- Joined
- Aug 3, 2004
- Messages
- 24,346
how is it a flop because someone cant overclock to speeds that %99.99999999999999 if people will never need any time soon...
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Haswell a big flop?
- Thread starter XenatR
- Start date
WhoBeDaPlaya
2[H]4U
- Joined
- Dec 16, 2002
- Messages
- 2,860
FTFY. I'd wager a lot of us don't upgrade to the latest and greatest because of "need"how is it a flop because someone cant overclock to speeds that %99.99999999999999 if people will never want any time soon...
- Joined
- Apr 15, 2005
- Messages
- 22,092
I have seen a downward trend from Intel offering mainstream parts that seem to offer less and less for the enthusiast.
SB came and most of them were stong OC'ers. 4.4+ GHz was almost a sure-fire thing, even with a budget HSF. Getting one that would do 4.8+ GHz at lower Vcore and low temps was the golden egg.
IB came and it seemed that 4.4+ GHz was doable with most, but an adequate investment in cooling was a necessity once that OC'ing "heat wall" was surpassed...Usually at 4.6+ GHz for a lot of them. Getting one that would do 4.6+ GHz at lower Vcore and low temps (comparatively) was the golden egg.
Now HWL is here and my gut feeling is that the golden eggs will be the ones that can do 4.4+ GHz at lower Vcore. Going to need premium cooling for this one when OC'ing, regardless of what speed a given unit will max out at. Hopefully I'm wrong, but a lot of reviews seem to imply otherwise.
If there was one thing I wish that Intel would have listened to all the reviewers (professional and user) biggest complaint from the dawn of IB and implemented in HSW, that would have been fluxless solder under the IHS for the K parts, if not the entire lineup. With the sheer success of all 1155 parts, they have no reason to skimp on something like that going forward to save a miniscule amount per unit...they've made their money from us, why not listen and give something back as a token of gratitude?
So, is this an intentional crippling of the mainstream parts in order to move more enthusiast LGA2011 parts at an undoubtedly higher profit margin?
SB came and most of them were stong OC'ers. 4.4+ GHz was almost a sure-fire thing, even with a budget HSF. Getting one that would do 4.8+ GHz at lower Vcore and low temps was the golden egg.
IB came and it seemed that 4.4+ GHz was doable with most, but an adequate investment in cooling was a necessity once that OC'ing "heat wall" was surpassed...Usually at 4.6+ GHz for a lot of them. Getting one that would do 4.6+ GHz at lower Vcore and low temps (comparatively) was the golden egg.
Now HWL is here and my gut feeling is that the golden eggs will be the ones that can do 4.4+ GHz at lower Vcore. Going to need premium cooling for this one when OC'ing, regardless of what speed a given unit will max out at. Hopefully I'm wrong, but a lot of reviews seem to imply otherwise.
If there was one thing I wish that Intel would have listened to all the reviewers (professional and user) biggest complaint from the dawn of IB and implemented in HSW, that would have been fluxless solder under the IHS for the K parts, if not the entire lineup. With the sheer success of all 1155 parts, they have no reason to skimp on something like that going forward to save a miniscule amount per unit...they've made their money from us, why not listen and give something back as a token of gratitude?
So, is this an intentional crippling of the mainstream parts in order to move more enthusiast LGA2011 parts at an undoubtedly higher profit margin?
Last edited:
Not even. Most Celeron 300A's did 450MHz without an issue. There were a few (including two I owned) that didn't quite hit 450MHz but the majority of them seemed to do it fine. The truth is we've never seen this much variance from Intel in terms of overclocking.
- Joined
- Aug 24, 2006
- Messages
- 2,418
K-Series processors are designed for enthusastic like us not for 99% mainstream users.
You will definitely need some beefy cooling to get anything substantial out of Haswell in terms of overclocking.
Gratitude? Not likely. Companies throw people a bone only when the think the PR is good and will provide a return on investment in the long run. Few people give a shit how overclockable a processor is or isn't. Enthusiast parts are sometimes leveraged as PR but they don't impact the bottom line the same way those budget and mid-range offerings do. Personally I don't think it's about crippling anything for the sake of moving enthusiasts anywhere as much as it is making the desktop and mobile segments as profitable as possible. I don't think the enthusiast was really part of the equation really. The enthusiast market is an after thought and not ever the driving force behind Intel's business decisions.
5 cents for 600,000 units would be $30,000. So cost savings using inferior thermal interface materials for the heat spreader add up quickly. The desktop market is becoming more of a niche every generation as we shift to more and more mobile options. Even laptops are starting to dwindle in terms of shipments as people migrate away from such units to tablets and smart phones. Intel went with an IVR to make these CPUs more efficient for mobile use. Plain and simple. They cut the cost as much as they could in order to maximize profits.
Greetings
I'm guessing that its down to simple laws of chemistry and physics because if you have a look at the 2700K it has a TDP of 95 Watts whereas the 3770K has a TDP of 77 Watts, furthermore if you read this article here it states that SB is 216mm^2 and IB is 160mm^2.
This means that SB puts out (95/216) = 0.4398 watts per mm^2 and IB puts out (77/160) = 0.4813 watts per mm^2, a copper heatsink attached to either CPU is fixed and determined by the conductivity of copper so if I understand Newton's law of cooling correctly the heat loss will be proportional to the temperature difference so given IB will put out more heat the temperature will have to rise until the temperature difference gets large enough so that the extra heat can be taken away. The end result is the CPU runs hotter and if you overclock you get closer to the thermal limit where the CPU throttles down, also I believe that if you overvolt the power consumed is the square of the voltage increase. Long term if you have the die shrinking faster than the power consumed (indicated by the TDP figure) you will probably see your ability to overclock will become less and less.
This is where the law of economics now comes into play. So for 99.99% of people who don't overclock or read this forum;
(a) consumers see a CPU that is marginally better than the previous one and uses 18 watts less power so definately a good thing from their perspective.
(b) manufacturers see that a smaller (read cheaper) power supply and CPU heatsink can now be used, also a good thing from their perspective.
(c) Intel who presumably spend a couple of thousand bucks (I'm guessing) on the wafer and manufacturing process can get what 300? SB chips off the wafer, alternatively they can get what 400? IB chips instead off the same wafer, since they can sell them for the same price its pretty obvious they are going to junk SB production of go with IB chips to make more money. Another thing to remember is that yields drop if there are defects especially more so if the die is larger, the Pentium 4 Extreme Edition was three times the size of an equivalent ordinary Pentium 4 CPU due to the extra cache so it would be interesting to see what proportion of those ended up being bad.
So for the 0.01% of customers left that are not happy, well Intel do not care because firstly the CPU's are rated for their stock speed and as far as they are concerned "you got what you paid for" and secondly AMD's cpu's don't even come close to Intel's stock speeds let alone any overclocked ones, and even if they do their IPC count is lower, they chew more power and generate more heat so they don't even appear on my radar.
Given Haswell's TDP is going up from 77 watts to 84 Watts my gut feeling is that you'll probably struggle to even get that speed.
Would it have made a signifiant differrence, the first thing people did when IB came out was to de-lid the IHS off the IB and attach the heat sink directly and from what I understand they had better results but the increase was not as much as they had hoped for.
I think if you look back in the past you've never had it so good, you have their premium cpu (2700K/3770K) for only three hundred bucks and an almost equally capable cpu (2500K/3570K) for two hundred. I remember the good old days of a 3.0 Ghz P4 for AUD$500 and a 3.4 Ghz P4 for AUD$750.00, going back a bit further I remember 700 Mhz Celerons and 1.4 Ghz P4's having a ten fold difference in price being AUD$140 vs AUD$1400, and going back still further if you were the proud owner of a PC with an 80386 (or was it an 80286?) CPU I vaguely recall an ad in Byte magazine where you could buy an 80387 (80287?) math co-processor for ONLY $1600. Don't forget as well that money back then would have bought a lot more compared to today.
No I don't think so for the reasons stated above and also the following.
Firstly, Intel has problems running at high frequencies and this started with the end of the Northwood P4 3.40 Ghz cpu ten years ago!, the Prescott's only got up to 3.80 Ghz and then were replaced by the Core CPU's at much lower frequencies and fortunately they (a) had a higher IPC count and (b) came in dual and quad cores.
Secondly, most software/games etc in general usage at most only really uses a 4C/4T cpu like the 2500K/3570K, you don't get that much of a benefit increase going to a 4C/8T cpu like the 2600K/3770K much less a hexacore 6C/12T X58/X79 one either.
Thirdly, the base model cost of a S2011 I7-3820 cpu is comparable to the top of the line S1155 cpu for pretty much the same set of features and motherboard prices are reasonably similiar. Your argument would be valid if the 3820 didn't exist and you were forced to buy a $600 (or more) hexacore cpu as a minimum purchase.
Conclusion:
Unfortunately, the ability to overclock is declining as we move forward due to the variety of reasons I have stated above (and none that appear deliberate by Intel) and there is not much you can do about it, I think that generally people should be grateful for what overclockability they did get in the past and in the future all we can probably look forward to is getting effectively non-overclockable cpu's with maybe even more cores on them to compensate for this but we will have to wait and see if the software written for them keeps up.
Cheers
I'm guessing that its down to simple laws of chemistry and physics because if you have a look at the 2700K it has a TDP of 95 Watts whereas the 3770K has a TDP of 77 Watts, furthermore if you read this article here it states that SB is 216mm^2 and IB is 160mm^2.
This means that SB puts out (95/216) = 0.4398 watts per mm^2 and IB puts out (77/160) = 0.4813 watts per mm^2, a copper heatsink attached to either CPU is fixed and determined by the conductivity of copper so if I understand Newton's law of cooling correctly the heat loss will be proportional to the temperature difference so given IB will put out more heat the temperature will have to rise until the temperature difference gets large enough so that the extra heat can be taken away. The end result is the CPU runs hotter and if you overclock you get closer to the thermal limit where the CPU throttles down, also I believe that if you overvolt the power consumed is the square of the voltage increase. Long term if you have the die shrinking faster than the power consumed (indicated by the TDP figure) you will probably see your ability to overclock will become less and less.
This is where the law of economics now comes into play. So for 99.99% of people who don't overclock or read this forum;
(a) consumers see a CPU that is marginally better than the previous one and uses 18 watts less power so definately a good thing from their perspective.
(b) manufacturers see that a smaller (read cheaper) power supply and CPU heatsink can now be used, also a good thing from their perspective.
(c) Intel who presumably spend a couple of thousand bucks (I'm guessing) on the wafer and manufacturing process can get what 300? SB chips off the wafer, alternatively they can get what 400? IB chips instead off the same wafer, since they can sell them for the same price its pretty obvious they are going to junk SB production of go with IB chips to make more money. Another thing to remember is that yields drop if there are defects especially more so if the die is larger, the Pentium 4 Extreme Edition was three times the size of an equivalent ordinary Pentium 4 CPU due to the extra cache so it would be interesting to see what proportion of those ended up being bad.
So for the 0.01% of customers left that are not happy, well Intel do not care because firstly the CPU's are rated for their stock speed and as far as they are concerned "you got what you paid for" and secondly AMD's cpu's don't even come close to Intel's stock speeds let alone any overclocked ones, and even if they do their IPC count is lower, they chew more power and generate more heat so they don't even appear on my radar.
Given Haswell's TDP is going up from 77 watts to 84 Watts my gut feeling is that you'll probably struggle to even get that speed.
Would it have made a signifiant differrence, the first thing people did when IB came out was to de-lid the IHS off the IB and attach the heat sink directly and from what I understand they had better results but the increase was not as much as they had hoped for.
I think if you look back in the past you've never had it so good, you have their premium cpu (2700K/3770K) for only three hundred bucks and an almost equally capable cpu (2500K/3570K) for two hundred. I remember the good old days of a 3.0 Ghz P4 for AUD$500 and a 3.4 Ghz P4 for AUD$750.00, going back a bit further I remember 700 Mhz Celerons and 1.4 Ghz P4's having a ten fold difference in price being AUD$140 vs AUD$1400, and going back still further if you were the proud owner of a PC with an 80386 (or was it an 80286?) CPU I vaguely recall an ad in Byte magazine where you could buy an 80387 (80287?) math co-processor for ONLY $1600. Don't forget as well that money back then would have bought a lot more compared to today.
No I don't think so for the reasons stated above and also the following.
Firstly, Intel has problems running at high frequencies and this started with the end of the Northwood P4 3.40 Ghz cpu ten years ago!, the Prescott's only got up to 3.80 Ghz and then were replaced by the Core CPU's at much lower frequencies and fortunately they (a) had a higher IPC count and (b) came in dual and quad cores.
Secondly, most software/games etc in general usage at most only really uses a 4C/4T cpu like the 2500K/3570K, you don't get that much of a benefit increase going to a 4C/8T cpu like the 2600K/3770K much less a hexacore 6C/12T X58/X79 one either.
Thirdly, the base model cost of a S2011 I7-3820 cpu is comparable to the top of the line S1155 cpu for pretty much the same set of features and motherboard prices are reasonably similiar. Your argument would be valid if the 3820 didn't exist and you were forced to buy a $600 (or more) hexacore cpu as a minimum purchase.
Conclusion:
Unfortunately, the ability to overclock is declining as we move forward due to the variety of reasons I have stated above (and none that appear deliberate by Intel) and there is not much you can do about it, I think that generally people should be grateful for what overclockability they did get in the past and in the future all we can probably look forward to is getting effectively non-overclockable cpu's with maybe even more cores on them to compensate for this but we will have to wait and see if the software written for them keeps up.
Cheers
InternationalHat
[H]ard|Gawd
- Joined
- Aug 13, 2004
- Messages
- 1,481
Mobile it looks great, desktop on the high end it's a joke.
- Joined
- Apr 12, 2003
- Messages
- 7,534
My 300A hit 464 easy no problem, but I did have numerous 333 and 366's that did not hit the target. That is the game of the draw with overclocking though.
$30.000 in a $145.200.000 sale is peanuts. I wouldn't mind paying more 5 cents to have proper thermal material. I would even pay $10 more.
At least the "high-end" price is a serious joke. They never seem to go down and each new chip bringing barely 5% solid gains with $30-$50 price bump isn't making the case either. If these new chips were dropping prices that would be something else, but they're raising them and until AMD gets their act together we're all going to suffer with paying for overpriced chips with a iGPU that supposedly can run Crysis 4 @ 30FPS because that's what EVERY desktop user wants.
I don't see i7 3770K dropping below $299 before it becomes retired unless there is some sales near the end of the inventory.
i dont really see a need to upgrade from my i5 2500k.. looks like it'll be another generation or two.
does that make the chip a flop? not really. i was hoping for more from the new architecture but it looks like they focused hard on the power saving aspect of things which is cool for laptops for sure.
anyways.. if i was building a new system i would definitely get haswell.. otherwise there is no need the performance increase is so small hence the close price points
does that make the chip a flop? not really. i was hoping for more from the new architecture but it looks like they focused hard on the power saving aspect of things which is cool for laptops for sure.
anyways.. if i was building a new system i would definitely get haswell.. otherwise there is no need the performance increase is so small hence the close price points
Parja
[H]F Junkie
- Joined
- Oct 4, 2002
- Messages
- 12,671
Doubtful. New steppings haven't had a significant effect on overclockability in at least a few generations now.
- Joined
- Apr 12, 2003
- Messages
- 7,534
I see a lot of threads with people concerned with clock for clock performance, and i think people are ignoring the new avx2 instructions which i believe will make a very big difference once all the software and games and emulators start using it over sse4.
I'm still running an old i7 920 @ 4.2ghz, and wish it had at least avx1.
It could eventually be the difference between a Pentium, and a Pentium pro "with mmx" which is a big deal if you want to say listen to mp3's in stereo in windows 95.
The new codemasters game grid 2 has a separate executable with avx instructions i'm curious about.
I was going to grab haswell at launch, but i think i will wait for a second stepping like the d0 i7 920
I'm still running an old i7 920 @ 4.2ghz, and wish it had at least avx1.
It could eventually be the difference between a Pentium, and a Pentium pro "with mmx" which is a big deal if you want to say listen to mp3's in stereo in windows 95.
The new codemasters game grid 2 has a separate executable with avx instructions i'm curious about.
I was going to grab haswell at launch, but i think i will wait for a second stepping like the d0 i7 920
- Joined
- Apr 12, 2003
- Messages
- 7,534
http://www.youtube.com/watch?v=xd3wZw_KWCc
It helps alot, and I'm sure Haswell, with AVX2 (over AVX on Sandy/Ivy) will help too. Once more software starts utilizing this and FMA3, Haswell will pull ahead and show it's stuff I predict. When the first wave of ps4/xbox1 games start coming it will start integrating itself better I'm sure.
StillFunkyB
[H]ard|Gawd
- Joined
- Oct 6, 2007
- Messages
- 1,386
Flop? No.
Disappointing? Yeah, a little.
Disappointing? Yeah, a little.
agreed that it's more of a disappointment then flop...I see no reason to upgrade even from my Gulftown CPU let alone people with Sandy or Ivy Bridge...the enthusiast desktop in general is fading so this is probably just a sign of things to come
It was an example. Nothing more. I do not know how much Intel saves by doing cutting the cost of anything. It's a simple example to show how 5 cents adds up quickly. We are most likely talking about a lot more than 4 cents per CPU and probably a hell of a lot more than 600,000 units over the next two years.