Socket 1366 obsolete, SMT a 'gimmick'

Page 9 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
CPU Thermal Design Power of i7-860 is 35W less than 920.
http://processorfinder.intel.com/details.aspx?sSpec=SLBJJ
http://processorfinder.intel.com/details.aspx?sSpec=SLBEJ

Amazing, the chart below shows 860 SYSTEM uses 30W(which is very close to 35W) less than 920.
cjl: Nope this is due to LGA1366's 1 additional channel consuming 30W(which is 30/115=26% of the total SYSTEM idle power) more power.
20087.png

Now, let's assume LGA1156 i7-860 runs as hot and use as much power as LGA1366 i7-920 which is what cjl THINKS.

1) i7-860 runs at higher frequency.
2) i7-860 has more on-die features(DMI, on-die NB, PCIE controller, 2channel IMC) and so 43 millions more transistor than 920.
( http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634 )
3) i7-860 runs as hot and use as much power as 920, assumed.

This means 860 CPU powers up more features(43 millions more transistors) and runs at higher frequency but use the same power as 920.
Hence, 860 is more efficient.

I can imagine what cjl is going to write.
He will be saying:

Nope. Try again.

LGA1366's CPU's (3channel IMC and QPI) consumes more power than the LGA1156's (DMI, on-die NB, PCIE controller and 2channel IMC)

OR

Nope. Try again.

i7-860 runs much hotter than 920.

OR

Nope. Try again.

The one additional channel ALONE consumes 26%(30W/115W=0.26) of the total SYSTEM power.

(So tri-channel consumes 3 x 30 = 90W which is 78.3% of the total system idle power????????!!!!!!!!!!)

You need to take LGA1366's off-die NB into consideration, because me, cjl, only care off-die NB and on-die NB is ignored.

OR

All of the above!
 



Well, I really did grave this thread out. Andy we need that you keep replying since looks like you're the only one "defending" 1156 plataform. Well let me show some "insights"

1) The fact the leap from 4 nm5 to 32 nm had been done 5 months later is much more related to market and I'm sure they're working on this transition far far before the release of D0 there are much complex issues that came with this "leap".

2)
As for my frequency/voltage comment, you really don't understand the idea of multiple clock domains, do you. A QPI link might need a higher clock speed than a PCI-E controller for example (both of which are on separate clock domains from the core). That is why you need to account for what the additional transistors are used for in a CPU, not just how many there are. !

Andy, Cjl's point here is extremely right. Think, we have the leakage factor so when we run at faster speeds the transistors switch between states more times in the same delta time. So if the 43 million transistors that 1156 processors have more than 1366 processors are there making tasks that run in a lower clock domain inside the processors (I believe in this scenario), the leakage from 1156 doesn't need to be smaller.

3) Think with me, 1366 processors have 2 cores linked by a bus while 1156 have a much more elegant unicore architecture, right? You know that the frequency and the width of this bus, the data being exchanged in 1366 dissipate more heat since the cores are physically far on die, multiply the little amount of heat being "lost" in this path every time by the frequency * bus width factor and you got some watts from Joule's efect (not the leakage). I believe that this dissipates much more heat than the 43 million extra transistors working at lower clocks in 1156 doing the same tasks adressed to the north bridge in 1136 plataform.

4) So in my opinion the major improvement in the 1156 is the unicore. Even though, I still prefer the north bridge doing more things. 1156 is the mainstream because of that just look at the I5 6xx series for me it's pure nonsense that onboard "VGA" keep that to the I3 family only and IMO there won't be new I5 750 like processors (without that VGA) with 32 nm released to 1156 plataform.

5) I do believe the transistors have the same leakage levels. Why would intel make such a change in their processes in the "last generation of 45 nm processors" (as far I know/expect)? All the above are phisically related arguments, so there's no doubt about it, unless thermodinamics laws had changed.

6) Now in my scenario: Heavy overclock. What would i preffer? A processor with a "north bridge" built in? No. Why? Because in 1136 I can cool the two things apart and have exclusive cooling solution for each one.

7) I don't even considered the I5 750 since, yes, I will benefit from HT. Here I7 860 is 25 dolars more expensive than I7 920 D0. The motherboard difference was about 70 dollaras (45 discounting the processors diference), and the memory well I paid more but I got 6gb instead 4gb (I use more than 4gb very often, but 6gb is enough). So in my case, I've spent less money in one 3x 2gb kit than in two 2 x 2 gb kits).

8) 1156 processors (except gulftow in 1366) have a more agressive Turbo mode. So the 1156 processors are much more likelly being more efficient in speeding up/down. But to test the architecture efficiency these benchmarks with turbo mode enabled at stock doesn't fit. The ideal scenario would be the two processors at the same clock with cpu clock stepping disabled and at stock voltages.

9) I do kwow that HT make the processor dissipate more heat. But in you charts how do you justify the I7 860 using 4w (or 5w) less than I5 750. Were those tests made at the same system (just switching processors) ? I'm presuming HT was enabled since it is a stock benchmark.

10) Finally Andy, I agree if you will not benefit from HT (I will) and Triple channel (most probably you won't benefit from this second) you'll just be fine with a good I5 750. Otherwise, IMO (se my case) between the two I7 families I still preffer the 9xx family.
 
Another thing: it's totally aceptable that a in real world applications a higher clock processor run cooler than a slower one, since it computes faster and came back to idle sooner (with or without any kind of speed stepping). For instance, the notebook c2d processors in macbooks the 1.88Ghz version runs hotter than 2.16ghz (I am saying this from my experience with 2 macbooks of the same generation with the same amount of cache.
 
@armartins:

Believe what you believe.

I don't care about what you believe, such as LGA1366 i7-920/LGA1156 i5-750 system costing about the same in NZ and LGA1366 being as efficient as LGA1156. Because:

1.
Don't lie! I am in NZ too.


Cheapest i5-750 $328.42


Cheapest i7-930 $492.58


The difference in CPU price alone is $164.16. Hence, it is impossible for 1366 to be on the same cost as 1156 unless you are comparing crappy X58 MB + highest end P55 MB.

Upper mid-range X58 & P55 Comparison:

GA-X58A-UD5: $516

GA-P55A-UD5: $332.25

Price difference: $183.75

2.
CPU Thermal Design Power of i7-860 is 35W less than 920.
http://processorfinder.intel.com/details.aspx?sSpec=SLBJJ
http://processorfinder.intel.com/details.aspx?sSpec=SLBEJ

Amazing, the chart below shows 860 SYSTEM uses 30W(which is very close to 35W) less than 920.
This is due to LGA1366's 1 additional channel consuming 30W(which is 30/115=26% of the total SYSTEM idle power) more power?!!!!!
20087.png


Now, let's assume LGA1156 i7-860 runs as hot and use as much power as LGA1366 i7-920.

1) i7-860 runs at higher frequency.
2) i7-860 has more on-die features(DMI, on-die NB, PCIE controller, 2channel IMC) and so 43 millions more transistor than 920.
( http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634 )
3) i7-860 runs as hot and use as much power as 920, assumed.

This means 860 CPU powers up more features(43 millions more transistors) and runs at higher frequency but use the same power as 920.
Hence, 860 is more efficient.

Nope the 30W more power is due to:

LGA1366's CPU's (3channel IMC and QPI) consumes more power than the LGA1156's (DMI, on-die NB, PCIE controller and 2channel IMC)?!!!!!

OR

i7-860 runs much hotter than 920??????!!!!!!!!!!!!!!!

OR

The one additional channel ALONE consumes 26%(30W/115W=0.26) of the total SYSTEM power??????!!!!!!!!!!!!!!

(So tri-channel consumes 3 x 30 = 90W which is 78.3% of the total system idle power????????!!!!!!!!!!)

OR

All of the above!

3.

1) You must be Intel's CEO that you know their pricing 2 years later!

2) 2-5% with 2x HD5870 means 3-4fps. Is it worth the extra $70?!!! I could get a better gfx with extra $70 to beat you by 10fps+.

3) Can you put the $999 Gulftown in your LGA1366 if you can't afford it?
2 Years from now when I'll be switching my I7 920 I'll get a six core for U$300.00 as I did with the 920
Even Intel's CEO doesn't know their pricing 2 years later!
Do you know how expensive C2Q extreme is now? Extreme will never become that cheap.
In addition, people will be using the cheaper and faster sandy bridge two years later.

4) 99% games/apps can't even take the advantage of 4-core CPUs, let alone 6-core.
Later games can take the advantage of 6-core soon
It is impossible considering it took 3 years for most games to take the advantage of merely 2/3-core.
And further more try to keep a 4.2ghz+ OC 24/7 on a I5 750 or a I7 8xx and let's how much Voltage you need
The voltage required on LGA1366 is lower, because P=VI and LGA1366 has worse transistors that generate a lot of useless/harmful leakage current which is the reason why LGA1366 CPU run so darn hot.

Isn't keeping CPU cool the reason why you want lower voltage? LGA1156 give you lower temps even with higher voltage and this also means that it is much more efficient than LGA1366.

In addition, the CPU voltage on LGA1156 not only supply core voltage but on-die PCIE controller voltage as well.
 
@armartins:

Truth is truth and it will never become a lie no matter how many peoples disagrees.

1)
Pricespy.co.nz shows that LGA1366 costs much more than LGA1156 in NZ.

However, you BELIEVE they cost about the same in NZ. What a great math background there you have!

2)
The Anandtech shows that LGA1156 i7-860 system performs better than i7-920 system although i7-860 system consumes 30W less.

You BELIEVE LGA1366 i7-920 is as efficient as LGA1156 even though LGA1366 920 performs worse and consumes 30W more power than 860 system. Such a great physics background there you have!


Hence, I don't bother listening whatever your beliefs are!
 
Honestly all of you guys are a bunch of hypocrites and I have argued your points that you now use in your favor.

I was a mad pusher of dual cores when everyone else pushed the q6600 and now I hear crap like this:

4) 99% games/apps can't even take the advantage of 4-core CPUs, let alone 6-core.

*Slams face into the ground*

Make up your friggin minds already - don't change it just to suit a current debate you're in now. At this point if you have not figured out future proofing your HD to match outdated software is a dumb idea then I don't know how to help any of you.

1156 > unless you can utilize the higher logical threads. 99% of us can't use a quad core much less a quad core with 4 logical threads attached. One thing I will say is nobody saw the 1366 socket to be a bogus socket until the i5 release so nobody really wins, a lot of people just lose. Screw Intel.
 


I have always been a quad pusher. It has benefited me more than it has not and since newer quads these days overclock as high as duals and are not that expensive, there is no reason not to get a quad. Its just like when dual cores first came out. They couldn't overclock as high as a single core, most apps and games couldn't use more than one core but after a while we went full dual cores only.

Its the cycle of PC life.

I think the 1366 socket will be more viable in the long run for certain things and in gaming way down the line. I am pretty sure that a gamer in 3 years with a 1366 will be fine but with a 1156, 775 or AM3 even will have to upgrade.

But thats just my views on it. I bought a quad when duals were kings, and now my PC will last me another 2 years meaning 4 total before I really would need a CPU upgrade.
 
Bold Letters! Caps!! [:jaydeejohn:7]

Babble
Burble
Banter
Bicker bicker bicker
Brouhaha
Balderdash
Ballyhoo

Comments
Cliches
Commentary
Controversy
Chatter
Chit-chat
Chit-chat
Chit-chat
Conversation
Contradiction
Criticism


Debates
Discussions
(These are words with a 'D' this time!)
Dialogue
Dualogue
Diatribe
Dissention
Declamation





Expugnations! Exclamations!! Enfadulations!!



[:jaydeejohn:7] [:jaydeejohn:7] [:jaydeejohn:7]

 


+1 [:jaydeejohn:7] [:jaydeejohn:7] [:jaydeejohn:7] [:jaydeejohn:7]

Why are people arguing a such a pointless thing......
 


And yet he posts twice for every contradiction to his onw beliefs. This guy is hilarious.

[:jaydeejohn:7]
 


Right now I'm pretty pissed because chrome just reloaded this page and I've lost my anwer.

Yes, indeed I'm realizing that you "don't care" you're like a rock. By the way if you really don't care why do you answer?

I've never said I was N.Z I'm Brazilian and buy my stuff mostly at Paraguay. I said I7 920 is cheaper than I7 860 not I5 750. First you READ then WRITE and don't change the subject to N.Z or anything else. Lets difference the mens from boys.

86078 | CPU INTEL-1156 i7-860 HT 2.80 2.5GT/S 8M | 345.00
92048 | CPU INTEL-1366 i7-920 HT 2.66 4.8GT/S 8M | 323.00
http://www.master10.com.py/importar/arquivos/lista.master.txt

Prices from one of many stores, note that I bought mine 920 for U$310 than it's U$35 not U$25 dollars cheaper I was wrong.

You're saying that 43 million transistor more on 8xx would justify more heat. You can't look at this alone, It's like saying a company with 30 employees can't spend more money paying then that another company with 40 employees. Why? You simply ignore they're functions (speeds) and don't assume that each one may receive a bigger or smaller paycheck (heat lost). You ignore that those 43 million transistor are realizing tasks of the 9xx north bridgh and in your own quote you've answered. The physics I was saing is in the earlyier post (you probably didn't saw) that the extra heat came from the QPI (mainly) linking the two physical cores on 9xx, the unicore at 8xx series it's the big improvement of the series. There are (since long long time ago) many actual speeds inside the processors, cache, bus, may work at different clocks (multiple domain clocks) and guess what? They are made of transistors. And about leakage to change the leakage the whole process would need to be revised this goes very deep in physics and chemistry, dopping, semi-conducting, MONEY, research,etc such a big change ocurs along with changes in transistor's size . Just try again quote my last post, don't keep requoting you and talking about others, N.Z. and stuff. Processor with north bridge built in? I'm out. Because? See last post, higher voltages, more room to vDrops, more points of failure INSIDE the processor. And yes I keep thinking that in 2 years 6 core processors with an MSPR of U$300,00 to U$350,00 (same as I7 920) are pretty viable.

Andy honestly YOU FAIL, try again. (and please don't take this as personal)
 


The whole point is that the stronger turbo mode on 8xx make the diference too, I was not contradicting myself just READ what I wrote. It's overclock with no turbo mode my case. Your name says all.
 
It was me who bought up the pricing regarding 1156 and 1366 being almost the same in NZ originally. Just because in the big scheme of things, it didn't seem to be a big difference, NZD150 i7 vs i7 or NZD300 i5 vs i7 (comparable Asus boards with 16 phase power). If you want more than 16 PCI-E lanes for expansion, then the price gap becomes much closer because only (as far as I know) the high end boards come with NF200 chips.

But that's probably not relevant, since 99% of people probably wont need more than 16. Getting a board with USB 3.0 is probably a better investment. I'm still glad I've got what I have though, mostly because everything works properly, mainstream boards in the past have let me down many times.

Also with the 32nm i980X coming out, hopefully we get lower 1366 models in 32nm as well, which I would expect would mean new plant, which should mean any low leakage transistor tech would flow on to those as well.

andy - bummer about your power usage not reporting correctly, not that it matters really. O/C is 3.6 @ 1.24V, large FFT core temps are ~73C max on two cores, and 77C max on the others, in games it stays under 70C which is fine but I want 4GHz without jumping around in the 80's. I have another cooler turning up today (Thermolab BARAM) to try. Also heard I should spin my S1283 around 90 degrees (which would explain why two cores are 5C hotter than the others). If this doesn't work, just going to water cool the bastard.
 

You fail in lying because of pricespy and Anandtech.

Try again!
 

Because there are many hilarious readers just won't listen if I speak only once. Actually, they don't listen no matter how many times I said.

Furthermore, it is not my belief but fact.

It shows clearly in the chart that LGA1156 860 performs better and consumes 30W less power than LGA1366 920.

LGA1366 owners are so hesitate to admit it just like they are hesitate to admit that the LGA1366 upgrade path is unaffordable to 99.99% and keep whining something they can't afford.

arguing-with-a-brick-wall-democrat-republican-politics-obama-demotivational-poster-1242061037.jpg
 

1) Yes, CPUs should be fine as long as they are under spec temp.

But why do people always want lower temp? Because lower temp will make CPU much healthier.

2) my i5-750 OC 3.6@1.21V, all full load temps under 50C(48C-44C-44C-43C) FYI.
 
Nice.
I'm much happier now.

Replaced my S1283 with a BARAM, and also rotated it 90 degrees, and used fresh MX-2 instead of old AS5 I had lying around. Results I'm happy with. Only problem is I need more fan airflow, and I have better fans, but Asus thought it'd be hilarious to make the CPU fan control PWM only, and I only have 1 PWM 120mm fan grrr

But anyway, Large FFT its 63-63-65-65 for the first 5 minutes, then slowly rises to 71-71-68-68. Those being core temps, the "CPU Temp" reading would be within Intel spec of 67.5. This is at 4.07Ghz @ 1.3V, so I'm pretty happy. I haven't even tuned it yet, just threw those numbers in, so I could probably raise the frequency or lower the voltage.

My DDR3-1333 is running at 1450ish as well, CL7, happy with that too since it was cheap memory ($NZ130 for 3Gb 7-7-7-20).

Apologies to everyone else for the off-topic.
 



The brick wall is you Andy. Can't you read things? I'm not saying that at stock an I7 860 can't beat, consume less power and be cooler than an 920 consider my scenario please and stop with this smalltalk "talking about me with others".


| 79741|PROC. INTEL CORE I7 860 S1156 2.80GHZ/2.5GT/CACHE | 345.00
| 66908|PROC. INTEL CORE I7 920 S1366 2.66GHZ/4.8GT/CACHE | 315.00
http://www.navenet.com/home/lista/ListaNave.txt


Just another store over here. You don't live here and you don't know the pricing of things over here. And another thing do you have my bank account password to know what I can or not affort? I can post here updated threads from like 10 foruns with people comparing I7 8xx with 9xx for OC and they keep going to the 9xx... even paying more for x58 motherboards (and processors over there) and that's a fact.
 
@armartins:

| 79741|PROC. INTEL CORE I7 860 S1156 2.80GHZ/2.5GT/CACHE | 345.00
| 66908|PROC. INTEL CORE I7 920 S1366 2.66GHZ/4.8GT/CACHE | 315.00
http://www.navenet.com/home/lista/ListaNave.txt
920 cost a little more or exactly the same as 860 in most countries prior to the release of 930.

Yes, 920 is now cheaper. However, don't forget that the X58 MBs cost much more than the P55 ones and 920 performs worse as well.

And another thing do you have my bank account password to know what I can or not affort?
I don't know if you can afford it or not. However, I do know that 99.99% of LGA1366 owners can't afford or not willing to pay for that much.
This is the reason why I said LGA1366 owners should stop whining how upgradable X58 is.

I can post here updated threads from like 10 foruns with people comparing I7 8xx with 9xx for OC and they keep going to the 9xx... even paying more for x58 motherboards (and processors over there) and that's a fact.
No doubt, that is a fact. However, the fact proves nothing regarding the performance and OC potential.

920 OCs higher? No, the maximum OC for both 860 and 920 are 4.2GHz on average. Furthermore, are you aware of that HT has to be turned off for 4.2GHz OC on most chips and therefore they are no different from the i5-750 at 4.2GHz?

Now assume that the maximum OC of 860 and 920 are 4GHz and 4.2GHz respectively which is what LGA1366 owners believe. Do you think that the 920 at 4.2GHz without HT would beat 860 at 4GHz with HT?

In addition, have you noticed that 920(HT supported) loses to 750(HT not supported) in many COMMON apps(including games which are the most popular apps) even they are at the same frequency? The 920 can only beat 750 in apps like 3D rendering and Video encoding which are used by very few percentage of people, but you can always get 860 if you are this kind of user.

Read through the WHOLE article linked below for 920 vs 860 vs 750:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3641&p=1

I did do the search before purchasing the P55 platform instead of just listening what the majority said.
Majority's thought are always right? No. In many cases, following the majority only makes you SEEMS to be right instead of BEING right.

Why more peoples recommend LGA1366 then?
Because LGA1366 was released about 1 year earlier and so it has much more owners than LGA1156. Hence, a lot of peoples who have no idea on what the difference between both platforms are got overwhelmed by the huge numbers of LGA1366 owners and therefore results in more and more LGA1366 supporters. Besides, LGA1156 has just released for about a little more than 6 months.

stop with this smalltalk "talking about me with others".
My bad. I was focusing on programing instead of this. It was Spidersweb instead of you. Sorry.
 
Status
Not open for further replies.