Am I reading this right that they are planning on somehow limiting the bclock through hardware? If so that's a great way to give AMD marketshare.
Printable View
Am I reading this right that they are planning on somehow limiting the bclock through hardware? If so that's a great way to give AMD marketshare.
Oh.. seems we get a new brand of intel cpu, someting inbetween the normal and the XE moddels. Wonder at what price they will come.
For mainstream it seems so, according to the slides. When i first looked at slides before even noticing the fine print, I disliked the integrated clockgenartor solution. But since there will be fully and partially unlocked cpu (none XE) we'll have to wait and see how this turns out. If they unlocked cpus are only a few bug more expensive then the locked ones -> no big deal.
For not so expirienced users this is even better, since they only have to mess around with one or two variables (multi and core voltage).
:toast:
about everything else... i guess your right... i have no idea why i thought 1156 had a lower ipc, must have been cause i only played with that platform pre-release and the bios was far from final... the gmch perf was a nightmare back then, memory performance was t3h sux which strangled the cpus in every benchmark and caused a 5-10% hit in 3dmark and more in cinebench...
ovrclocking? on sandybridge? :lol:
you mean you guys havent heard?
there is no overclocking on sb, forget about it...
they use a single clock for bclock and dmi, and you all know how well dmi overclocks...
funny how intel blames PEG ie pciE for it when they KNOW its THEIR interface or most likely just the protocol that cant handle higher clocks.
its even more ironic that they arent able to fix this, shouldnt be that hard at all, x58 can reach pciE clocks in the 130 range and they didnt even tweak it... anyways... even with tweaks it would be pretty limited overclocking, right now we are at 133 and can go up to what, 233+? that kind of overclocking is impossible with dmi even with serious tweaks...
from what ive heard the highest clocks are around 105dmi on air and 112 on ln2... since the ref clock is 100 by default thats a 5% and 12% overclock respectively. so yes, the only way intel CAN sell that platform is with cheap cpu multiplier unlocked chips.... and i think their K cpus for 1156 were just a test to see how the market responds and to check how they should price their multiplier unlocked 1155 cpus...
i heard thats what intel is telling their partners though... they broke overclocking and thought it wouldnt matter, now they are backpedaling and trying to implement a divider for dmi/pciE so bclock overclocking is possible again. but its too late to do that for 1155 and might even be too late to do it for 1356. either way, intel is definately suggesting mobo partners to build an enthusiast platform based on 2011(?) from what i heard... like i said before, it seems that SR2 has made them want to resurrect skulltrail...
cause they broke overclocking and you cant sell a an enthusiast platform based on a cpu that... cant overlock... lol :D
and the 8+core cpu i heard of was for 2011(?) with quad channel, not 1356... there might be 8+core cpus on 1356 as well, but im not too sure about that... then again, a 6core 12 thread cpu doesnt even make sense and that didnt stop intel and amd from launching them either... even if they cant really make a lot of money off of them...
PS: when they say 57 bins in the slide... does anybody know what they mean?
cause it almost sounds like the max cpu multiplier is 57x?
57x100=5700mhz... thats not good at all...
think, 57x is CPU multiplier too...Im curious for more infos.
Wow are you dead serious? How on earth does intel expect to sell their new hot "sandybridge" like hot cakes? Intel have taken away the only thing i enjoy and love fiddling around with.. Wow intel have really f***** up.. Even if their new sandybridge cpu range is 8 core it doesnt change anything, Hopefully they fix the way these are manufactured! if not ill be sticking with a 990x or going to the red side n getting bulldozer!..:mad::mad::mad::mad::mad::shakes:
Remember that in the post-conroe, pre-nehalem days, it was long rumored that nehalem would not be overclockable, etc. and that intel had "locked the platform down". Not the first time we have heard this.....
If I had a penny each time I hear the next gen Intel CPUs won't allow OC, I would surpass Bill Gates. People are gullible that's for sure...
Wasn't the same thing about axed overclocking being waved since the i915X chipset -- the one with the locked PLL, or something?
Well, at least now we have some sort of official confirmation, not just groundless rumor... :shrug:
and initially that was true... on x58 they originally wanted the IOH to buffer the clock and distribute it, so they could get rid of an external pll, or well, at least simplify the design to cut the cost down a few bugs. turns out that it limited overclocking to around 140bclock, which is what you still see on x58 based laptops (yes they exist) and on server boards. cause they use the buffered clock mode... intel wasnt sure if the buffered mode causes some issues, so they had a backup plan to distribute clocks throughout the platform in the traditional way, and thats what intel went for in the end, which is what allowed bclocks in excess of 140.
on p55 i guess it was about the same, but intel had their test runs and the buffered mode DID work, so while they still supported the legacy way of clocking and hence overclocking, they didnt implement this on sandybridge. you guys have to understand that the guys who make these decisions at intel work closely with intels largest customers, ie the oems, and they couldnt care less about overclocking and rather want to save 1-2 us$ per platform and simplify the design... so thats what intel did... and their backup plan was to release unlocked cpus, though i think that wasnt originally they backup plan but an emergency decision when they got feedback from mainboard makers that they wont be able to position their products properly without any overclocking.
its kinda ironic, its the good old agp/pci lock all over again... remember when there were no dividers and we couldnt overclock the fsb past certain speeds cause the pci and agp clocks went too high? same thing really... annoying to see it all happening again :D
hah, yeah, what a joke... their oc lock was to reduce the chipset voltage so much that it would safely run at stock speed but not much higher... :rolleyes:
how would they fix this issue? the only way is to redesign the cpu and implement a separate divider for dmi that keeps dmi at 100 all the time no matter what clock you feed the cpu. i find it hard to believe that intel will manage to add this within 3-4 months. and from what ive heard they dont even plan to do it... as most customers, in intels perception, dont care and dont want to pay for the added logic and transistors.
but sure, keep your thumbs pressed, intel might add that divider after all, and it might even work... either way, its not like its going to turn sb into an awesome platform... besides the igp and avx there isnt really anything new, and the fact that 1366 continues to be the highend platform for a while after sb launches really says it all, doesnt it? :P
well, if intel does a good job with the turbo modes, i might get a snadybridge laptop after and it wont matter that it doesnt allow overclocking :D
Wow intel just killed mainstream SB if that's true about no overclock. And lol, 3 brand new sockets.
If it were any consolation...
1. Nehalem on 1156 / 3.8 will probably hold very well against stock SBs albeit at higher power costs
2. There has been K models (rather pointless) on 1156. Test run for 1155 K models? If BCLK really is latched to DMI, look for them coming. They won't be cheap though.
Actually there will be 2 (3) models for mainstream SB, fully unlocked (up to x57 multi), partially unlocked and (I am not sure about his) standard models.
If the (none XE) unlocked models come for the same price as the current i5/i7-K models its ~30€ more then the same locked model.
It all depends on the price and this is still ab bit away.
But the one thing that changes, is that ocing for mainstream SB will get a lot easier (and some see it as less chellenging). In the end it doesn't really matter if you oc by changing the multiplicator or the bclck.
There are some parts that don't make sense if you apply a bit logic. Why should the S1356 platform should be broken, if it is using the classic approach (external clock generator) with adjustable PCIE/DMI multis. Also for me it still don't makes sense that intel want's to use S2011 as enthusiast platform, it would be the same as intel would recommend to use S1567 as enthusiast platform... hell not even skulltrail used the MP platform, it still was the DP platform. (just if someone does not remember, the MP platform back then was S604 and skulltrail as S771 the DP platform)
We already have pretty much a good impression what the different platforms get in regards to core count:
S1155 -> up to 4 cores
S1356 -> up to 8 cores
S2011 -> up to 10 cores (maybe more)
So why are you implying that S1356 will get more then 8 cores?
Personally i wouldn't mind seeing a S2011 enthusiast platform, but i doubt this will happen. This would be the first time consumers would get enterprise grade hardware, but i doubt that somehow...
Bulldozer here we come, @ saaya im not trusting what you say as we've herd it before... SB isnt released yet so u dont know what intel can do..
I guess Intel has learned a lesson from i7, when we all bought a sub $300 920 and OCed it to beat the $1000 975 for 24/7 use. It was difficult to sell high-end CPUs then.
It is natural if Intel try to distinguish the OC-ability between mainstream and high-end to prevent a direct competition. But the SB is not out yet, and it remains to be seen how the final product will come out.
but some guys are now testing ES SB. Example Coolaler.
These early ES may change a lot, as my X5650 A0-ES with x19 multi and 20 turbo has changed to x20 multi and x22 turbo later. So we should be careful to speculate too much based on early ESes.
There will be OCing on SB mainstream models... It probably won't break any OC records but OC will be possible.
so will ram be overclockable at all?
Considering it can use DDR3 2666+ what do you think ?
http://www.youtube.com/watch?v=OkEgqpf_xRM
2:07 into the video states : 3)Limited base clock margining(sic) Vs. previous generations .
How much limited is the question.
edit:
2:53 into the video explains the limitation : 2-3% overclock only (DMICLK ) due to processor DMI/PEG PLL limitation
still we will have Non-XE unlocked parts though,so DMICLK limitation will be circumvented.
anyway im sure that mobo makes like evga will figure something out .... dont dispare people ... there will always be oc'ing
So, sandy bridge is a new socket correct? And these ones in the OP are the low end ones? So are there high end ones with Hex/Octa cores coming out for the same socket next year that will take over the high end from 1366?
i cant keep up with all the platforms and numbers and lingo
i figured they would provide different settings for the ram and you just choose what yours is, but that does not really mean you can OC, i can get 1600mhz stuff and set it to 2133 maybe, but could i ever go past that?
From the slides it looks like b-clock overclocking is out, or at least limited. But it looks like even normal chips might have a little more (but limited) multiplier freedom then before to make up for it.
It's not really ideal for budget overclockers, but I can understand why Intel would do it that way from a business perspective.
Who buys chips with integrated graphics for OC ? Having the graphics on die ( more integrated that everything else on the market ) means some limitations to the OC capability since you're affecting directly the graphics chip. Obviously, as long as you know what you are doing or the GPU is disabled you should still be able to OC.
The performance version with 6 and 8 core without graphics should not be any different in OC than current Nehalems.
From the looks of it, overclocking of mainstream Intel chips will be canned, unless you pay more. Greedy Intel!
Greater share for AMD then!
yeah, it is possible intel feels their brand name is more superior than simple benchmarking and overclocking, then all they want is to get the most money per buyer, knowing that a percentage of their market share will not leave them.
so, still i dont know, what coming as highend SB, LGA1356 or LGA2011 (last time this packet was only for servers)
For the lower price segment/bracket, AMD is the only choice left available for overclocking. Unless AMD follows Intel and break overclocking?
Sry but what dodge?
Of you buy ram that is specificed for 1600mhz and your running it with 1800mhz your ocing it. It doesn't matter if its due to increasing the base frquency or the multiplier... and for the other question, sure when its possible to raise the bclck (even only a few mhz) you can get past the 2133mhz.
it'll be interesting to see if anyone could/would whip up a bios that could bypass the bclk limit. the limit is purely artificial. i could see it making sense when comparing bloomfield v. lynnfield so i think the limit could be very restrictive.
Well according to the slides, there is no limit, but the whole system will get instable. Just look how fast S1336 systems get instable when you increase pcie frequency above a certain point. sata controlelr crapping out comes to mind and other stuff.
It might be possible that the clock generator it self in cougar point is capable of much more then 2-5%, but if everything else craps out there isn't much point to it.
This is a new architecture, and we should be careful to speculate too much based on current architecture.
As I've written in another tread, it was many similar speculations before the release Nehalem too . Some where speculating on problems with OCing RAM on Nehalem, and issues with using 2v+ RAM on new system. All was a part of new and better 3-channel architecture, but those who didn't understand it at that time were making negative scores on it.
its not a bios limit, so how could a bios fix it?
unless its possible to flash new firmware onto the PCH via a bios update and the dmi protocoll really IS the limitation AND can be updated... i wouldnt hold my breath... even if that would work, youd get how high? 120mhz? 130mhz maybe? thats a 30% overclock then, not bad, but compare that to 1156 and 1366 bclock overclocking...
dont compare it to 1366, compare it to 1156... whats the max pciE clock youve seen on 1156? 1366 actually did "ok" pciE clock wise...
anyways, i actually havent heard anybody mention 1356 or 2011, all i heard was that sb bclock ocing is broken as its linked to dmi, and the current plan of action is to fix that on the enthusiast platform that comes out later, a while later, and thats gonna be quad channel and 8+ cores... those are tidbits from several sources and most of it overlapped... it was me who assumed that platform is 2011, but 1356 def isnt quad channel and unlikely to have 8+ cores, so...
i havent actually heard anybody mention 1356 in a long time... maybe its cause its coming out later, but why would it come out after socket 2011? and why would intel focus on 2011 for highend and not 1356? i have no idea...
Both, socket R and socket B2 are scheduled for H2 2011. And how can DMI be broken it works as intended. Overclocking is running thing out of specification. It can work but there is no obligation from intel that it should work outside there spesifications. You can call it broken if it doesn't even wirk at there own specifications.
And again I still doubt we see S2011 for consumers, but at least we'll see some news on next IDF.
true, its not broken... from intels prespective at least :D