Printable View
XFastest has posted preliminary clock speeds for the upcoming sandy bridge processors from Intel. Clocks are interesting; the i7 2600 at 3.4Ghz seems rather interesting if one takes into account the aggressive turbo I'm sure Intel would implement in sandy bridge. 4Ghz, anyone?
http://img204.imageshack.us/img204/273/sandys.png
Would be around 3.5 not 4ghz though.. the i7 2600 is that high end or low end cpu's? Any news on how it scales to say a i7 920 or 930? its obviously faster at default but how many cores?
this chart is purely for reference :D
the core i3/i5/i7 brand is starting to become strong, of course they'll keep it...
anybody saw SB running CB 10?
http://www.youtube.com/watch?v=TkP4r...layer_embedded
do all of these have integrated graphics?
the LGA1155 CPUs (2-4 cores) will replace clarkdale/lynnfield (LGA1156), not the higher end bloomfield/gulftown i7s ;)Quote:
Originally Posted by Johnny87au
Is it known yet how much faster they will be?Quote:
Originally Posted by -Sweeper_
Hmm i swear i've read somewhere that sandybridge will replace the i7's ..hmmm "scratches head"
That's what I thought too.......Quote:
Originally Posted by Johnny87au
an hexa/octo-core version of SB will replace the current i7s, not the dual/quad-core models expected for early 2011Quote:
Originally Posted by Johnny87au
Gotta wait for someone who has more knowledge to post :ROTF::up:Quote:
Originally Posted by kuhla
Minor speed improvements over Lynnfield clock 4 clock, but it will have an improved memory controller, AVX and should have better stock frequency (which should translate to good overclocks too).Quote:
Originally Posted by natty
A little info here, but plenty more if you look around.
Sandybridge is a micro-architecture and Core i7 is a brand. If Intel is happy with the current naming scheme they are not required to change it. The "Core" branding that started with Merom in 2006, changed to Core 2 and then Core i7. The i7 branding isn't even two years old. Certainly Sandybridge will replace the Nehalem architecture, but it doesn't require Intel replace the i7 branding. The certainly can if the need to, but it doesn't mean they will.Quote:
Originally Posted by Johnny87au
My confusion lies with the current socket situation, not the branding which I don't think will change.Quote:
Originally Posted by sdsdv10
This may be obvious to everyone else but I like to think type out loud.... after looking more into it, I think Sandy Bridge on the LGA1155 socket will initially replace LGA1156 and then probably down the line replace LGA1366 with LGA1356. My basis for that reasoning is the motherboards that have been shown off so far. The P67, H67, Q67, etc. that we have seen so far with the LGA1155 have been dual channel with onboard video yadada so just like the current LGA1156 stuff. I would expect X68, or whatever replaces the X58 chipset, for the new enthusiast level stuff to have triple channel memory setups. Since we have not seen that yet, therefore Sandy Bridge is just LGA1156 replacement for now.
We shouldn't exchange these CPUs with the high-end.
We get performance and mainstream CPUs in this round, but high-end Sandy Bridge is expected much later, in Q3 2011 according to current info.
1155 = BCLK 100 = more multi = expect more fun overclocking
any news of 1366 replacement? (1567?)
4c/8t 8MB l3 cache and the "same" 4c/8t 6MB ???? Disabled some cores in 8MB l3 version???
disabled cache*Quote:
Originally Posted by FlanK3r
Intel does this a lot, nothing new here, move along.
These things have the iGPU right? The TDP should be 95W,even for the 3.4Ghz 8T model?
SB vs fastest i7 on market and it could render CB10 30 seconds faster?:shocked::shocked:
fastest i7 refers to i7 980x or i7 975?
We have been over that topic before...
The 720p version of the video shows a time of 44 seconds an apparentlyQuote:
Originally Posted by haylui
a score of 19641 for Sandy Bridge. This puts the Sandy Bridge in the
range of an i7 975 / i7 965.
The 64 bit version of Cinebench 10 is used.
http://www.legitreviews.com/images/r...inebench10.jpg
http://www.legitreviews.com/article/986/8/
http://images.tweaktown.com/imagebank/corei7975g_05.gif
http://www.tweaktown.com/reviews/277...ed/index6.html
Regards, Hans
Quote:
Originally Posted by Hans de Vries
just wonder how could it be faster than the other demo machine by such a huge margin
according to the Coolaler leak of SB early benchmark, it is about 10~15% faster than Nehalem
Possibly may have been a Performance / Watt comparison using a core i7M chip @ something like 2ghz, Vs SB at 3ghz or so
Mobile chips is a good guess since their "top of the line i7" doesn't score anywhere close to the actual top of the line i7 in CB10(and I mean QC not Gulftown)...
the video is on a mobile platform SB vs 920XM...
last 2-3 minute here:
http://www.youtube.com/watch?v=kjEjP...layer_embedded
Thanks for posting the video.So it's a mobile platform after all. No wonder why i7 scored lower,it runs @ 2Ghz when all the cores are loaded. We already have SB numbers in Cinebench over at coolalers website.Quote:
Originally Posted by IDANPE
Quote:
Originally Posted by IDANPE
according to what I've seen, SB completed in about 40sec while 920XM uses 68sec
in terms of time, it is about 1.7 times faster for SB to render the same scene???:shocked::shocked:
I wonder if this is the job by AVX?
Maybe it's a "modded" version of Cinebench making use of AVX? :shrug:
or it's running at 3+Ghz...
mobile 3GHZ 4c/8T cen't be bad thing :yepp:
I wonder why the desktop version of Sandybridge only clock 3.4Ghz while the mobile version clock 3Ghz+......
if its mobile platform vs desktop platform then its fail again intel .....
maybe they had fear show it in better R11.5 version, this version is more "reality" than lagging R10 at more than Quadcores...
It's called higher clocks + higher Turbo(thanks to 32nm).Quote:
Originally Posted by haylui
well, the 920XM platform, while it IS a mobile platform, isn't very mobile like.. 55w TDP + the chipset .Quote:
Originally Posted by superrugal
While i'm not discounting SB's performance here, I think if anything it shows exactly what SB is aimed at.. very good performance/watt.
To show this in the BEST possible light (which is the smart thing to do when you're marketing something) it makes sense to use a very high power consumption mobile platform as a comparison for perf/watt improvements
it's mobile vs mobile.Quote:
Originally Posted by Sn0wm@n
only 3.4 max?
same clocks as the current chips... what gives? :shrug:
i wonder why they keep the guaranteed clocks low and boost turbo instead...
and 10-15% higher ipc of sb vs current gen? where? its half of that from what ive seen, and thats vs lynnfield/clarkdale which have a lower ipc than bloomfield/gufftown...
gulftown and even bloomfield continuing in the market for quite a while until sb launches says a lot about how much faster sb will be, doesnt it? :P
ok, can anyone reconcile the chart with dresden's finding in BOINC?
/citavia.blog.deQuote:
P.S.: Since looking for BOINC stats of Ontario and Llano was successful I tried the same for Sandy Bridge (although there are other benchmark results out there):
Sandy Bridge Stepping 3, 2.2GHz
Sandy Bridge Stepping 2, 2.0GHz
Sandy Bridge Stepping 0, 2.2GHz
Sandy Bridge Stepping 3, 2.4GHz
(see end of blog post)
Fudzilla talked about the same DDR3-1333 only officially supported:shrug:Quote:
Originally Posted by EvilOne
Did you say that it is same clock as the current chip ?Quote:
Originally Posted by saaya
i7 860 is 2.8GHz, and i7 870 is only 2.93Ghz !
Also, SB is supposed to have better performance per clock, and better performance per watt
3.4 GHz without turbo is great default clock!
:ROTF:Quote:
Originally Posted by saaya
http://www.anandtech.com/bench/Product/46?vs=107
lynnfield has the exact same performance as bloomfield, there are only very very few apps where the additional bandwidth helps bloomfield... get your facts right. :cool:
Quote:
Originally Posted by FlanK3r
Specially when 1090T get i7 965 numbers. :rolleyes:
while it's $600 cheaper.Quote:
Originally Posted by Nintendork
you linked a bad comparison to make your point, ended up reinforcing saaya's instead ;)Quote:
Originally Posted by Hornet331
The i7 870 has a higher Turbo freq (3.6 vs 3.2Ghz) , and not all those tests are fully multithreaded.. ones that are do show a few percent on avg higher performance on Bloomfield :p
I thought it was pretty obvious it was a mobile chip IE 920xm when they said their fastest i7 and it was only topping out at 2ghz. That and the fact that intel and everyone in the know has said over and over and over and over again that the first SB parts will be mobile and desktop mainstream. Why would they then compare it to a high end chip? Why is it so hard to draw simple conclusions but so easy to make random conspiracy theories?
How much of a jump are these going to be, performance wise, compared to the current 1156? I've been thinking i3 530 @ 4.2ghz or greater, but is this really worth it? It's not like we can't drop these into 1156 boards (Thanks a lot INTEL!).
At least keep the 1156 mounting holes. We don't need ANOTHER set of coolers for this.
I like the idea of say 31x100 = 3.1ghz. Imagine OCing something like that. 31 x150 = 4.65ghz. I can see people putting LN2 on these and getting 7+ ghz no issue.
Interesting strategies Intel and AMD are using. Intel is releasing midrange to begin with where AMD is going full throttle and releasing their high end right away. Then again AMD is banking everything on there chips.
Quote:
Originally Posted by [XC] Synthetickiller
last i heard about it is the placement of some pins that is different so different thing different place = non compatibility ... so no ... but yes we will still blame intel for a long platform longevity .........
Intel and Socket longevity are 2 things that will never coincide. This is mainly because they change so much that between the architecture if they kept the same board problems would pop up out of nowhere. Intel's strategy of changing the socket with the processor keeps these problems from occurring.
Thank you, sandybridge is just to replace the 1156's.. Hope they bring out the high end sandy bridge cpu's before 2011 :down:Quote:
Originally Posted by Sam_oslo
octo-core in 2012?Quote:
Originally Posted by -Sweeper_
This isnt the best move for Intel to make seeing as their competition is starting by releasing their new architecture with 8 cores 16 threads in the beginning of the year. The Bulldozer architecture also has the AVX so Intel had better hope their CPUs Clock like nuts.
Even if nobody wanted the new intel chips over the new amd chips, intel would still sell more by virtue of their manufacturing edge they have over amd. Intel survived Athlon 64 on sheer marketing and manufacturing alone. Intel also advertises, so they have that brand name zing, while AMD... are f'ing retards when it comes to advertising their products. AMD could seriously take Nvidia to task if they would just start advertising Radeon graphics as a brand name... instead Nvidia is still the brand name in the mind of the uninformed.Quote:
Originally Posted by =SOC= Admiral
16 threads !?Quote:
Originally Posted by =SOC= Admiral
Bulldozer don't support hyperthreading. Please do some research
i think you're missing out on one dirty B word.Quote:
Originally Posted by Dalten
Then compare it with the 950, same multi in turbo with load higher then 2 cores, nearly the same results.Quote:
Originally Posted by mAJORD
Its not Hyperthreading. AMD is using their Cluster Core design which is 2 sets of execution units fused between 2 128bit FP schedulers (AVX). Each set of execution units gets its own thread. Hence one Cluster Core and 2 threads. Bulldozers full design calls for 8 Cluster Cores and 16 threads And shared L3 between all cores.Quote:
Originally Posted by dartaz
Intel seems to be keeping with their same tried and true system but on a new architecture.
execution units are "pipelines." 8 for int and 2 for fpu(FMAC) per "cluster core."Quote:
Originally Posted by =SOC= Admiral
keep in mind that top-end BD will only have 8 cores, 8 threads, 4 "cluster cores" for desktop(zambezi).
16 cores, 16 threads, 8 "cluster cores" is interlagos. a server part only.
SB 3Q 2011 top-end desktop will be capable of 8x256 bit AVX(although MADD only; or 8x128 bit fpu.)
Their whole naming and designation scheme has gone every-witch way, thats why i was slightly messed up with my info. After reading a lot more in some different places its slightly clearer but cant wait till next months info dump. Hope Bulldozer clocks like these Hex cores do. But the 2011 round of the Intel VS AMD should be very interesting.
q3/2011 ;)Quote:
Originally Posted by hondacity
Yeh my bet is intel >amd like it has for the last few years :shakes:Quote:
Originally Posted by =SOC= Admiral
those are ancient 45nm lynnfield chips, big deal...Quote:
Originally Posted by dartaz
compare it to a 32nm chip ie the i5 600 series and suddenly those clocks dont look very impressive at all...
an i5 680 comes clocked at 3.6ghz, and no, thats the default clock, not turbo mode...
sb looks like a quadcore clarkdale with better igp, but who cares about 2 more cores and igp? :/
lets face it, in a normal usage scenario you wont be able to tell a sb and clarkdale machine apart...
yeah like 6% and who cares about power in a desktop machine... :PQuote:
Originally Posted by dartaz
1. i talked about ipc, and not the performance of cpu a vs b, what intel decides to bin turbo mode at etc has nothing to do with what a cpu design is capable ofQuote:
Originally Posted by Hornet331
2. at stock speed and looking at the cpu level perf with turbo mode and all, yeah, but when overclocked even that comparison shows a lead of gulftown/bloomfield.
again, if you talk about IPC they are the same.
Bloomfield sometimes has advantages when there are apps that need lots of bandwidth, but in general they have the same IPC if you remove that advantage by running bloomfield in dualchannle mode they performe exactly the same.
Again get your facts right. :p:
its harder got stock 3.6 GHz six or octa cores than dualcores...and TDP up to 130W too
where is my LGA 1366 32nm 260€ cpu :'(
*still dreaming*
sb is a cost down of 1155 for OEMs...Quote:
Originally Posted by Johnny87au
remember 30h6's clarkdale review?Quote:
Originally Posted by Hornet331
i dont know if the efficiency has improved now that everybody knows how to tweak the gmch and push for high qpi... at stock speeds maybe... but when overclocked... i dont think so... why do you think all top scores are still done on 1366 and 1156 is waaaaay behind?
god, you just love wearing your 4ss as your hat, dont you? ^^Quote:
Originally Posted by Hornet331
Wish I could get away with a comment a like that. Oh well, maybe when I get 34,000 posts. :rofl:Quote:
Originally Posted by saaya
Because there are some limitations for socket S1156, e.g power layout of the socket iteself, the obviouse dual vs tripple channle and more pci-e lanes and some other minor detail, which makes S1336 the better ocing platform for extreme ocing. But this doesn't change the fact that they have the same IPC.Quote:
Originally Posted by saaya
If you look even closer, you see that normal 9xx dont clock much higher then the 8xx (check records on hwbot, they are all ~5,7-5,8ghz the exeptions are the 975 and the 980
Yes I do, the facts are there I just pointed them out. It's not my fault that they don't match with your opinion... :p:Quote:
Originally Posted by saaya
I think they'll be interesting and hope to get my hands on 'em early :p: :D Liked 670's for Pi and frequency shots, sandybridge might be doing good there :hehe:
Do you know when they will replace the 1366 high end bloomfield's? or yet nehalem?????Quote:
Originally Posted by saaya
what, im calling him rude, whats wrong with that? i didnt mean it as an insult and i dont think he saw it as one...Quote:
Originally Posted by OhNoes!
hornet?
compare the highest scores of 1156 and 1366... there is a pretty huge gap, not just 5% or 10%...Quote:
Originally Posted by Hornet331
well thats rather a hint that the extra vcc and vss pins dont really matter all that much, doesnt it? :DQuote:
Originally Posted by Hornet331
no, i agree, you were right, the ipc at stock clocks is about the same for 1156 and 1366... what i meant is that you seem to really enjoy pointing out that i was wrong :DQuote:
Originally Posted by Hornet331
thats still a bit ahead... 10 core, 4 channel will come in q2 a far as i know. and yes, there are already plans for dual socket highend boards... seem like evgas sr2 has motivated intel to bring skulltrail back to live :DQuote:
Originally Posted by Johnny87au
i really dont see the point in a 20 core 40 thread 8 channel highend platform though when its hard to find a game that uses more than 3 threads and scales with more than 2 channels... actually most games barely scale when you go from single to dual channel, hah... :D
for crunchers who want plenty of cores sb will be interesting, same goes for servers... i just really really hope the dualcore mobile sb chips wont dissapoint... mobile clarkdales offer better performance than culv but you lose 1-2h battery life, which for me just isnt worth it...
10-core SB? i thought they'd just release an octo-core and westmere-EX (10-core)...Quote:
Originally Posted by saaya
aww hell no !
not going to upgrade this soon again.. stop making new stuff intel !
i heard its more than 8 and i dont think itll be 12, so...Quote:
Originally Posted by -Sweeper_
jesus christ, 10 core yet their still bringing out the 990x..Quote:
Originally Posted by saaya
what insult... :pQuote:
Originally Posted by saaya
only if you compare it to the 980x, but this baby plays in its own league. All the 45nm nehalems are quite in line on cpu bound benchmarks, when you start looking at vantage the table turn, thx to more pci-e lanes for SLi/CF.Quote:
Originally Posted by saaya
Look at the benchmarks i provided, there you have a i7 9xx and a i7 8xx clocked @ 4ghz and the perform tha same. It doesn't matter what what clock they run, stock or overclocked. If IPC is the same you performance is determined by max achivable clockrate and there S1336 has the edge compared to S1156.Quote:
Originally Posted by saaya
And no im no enjoying it to correct someone who is wrong...actually its a pain in the ass, cause I have to find data and I am a lazy fella... :yepp:
Can't decide, search for a better i7-860 CPU (atm running 4.0GHz 1.392v / 1.36v load HT off cuz of summer temp probs, max 77C at this config) or just wait til SB arrives and upgrade? Money isn't a problem atm so I don't care if I wouldn't gain that much. Would be nice to know how well SB overclocks. ^^
I'm not interested in the 8(+) core versions myself, I can't seem myself be in need of more than 4 cores for quite a while.
Bring on the cores!Quote:
Originally Posted by saaya
But really, how does Intel plan to fit 10 cores on a desktop (non EX) socket without a 22nm 'tock'? 6 Westmere cores already span 32nm, SB will be 32nm for a while until they shrink down to 22nm.
Are SB cores smaller or arranged differently?
yeah, nehalem cores measure 24mm˛, while SB cores shall be more like ~20mm˛Quote:
Originally Posted by trn
the desktop SB quad-core + IGP will end up with 65w TDP
an octo-core and even a 10-core processor without the graphics part is totally possible...
Definately low end cpu's if thats the case :up:Quote:
Originally Posted by -Sweeper_
Only changes from i7 to sandybridge are listed below..
Sandy Bridge DT (LGA 1155) will be a direct replacement for Core i5/i7 LGA 1156, but I doubt whether the owner of the latter would bother too much with the upgrade since the only notable differences are :
(1) a few new instruction sets, the most notable being AVX, that will still require software to be written to take advantage of them-probably around the two-year timeframe for AVX usage to become anything more than a novelty
(2) native SATA 6Gb support, as opposed to a third party onboard controller chip.
(3) Upgraded DMI (4 lanes PCI-E v2.0 for "Southbridge" duties)
The new chipset retains the dual channel DDR3-1333 (official support) and sixteen lanes of PCI-E 2.0 that the P55 enjoys.
So, LGA 1366/X58 will remain the enthusiast setup of choice* until Sandy Bridge B2 (LGA 2011) launches in late 2011.
* Depending upon whether Bulldozer mounts any worthwhile challenge, assuming it launches before LGA 2011....not a given at this stage.
Would really love to see how their native 6gb support scales to the gimmicky current third party onboard controllers?
Q3, not late 2011.Quote:
Originally Posted by Johnny87au
whats bull*hit? What i know, EX Sandy Bridge for highend s1355 and 8c/16t, not more
No, you are completely wrong, Johnny87au is right. Next enthusiast platform will be LGA 2011, till will be High-end X58+Core i7Quote:
Originally Posted by FlanK3r
I thought 2011 is only for server segment...? (s1155 mainstream and lower highend, s1355 new highend, 2011 server). 2011 pins will too crazy and big chip (how large L3 cache than?)
yet it will be huge chip, but quad-channel, new logics, 8+ cores, 32 pci-e bridge needs more pins and bigger silicon then 1366 now.
Your talking about Sand Bridge EX aka replacement for nehalem EX... thats no enthuisast platform but rather a replacement for 4P+ server. :p:Quote:
Originally Posted by Zed_X
he asked for EX
Ehhh.. not really S1336 will be replaced by Socket S1356 , Wikipedia is wrong in this regard, since it makes no sense to have such a huge socket for normal uniprocessor server and highend desktop.Quote:
Originally Posted by Zed_X
S1156-> S1155
S1336-> S1356
S1567-> S2011
Three segements three sockets. ;)
Why should intel suddenly change a scheme that worked for the last decade.
hornet i strongly disagree with you here, I highly doubt the 1366's >1356.. 1356 is the new sandybridge socket which will be high end , say 6/8/10 cores.. not sure but im assuming 6-8 cores..
i assume 1356 same size heatsink hole as 1366????
would 1356 still tripple ch. ddr3 or it would go to quad ch. ddr3????
Not sure mate, definately gonna be 8core/12threads though.. Cant wait :DQuote:
Originally Posted by PatRaceTin
what about L3 size chache for LG1356 ?
http://img80.imageshack.us/img80/744...ybridge511.jpg
LGA1356 will probably harbor salvage parts from the full-scale LGA2011 (quad channel) Sandy Bridge processors.
Uhhh... we where talking about the new enthusiast platform.. and if you belive intel makes the new MP platform the new enthusiast platform then you couldn't be more wrong. Mother board prices would be insane due to strict desgin requirements, ocing will be next to impossible...Quote:
Originally Posted by Johnny87au
The rumored S1356 has the same target field as the current S1336 -> Singel or Dual socket boards for highend desktop and workstations/lowend servers.
S2011 is the MP platform for 4 Socket+ with the chance for some 2S boards.
It makes no sense that intel extends MP platform down to singel socket workstations...
Im not exactly sure what you mean by MP platform... As far as my knowledge lets me go i know their is alot of changes which in no way targets the same idea of the 1336.. As for expensive mobo's yes unfortunately people will have to upgrade mobo's but CPU's will be fine..
MP = Multi Processor -> 4sockets+Quote:
Originally Posted by Johnny87au
DP = Dual Processor -> 2 sockets
UP =Uni Processor -> one socket
You should know this when you talk about this kind of stuff, since is basic knowledge.
Hell, Intel even uses this as there official nomenclature for there cpus...
Enthusiasts wont get S2011 cause its the MP platform aka Sandybridge EX, why do you want stuff like hardware RAS and other reliability related features that come with a hefty price tag for consumers... :confused:
i cant understand chinese lang., but here is some new video about SB
Sandybridge OC?
Thanks for the link. Here are some slides from the video...Quote:
Originally Posted by FlanK3r