Sandy Bridge may be Un-Overclockable

[VirtualLarry]

Has Intel ever officially stated any of their CPU's were overclockable?

They have produced motherboards with overclocking features.

 

[VirtualLarry]

Actually there's good info from that WoW demo. First look at the Anandtech results: http://www.anandtech.com/show/2901/4

The GMA HD is only 30-40% faster in WoW than the G45. With both system running WoW set to high, the 65nm-based GMA 4500 was getting roughly 3-4 fps. The GMA HD wouldn't do much to make it playable in this case. The Sandy Bridge graphics? It had perfectly playable fps which suggest at least 20 fps.

I remember playing WoW on my A64 3800+ single-core, and NV 6150 graphics, and according to the on-screen display, I was getting between 12 and 18 FPS, at whatever low res I was runnning in (I forget), and I sure didn't think that was smooth enough for regular gameplay. So I wouldn't say 20FPS is "perfectly playable".
[/QUOTE]

 

[ShawnD1]

A GMA950 can run WoW playably... if you put settings all low. Did you see the settings on the video?

Apparently it sucks compared to a GeForce 9400m, which is basically the standard graphics found in a midrange Intel laptop.

http://www.notebookcheck.net/Mobile-Graphics-Cards-Benchmark-List.844.0.html
9400m - rank 154
GMA950 - rank 235

The top Intel graphics on that list is the GMA HD at rank 171. To put it into perspective of how far behind Intel is, I want to link you to the cheapest AMD laptop on Best Buy's website: link. This $350 computer comes with Radeon 4250 graphics. According to that list of integrated laptop graphics, the Radeon 4250 (rank 169) is actually better than Intel's top graphics, the GMA HD (rank 171).

So really, Intel's best graphics is worse than the cheapest piece of shit sold by Best Buy. Think about that for a minute. Any "revolutionary" step by Intel would simply be matching what AMD and Nvidia are already doing. Maybe they'll do something that has been totally impossible up until this time - make Intel graphics as fast as the GeForce 9400.

 

[Ben90]

You forget Intel's target consumer.

My grandma has one of the first Pentium 4's on XP with so much stuff going on in the background that it takes a good 5 minutes to boot up. She used to have 6Mb/s cable, but downgraded to dial up after she realized she didn't need the speed. My ex-girlfriends father uses a Pentium III and couldn't be happier. We have some 11" laptops at work using Penium IIs and they are absolutely amazing for what we need them to do.

You have to realize that most people don't really have an interest in playing Crysis or Starcraft II, or even flash games online. To them a GeForce 9400m is just a waste of money.

 

[ShawnD1]

You have to realize that most people don't really have an interest in playing Crysis or Starcraft II, or even flash games online. To them a GeForce 9400m is just a waste of money.

As you implied in your first paragraph, old people are the ones who need the overclocked i7 system more than anyone on Anandtech does. One of my bosses refused to run multiple programs, so he would close Firefox (I assume his daughter installed it) before opening MS Word, and he would close Word when going back to Firefox. That's the guy who needs top of the line hardware just to get an acceptable computing experience because he's using his hard drive to do what most of us do with RAM. My mom does the exact opposite - she'll run 20 Firefox tabs at the same time and never close anything; some of those tabs are things like Map Quest looking up a location she went to one time a few months ago.

Going back to the car analogy, it's usually people who don't understand cars that are the hardest on them. A 40 year old car mechanic knows how to drive a standard. A retarded 16 year old who doesn't quite understand what a clutch does will burn the shit out of it and need to replace the clutch every year. Similarly, old people who don't know how to use a computer are the ones who need 6 cores and solid state hard drives to get the basic computing experience because they run 3 antivirus programs at the same time, the computer has 20 different programs that launch on startup, and the person still uses "power off" instead of "sleep" when they are done with the computer, which forces it to reload all of those programs from the hard drive.

I don't know how an old person could need a 9400m video card, but I'm sure they'll find a way. I didn't think anyone would insist on closing Firefox before opening Excel since we've had multitasking for the past 20 years, but I still found that person. Somewhere out there is someone who tries to run Crysis, while doing a GPU video test, while doing GPU folding, and is wondering why everything is so jittery.

 

[aigomorla]

im so tired of 1155... seriously...

Im at the point where now im even wondering who is telling the truth and who is just repeating bullshit.

Meh im seriously gonna start overclocking netboxes, and go on the enterprise machines for everything else.

 

[heyheybooboo]

His post said 'The good new is that the title of this thread is totally silly and wrong too'. The title is 'Intel plans to deliberately limit Sandy Bridge overclocking'. Now, Sandybridge might have limited overclocking because of a cost cutting design that tries to combine everything down to one clock signal, which results in CPUs that can't be overclocked as much while at the same time not being intentionally designed to kill overclocking.

The guy's statement could be entirely true (that Intel isn't intentionally trying to kill overclocking), while at the same time not really be denying that Sandybridge might have more limited overclocking compared to what people are used to.

I haven't read the entire thread so don't know if someone brought this up, but ...

Screwing with the clock signal could generate 'skew & jitter' which impacts the PLL timing.

It doesn't mean it can't be done, just makes it very difficult to get much of a bump. To really crank it would require extra voltage which would greatly increase the likelihood of frying something on the die ...





--

 

[Ben90]

I think the main problem booboo are the buses not tied to the processor e.g. SATA/USB. These are very sensitive to change and, if this news is true, are locked to the Bclock. The processor will have no problems overclocking using the old methods, but other sensitive things will not be able to handle the changes.

Asynchronous clocks are nothing new and if the motherboard makers decide there is a market for a more fine grained overclocking solution than just raising multipliers, they will implement it.

Personally I think Intels K-series will offer good enough overclocking potential for most people using the 1155 socket. Remember this isn't the enthusiast socket for a reason, if you really need that much performance you will be looking at 2011.

 

[DrMrLordX]

im so tired of 1155... seriously...

Im at the point where now im even wondering who is telling the truth and who is just repeating bullshit.

Meh im seriously gonna start overclocking netboxes, and go on the enterprise machines for everything else.

No no, overclock Magny-Cours systems with K10stat! It's all the rage.

 

[ilkhan]

Personally I think Intels K-series will offer good enough overclocking potential for most people using the 1155 socket. Remember this isn't the enthusiast socket for a reason, if you really need that much performance you will be looking at 2011.

Thats my thoughts as well.

Even if this is true (likely is, to my mind), those same slides suggest a basic "you can increase the turbo multi's within certain limits" and a k series "you can increase the base multi (and turbo goes beyond that, if on) without limit".

 

[aigomorla]

Dude if we only have multi overclocking that is gonna blow.

Its like putting an automatic in a car when you want to drive a manual.

We really need that bclk.

Otherwise its gonna blow.. everyone will have ballpark the same OC, because its multi, and wont be able to get that 10mhz more for epeenis rights.

Man... whats the fun in that?

 

[khon]

As with most "leaked" information there is obviously a very good chance that it's false, but I still can't help but wonder. We discussed the Sandy Bridge model numbers earlier (http://forums.anandtech.com/showthread.php?t=2089138), and none of them seem to be partially or fully unlocked (which I'm basing entirely on the lack of letters at the end of the name, i.e no k or x etc.)

So what are the chances that none of the processors in the initial launch are overclockable ? Could the unlocked version come out later ?

 

[Idontcare]

Dude if we only have multi overclocking that is gonna blow.

Its like putting an automatic in a car when you want to drive a manual.

We really need that bclk.

Otherwise its gonna blow.. everyone will have ballpark the same OC, because its multi, and wont be able to get that 10mhz more for epeenis rights.

Man... whats the fun in that?

Sounds like NASCAR...thats what you get whenever a cool hobby sport goes corporate.

 

[DrMrLordX]

We really need that bclk.

Agreed. Otherwise, how are you going to overclock the uncore? And memory overclocking, oh dear.

The thing I hate most about this is that once BCLK-based overclocking is mostly dead (well, on everything but LGA2011), you'll see memory kits redone to reflect this fact except for the small number of kits that are released for AMD machines or that will be released for LGA2011.

If people thought Clarkdale was a PITA to overclock, this is going to be worse.

 

[Ben90]

Agreed. Otherwise, how are you going to overclock the uncore? And memory overclocking, oh dear.

The thing I hate most about this is that once BCLK-based overclocking is mostly dead (well, on everything but LGA2011), you'll see memory kits redone to reflect this fact except for the small number of kits that are released for AMD machines or that will be released for LGA2011.

I do concur that its kinda lame, but I really want to test out how fine grained we can actually get it. I'm going to pick an arbitrary number that our components max out at to see if its really going to hold us back that much:
Core 4874mhz
Uncore 3521
RAM 987
Bclock 103 (I'm under the impression that 1155 is using a base Bclock of 100, correct me if I'm wrong)

The most important thing is almost always core speed so lets use:
101x48 = 4848mhz Core
101x34 = 3434mhz Uncore
101x9 = 909mhz RAM

Vs our current overclocking scheme of 133-230bclock
168x29 = 4872mhz Core
168x20 = 3360mhz Uncore
168x5 = 840mhz RAM


Now you could probably tweak the 133-230 bclock numbers more at the cost of core speed, but since I'm lazy and mostly wanted to compare core speeds this example will do.
We lose a whopping 24Mhz on the Core clock, but make up for this in Uncore/RAM speeds because of the lower Bclock. The 100 bclock basically saves the day as it allows more fine grained tuning to begin with even though we cannot adjust it all that much.

Also, now that Ive done the math I realize that the 4874 number (core speed wise) was stacked in the second setups favor. If I chose something like 4852 or 4891, basically every number would wind up in the new systems favor.

It also took me all of 5 seconds to find the optimal bclock for the 4874 number due to the limited number of bclocks availible, while I sat in the calculator for a couple minutes trying to find the closest bclock for the 168 system.

This is not the end of the world.

 

[A_Dying_Wren]

I do concur that its kinda lame, but I really want to test out how fine grained we can actually get it. I'm going to pick an arbitrary number that our components max out at to see if its really going to hold us back that much:
Core 4874mhz
Uncore 3521
RAM 987
Bclock 103 (I'm under the impression that 1155 is using a base Bclock of 100, correct me if I'm wrong)

The most important thing is almost always core speed so lets use:
101x48 = 4848mhz Core
101x34 = 3434mhz Uncore
101x9 = 909mhz RAM

Vs our current overclocking scheme of 133-230bclock
168x29 = 4872mhz Core
168x20 = 3360mhz Uncore
168x5 = 840mhz RAM


Now you could probably tweak the 133-230 bclock numbers more at the cost of core speed, but since I'm lazy and mostly wanted to compare core speeds this example will do.
We lose a whopping 24Mhz on the Core clock, but make up for this in Uncore/RAM speeds because of the lower Bclock. The 100 bclock basically saves the day as it allows more fine grained tuning to begin with even though we cannot adjust it all that much.

Also, now that Ive done the math I realize that the 4874 number (core speed wise) was stacked in the second setups favor. If I chose something like 4852 or 4891, basically every number would wind up in the new systems favor.

It also took me all of 5 seconds to find the optimal bclock for the 4874 number due to the limited number of bclocks availible, while I sat in the calculator for a couple minutes trying to find the closest bclock for the 168 system.

This is not the end of the world.

It wouldn't be inconceivable however for Intel to limit the multipliers on 1155 so I wouldn't bet too much on there ultimately being little difference.

 

[DrMrLordX]

It wouldn't be inconceivable however for Intel to limit the multipliers on 1155 so I wouldn't bet too much on there ultimately being little difference.

You can bet on that. A CPU with a $40 premium for an unlocked CPU multiplier may still not have all the memory multipliers that you might want.

And really, my point was that LGA1155's limitations would affect the entire consumer memory market, whether you were an LGA1155 owner or not.

 

[Cattykit]

As you implied in your first paragraph, old people are the ones who need the overclocked i7 system more than anyone on Anandtech does. One of my bosses refused to run multiple programs, so he would close Firefox (I assume his daughter installed it) before opening MS Word, and he would close Word when going back to Firefox. That's the guy who needs top of the line hardware just to get an acceptable computing experience because he's using his hard drive to do what most of us do with RAM. My mom does the exact opposite - she'll run 20 Firefox tabs at the same time and never close anything; some of those tabs are things like Map Quest looking up a location she went to one time a few months ago.

Going back to the car analogy, it's usually people who don't understand cars that are the hardest on them. A 40 year old car mechanic knows how to drive a standard. A retarded 16 year old who doesn't quite understand what a clutch does will burn the shit out of it and need to replace the clutch every year. Similarly, old people who don't know how to use a computer are the ones who need 6 cores and solid state hard drives to get the basic computing experience because they run 3 antivirus programs at the same time, the computer has 20 different programs that launch on startup, and the person still uses "power off" instead of "sleep" when they are done with the computer, which forces it to reload all of those programs from the hard drive.

I don't know how an old person could need a 9400m video card, but I'm sure they'll find a way. I didn't think anyone would insist on closing Firefox before opening Excel since we've had multitasking for the past 20 years, but I still found that person. Somewhere out there is someone who tries to run Crysis, while doing a GPU video test, while doing GPU folding, and is wondering why everything is so jittery.

100% agreed. The beauty of modern computing is that hardware has much more power than typical softwares every request. Like you said, unlike before, average people can still have so many crapwares running in the background yet the system handles 'em fine.

 

[Ben90]

It wouldn't be inconceivable however for Intel to limit the multipliers on 1155 so I wouldn't bet too much on there ultimately being little difference.

Read OP.

It is pretty obvious that Intel understands there is a market to overclocking, note how long they have had SKUs with unlocked multipliers, and how its mentioned in official Intel slides. I really don't understand all this freaking out about having an locked bclock, It only really affects the LN2 overclockers, and even then they will be using LGA2011 anyways.

If Intel has lied on their official stance and you cannot overclock 1% due to the external busses going corrupt mobo makers will add an external clock controller to sort shit out if they see a market for it. Its really not that hard to do, you need to add a whopping 2 ICs to your mobo design and you can ramp your bclock up to as high as your processor/mobo will allow. One in front of the Northbridge, and one on the PCI-e lanes. Problem solved.

 

[Nemesis 1]

What do you mean by that (serious question, I couldn't figure out what your post meant, can you reword it)?

Anyway so the questions really that haven't been answered (assuming dr who isn't just completely full of crap, which he might be, who knows):
1. How much can you raise the multiplier (or just the turbo multiplier maybe?) on the "locked" processors. Can this multiplier be used all the time on all cores?
2. What is the price premium on the unlocked processors? (ie, $25 sure, $75 maybe, $150 no thanks, people will buy AMD or save up for gulftowns or buy an 1156 or 1366 platform). Intel employs some really friggin smart people, so I think it's doubtful that they will cost more than $100 more than the locked parts--that would kill sales. My guess is $50-$75 more. What do you all think?

And the main question (I mean lets face it, it's not like anyone is expecting Sandy Bridge, at least the 1155 version, to outperform a 980x or whatever, so the million dollar question is...):

3. How good is the GPU??? NO ONE is talking about this, no numbers are out, all we got is a crappy little WoW demo that didn't show anything. So, how good is it? If Intel can bring, say, Radeon 55xx performance, this will be a mind blowing processor. That's unlikely, but COULD happen. Disclaimer here is that that performance would need to actually come with equivalent image quality--for example, the IGP on i3/i5 stuff can sometimes beat a 5450 performance wise, but that's only because the image quality is rofl night/day horribad (try sometime you'll see what I mean lol). So, how good is this GPU? What all does it support? DX10, right? Intel won't surprise us with DX11 will they? Will the GPU multiplier be unlocked on some or all SB cpus? I'm dying for some more info on this. (This is also why AMD's bobcat excites me so much).

Gee can't you wait for the release instead of all this mindless rambling. I should start a thread AMD BD won't overclock . Real classy threads around the forums as of late,
As for intel and DX11 intel doesn't have to support it . Intel has said many times they don't care what MS does as they can run anything . The HD version already uses DX11 like paths.

As for no info on the SB GPU there is good reason its being info blocked. Intel HD IGP was much better than any people here admitted . Fact is many still deny its a good little IGP.
But can it play crisis . Thats a question that speaks to the mindset of the AMD/NV groupes

Enough with the mindless ranting, please.
-ViRGE

 

[aigomorla]

dude ive been totally patient with this thread, and now i have offically have had.

Its funny how guys will ignore people who got samples.
And are trying to pass more stuff then such people who got samples.

Unless intel seriously fubared the line up, and gave out all CRAP samples.. (highly doubt)


Anyhow im seriously done with 1155.

If you feel you are the same type of person i am, who enjoys tuning system, i have said this 10000 times, and it will be my last time.

SB 1155 IS NOT FOR YOU.

If you dont care about overclocking.. then SB1155 will be the fastest stock system you can get.
Hands down...

But An Overclocked C2Q even would kill a stock SB any day of the week with its eyes closed.

 

[extra]

Aigo can you confirm or deny if there are socket 2011 systems out in the hands of enthusiasts yet?

 

[aigomorla]

Aigo can you confirm or deny if there are socket 2011 systems out in the hands of enthusiasts yet?

I bet there is.

But no one is talking.

2011 should break all the current benchmarks minus SR-2.

The SR-2 benchmarks will probably rank crown for another year.

 

[cbn]

dude ive been totally patient with this thread, and now i have offically have had.

Its funny how guys will ignore people who got samples.
And are trying to pass more stuff then such people who got samples.

Unless intel seriously fubared the line up, and gave out all CRAP samples.. (highly doubt)


Anyhow im seriously done with 1155.

If you feel you are the same type of person i am, who enjoys tuning system, i have said this 10000 times, and it will be my last time.

SB 1155 IS NOT FOR YOU.

If you dont care about overclocking.. then SB1155 will be the fastest stock system you can get.
Hands down...

But An Overclocked C2Q even would kill a stock SB any day of the week with its eyes closed.

You are making me wonder if SB could be nothing more than a quad core Clarkdale with a doubled up IGP.

 

[extra]

You are making me wonder if SB could be nothing more than a quad core Clarkdale with a doubled up IGP.

Well there will be dual cores and quads on 1155, that's been announced. It will also have AVX and other tweaks.

I'm willing to bet that SB is more than "doubled up" on the IGP side of things though. Gotta remember, yeah, Intel's graphics history might be...how do we say this...underwhelming...but they do employ some of the smartest people in the world. Go watch that sb graphics demo if you haven't--it did look like a *big* improvement, tho we didn't get any specifics. The current (the i3 one) intel graphics perform as well as the (albiet several years old now) amd integrated video, or better, but they do have image quality problems. I bet we get a 2.5-4x speed improvement over i3 integrated graphics, plus more effort put into Intel's graphics drivers. I think if SB's integrated graphics are good it has the potential to really re-vitalize the PC gaming market.

Think about it...if you're making a game with decent graphics, you're only able to target something like 25% (educated guess) of all new computers sold. Something like 75% (educated guess here) come with Intel GMA. That's a huge potential player base that can't buy and enjoy your product. If SB reaches the "good enough" stage...yeah.