammy Hagar? LOL??

"I can't DRIVE 55!!!"

(and, I don't have to really, & neither do you: But WE can take what WE have? & make the MOST of it).



(Why should I, if I don't have to... this isn't cars on a highway with limitations, other than what we can afford... buying single core CPU's today? Well, not as 'smart' as buying dual or more core cpu's, due to OS process scheduler abilities AND multithreaded code design, which IS prevalent!)

APK

P.S.=> The only "limits" on any of us, is the willingness to tune our systems to the max (this takes time & effort to SOME extent, granted), but also our wallets/bank accounts, etc.. THIS PART? I can concede...

However, why not get what gets you an extra 5% PER PROCESS? If you have a 100 processes running, it adds up TO OVERALL BETTER PERFORMANCE. Gannt (a genius imo) proved ALL of this, & I didn't believe it either... that is, until I was exposed to it, & did problems in it, by hand, back in the 1980's...

It seems that I have seen more than 1 time here, in fact several, that multithreaded design is often "scoffed" @, but when shown proofs of it making gains here (as I noted using Quake 4 1.2-1.3 SMP vs. its original single threaded design, & @ ANY resolution/AA+AF used)?

This STILL gets scoffed at... well, like I said: For those of you in academia, in either business disciplines, OR comp. sci. related fields? When you do GANNT charting & learn about scheduler designs?? Well...

"God'll send the rain..." , or rather, you'll send it to yourself, right into your own minds & viewpoints... & see it works! apk

 

[Tatty_Two]

yes but dont forget, you are talking to someone who had a 4200 x2 @2.9Gig and changed back to single core (albeit only till christmas) because I did not see any benefit at the time.

I will however become a hypocrite at christmas and go E6400!

 

[Alec§taar]

NOT an argument, but... read thru this one, closely, (in addition to the other stuff I put out earlier on the last page to this one) & see the example @ the end of THIS one, especially (it makes a point):

(Good discussion, & IF you can convince me/prove to me I am incorrect? I stand corrected, absolutely).

yes but dont forget, you are talking to someone who had a 4200 x2 @2.9Gig and changed back to single core (albeit only till christmas) because I did not see any benefit at the time.

In what - games?

This I can concede to an extent: Most games are STILL single-thread design... but, consider this:

You don't think/see/sense the world, @ the speed computers do either... sometimes, the gain's NOT apparent, though in a 'pitched battle' with the blood & particles flying? It will.

Not to you, as a human being - hence, WHY we are so "into" running benchmarks here @ this forums, as do many others populated by "performance fiends"...

I will however become a hypocrite at christmas and go E6400!

QUESTION & SCENARIO (very typical in fact):

Even IF say, a game is single-threaded...?

Let's say your other processes ARE multithreaded... & the OS can schedule their threads across multiple cores & get them done faster (which you concede is possible, & GANNT (genius imo @ least) methods show overall process scheduling gains are possible via concurrency))??

Does/WILL your SINGLE-THREADED game get more overall CPU time, & faster interrupt re-access to the CPU, even if single threaded, if the OS can process other proggies' interrupts, faster???

Remember - this is a multitasking system, & with many processes requesting time from CPU's available.

APK

P.S.=> Then again, (again): Quake 4 SMP 1.2-1.3, even in beta, illustrates UP TO 87% (averaging roughly 45% gains, thru ALL resolutions & tests @ AMDZone) gains via J. Carmack & crew's redesign of that game in fact...

Actually, probably even slightly HIGHER overall, in final build code (no debug symbols & routines usually left in final builds)....

Trog100 made the MOST logical 'argument' vs. this here I have seen:

He stated that using resolutions 1600x1200 & above in Quake 4 SMP doesn't show gains as large as lower res does, when using AA/AF in FULL FORCE settings... & the tests are "rigged" & impractical, & that "REAL GAMERS DON'T PLAY @ LESS THAN 1600x1200" or, something along those lines!

(I disagree strongly, & here is why):

Why use AA + AF, when the images do NOT distort @ those res., period? Pointless "overprocessing" of the imagery border frames!

E.G.-> I have one of the MOST powerful GPU's there is, & it plays MISERABLY (even in SMP builds) @ those resolutions ESPECIALLY w full AA+AF enabled, 1600x1200 is iirc, highest I can go...

TOO slow!

I'd rather play using AA+AF on FULL, @ say, 800x600, & get the gains possible via threadusage & good design... larger gains, %-wise, than gained in higher res.

J.C. & crew @ IDSoftware, imo @ least? Are some of the best out there (if there is such a thing), & they too, went w/ multithreaded designs & the results?

Well, speak for themselves... apk

 

[Ketxxx]

Hehe... well, don't know what to tell you guys, other than the above as proofs as to WHY this is used & done.

(If you guys are in academia, one day? Odds are, you'll be exposed to GANNT charting & process concurrency stuff (quite possibly, because you ARE "fellow nerds/geeks", scheduler designs), & see what I mean - in my day? You did the algorithms by hand, & proved it TO YOURSELF, it works!)

And then? "God'll send the rain..." (right into your brain!)

APK

your missing the point, just because you can do something, doesnt mean it needs doing. or to put it another way, are you personally doing something because you know it needs doing, or are you doing it because your being told it needs doing? theres a suttle, but substantial difference.