Haswell i7 Engineering Sample Pinned Up Against i7 3770K

[neon neophyte]

your math is terrible. 4.6x1.02=4.692

"extreme cooling" a 30 dollar cooler will do it, hell, i reused the cooler off my socket 775 system which only cost me 50 bucks 5 years ago

just stop talking, you dont have a clue what you are talking about

 

[blazorthon]

[citation][nom]neon neophyte[/nom]your math is terrible. 4.6x1.02=4.692"extreme cooling" a 30 dollar cooler will do it, hell, i reused the cooler off my socket 775 system which only cost me 50 bucks 5 years agojust stop talking, you dont have a clue what you are talking about[/citation]

My mathis terrible? How is it terrible? I did 4.6+2% = ~4.7. The difference between that and 4.692GHz is error of margin as far as performance goes. You might need some reading comprehension before accusing me of bad math.

Name a single $30 cooler that will get an average i5-2500K reliably to 5GHz. I've built dozens of computers with such coolers and many of them had that CPU and looking online, I don't even read of anyone getting such a frequency properly running with $30 coolers regardless of the motherboard. The only times I read about it are with either a few rare exceptional i5-2500Ks and still good cooling or with top-end cooling such as an NH-D14 or better.

You're not a good judge of who knows what if you failed to even read what I said before criticizing it.

 

[neon neophyte]

*eyeroll* how about my archaic 50 dollar cooler or an evo.

not only does sandy overclock considerably higher, it does it cooler.

"4.6GHz Ivy versus 5GHz Sandy. 4.6GHz+~10% is about equal to 5GHz." no, just no.

 

[Fulgurant]

not only does sandy overclock considerably higher, it does it cooler.

"4.6GHz Ivy versus 5GHz Sandy. 4.6GHz+~10% is about equal to 5GHz." no, just no.

4.6 * 1.1 = 5.06, no?

Not seeing a problem with that statement.

 

[blazorthon]

[citation][nom]neon neophyte[/nom]*eyeroll* how about my archaic 50 dollar cooler or an evo.not only does sandy overclock considerably higher, it does it cooler."4.6GHz Ivy versus 5GHz Sandy. 4.6GHz+~10% is about equal to 5GHz." no, just no.[/citation]

4.6+10% is about equal to 5GHz. "No, just no" is wrong. I never said that Ivy is 10% better per Hz than Sandy. I only said that about 10% is the difference between 4.6 and 5. Again, reading comprehension.

Sandy does not overclock better by enough to make a discernible difference and that it's somewhat cooler doesn't mater much.

You'd have to be more specific about your own cooler than merely the price and implied age. If by Evo you refer to the CM Hyper212 Evo, then it is a $30 cooler and it usually gets around 4.4GHz to 4.7GHz with the i5-2500K from what I've seen, most often around 4.6GHz. The i5-3570K tends to get around 4.3GHz to 4.5GHz or thereabouts and is closer on average than if you did a higher end cooling comparison due to Sandy getting better gains from higher end coolers than Ivy does, mostly because of Intel's crap paste between Ivy's IHS and die.

With coolers such as the NH-D14 and Phantek's top cooler, the difference on average between Sandy and Ivy is a little greater than the difference with lower end coolers, but again, it's still only a minor difference.

 

[neon neophyte]

im well aware of why ivy runs so hot

lets just put this to rest by saying theyre about equal once u take overclocking into account, except the ivy runs hotter to do it.

the only thing i like more about ivy is that you can delid it for even better overclocking, which puts it above sandy imo. delidding is not for the feint of heart though. way too many people end up with a dead cpu. still pretty cool though.

 

[blazorthon]

im well aware of why ivy runs so hot

lets just put this to rest by saying theyre about equal once u take overclocking into account, except the ivy runs hotter to do it.

the only thing i like more about ivy is that you can delid it for even better overclocking, which puts it above sandy imo. delidding is not for the feint of heart though. way too many people end up with a dead cpu. still pretty cool though.

Agreed on all points there.

 

[neon neophyte]

how many volts do you consider safe? ive read anything up to 1.5 is safe.

ive been running my cpu between 1.41 and 1.45 volts for the last 2 years without issue.

truth is sandybridge is an established work horse with very very few cases of the cpu degrading or dying.

 

[Fulgurant]

Here's the thing: you just simply can't substantiate the claim that it's uncommon for a 2500k to hit 5 GHz when there are over 17000 submissions with an average AIR overclock of 5 GHz.

I understand that hwbot attracts the overclocking cloud and not average Joe clueless about technology, but a handful of reviews is simply not enough to back up the claim that 5 GHz is an uncommon overclock.

What is uncommon is a 2500k that will hit 5 GHz on low enough volts to be safe/cool enough for 24/7 use on air cooling. But it's not uncommon at all that the chip will actually get to 5 GHz. So maybe we're talking about two different things here, but as far as what the chips will do - 5 GHz is certainly not uncommon.

For example, I can get into windows at 5.4 GHz on my 2600k, but it takes way too many volts to be usable on my NH-D14 (and it probably isn't stable for anything more than a CPU-z validation). The highest OC I can actually use on volts/temps I'm comfortable with is 4.8 GHz.

If it's not a sustainable overclock, then is it relevant to a discussion about the practical performance differences between Sandy and Ivy Bridge?

Also, and for what it's worth, I tend to agree with blazorthorn that reputable review sites are the best indication of what's practically achievable; reviewers' overclocks may be more conservative than they strictly need to be, but at least we have a reasonable expectation that they're truly stable. As you point out, the stability of the overclocks you see on hwbot is questionable. Some, if not most of those people are just pushing limits for the sake of pushing limits.

And there's nothing wrong with that; heck, I was once in that crowd, a long time ago. But nowadays I'll trade 10-20% in speed for guaranteed stability (and/or comfortable temps), every time -- and the comparison isn't even close. I suspect you agree, because you're not running your chip at 5 GHz.

In any case, I've enjoyed reading the discussion. Thanks to all parties.

 

[neon neophyte]

ive had my sandy overclocked to 5ghz on a cheap air cooler for 2 years now.

it is easily sustainable

 

[Fulgurant]

ive had my sandy overclocked to 5ghz on a cheap air cooler for 2 years now.

it is easily sustainable

You are only one example, though.

The argument, unless I missed something, concerns the average overclock. Just using this thread as a sample, we have you (at 5 GHz), and BigMack (at 4.6-4.8 GHz). Obviously, that's not conclusive one way or the other.

Then we have a multitude of reviews from sites like Tom's, Anandtech, and so on. FWIW, I don't think I've ever seen an article that listed a stable 5 GHz overclock unless the explicit point of the article was to push limits (testing extreme cooling solutions or whatever).

Then we have the user-generated data available from hwbot, which suggests that 5 GHz is technically reachable on the average chip, but it doesn't (AFAIK) offer us any insight into the stability and/or practical usefulness of those overclocks.

 

[neon neophyte]

bigmack could get 5ghz if he had the right mobo and was willing to push the volts.

he just doesnt like the idea of going over 1.4 volts, doesnt mean it cant be done.

i still think most sandys can do it with enough volts and the right motherboard.

you might say i have a golden chip but i guarantee it wouldnt do 5ghz on many other motherboards

 

[Fulgurant]

bigmack could get 5ghz if he had the right mobo and was willing to push the volts.

he just doesnt like the idea of going over 1.4 volts, doesnt mean it cant be done.

i still think most sandys can do it with enough volts and the right motherboard.

you might say i have a golden chip but i guarantee it wouldnt do 5ghz on many other motherboards

Maybe, maybe not. Some chips are better overclockers than others.

Anyway, I don't presume to speak for BigMack, but he did say that he doesn't like his chips to rise above 70 C in games, and 90 C under stress testing. Those are sensible limits, IMO. I sure as hell wouldn't push those temps for the sake of a couple hundred extra MHz, and I wouldn't consider anyone who does push those temps to be a representative example of a given CPU's sustainable overclockability.

YMMV. You may be right that the motherboard is the culprit, but if that's true -- if you need a certain class of motherboard to sustain 5 GHz on Sandy Bridge, then doesn't motherboard selection further limit the chip's average (or "common") overclockability? Which motherboard do you use, by the way?

 

[neon neophyte]

asrock extreme 4 z68, not gen 3

 

[Fulgurant]

asrock extreme 4 z68, not gen 3

Good to know, thanks.

That's an excellent motherboard, by all accounts. Still, if you're in for a $200 (at launch) motherboard, I'd surely hope that you could overclock well. At this point, given that you can pick up z75/77 boards for about $100, it seems that blazorthon's point about high-end cooling holds true, if only indirectly; strictly as a value proposition, there's very little practical difference between spending extra money on a cooler or spending extra money on a motherboard.

On the other hand, a more expensive motherboard might offer other features besides higher over-clocking headroom. An expensive cooler won't.

 

[neon neophyte]

the extreme 3 was just as capable an overclocker, for 120-130 dollars. just had less features.

 

[neon neophyte]

cant help but point out you have an asrock board as well, champs at overclocking i say, champs.

 

[blazorthon]

Yes, I was talking about practical overclocks, not impractical overclocks. Being able to hit over 5GHz doesn't matter in a performance discussion if it's not reasonably sustainable as an average overclock.

I do like ASRock too these days and one of my most commonly used board in client's computers over the last year has been the ASRock Extreme4 Z77. I occasionally find it for around $100 online and haven't found another motherboard deal online that's as good as it is without a combo or something like that.

 

[neon neophyte]

i dont see how 5ghz is impractical. ive been doing it for 2 years on a cheap cooler.

 

[blazorthon]

i dont see how 5ghz is impractical. ive been doing it for 2 years on a cheap cooler.

Most people don't get a practical overclock to 5GHz with an i5-2500K and a cheap cooler. Your chip may be exceptionally good at overclocking.

 

[neon neophyte]

i still think its the motherboard.

bigmack gets better results and he has a nicer asrock board. coincidence? hmm? ;p

it was a pretty well rated cooler back in the day to be honest, true 120. i got it for 50 bucks when i had a socket 775 rig, then transfered it over to my 1055. (asrock boards have the mounting holes for other sockets. yay.)

 

[blazorthon]

i still think its the motherboard.

bigmack gets better results and he has a nicer asrock board. coincidence? hmm? ;p

it was a pretty well rated cooler back in the day to be honest, true 120. i got it for 50 bucks when i had a socket 775 rig, then transfered it over to my 1055. (asrock boards have the mounting holes for other sockets. yay.)

Agree to disagree, I suppose.

 

[deksman]

I don't think AMD is so far behind Intel as people seem to think in the desktop segment (at least).
The FX-8350 seems to be faster than i7-3770K when it comes to multi-threaded programs (3d studio Max comes to mind - I've seen the benchmarks and the Intel quad is either on par or slower than FX-8350).
In most games, its more or less comparable to i5 and i7, but certain games such as F1 and some flight simulators seem to give Intel the edge by at least 10 to 25 FPS (making the 8350 appear to be on the level of i3 in those specific games).

Personally, I don't play any of the above games (and the ones I do play, the AMD can seemingly handle with ease, but I do use 3d Studio Max... so on that front, the FX-8350 seems to be more than enough (though granted, its power consumption is greater than Intel's - which can be undervolted on stock clocks and reduced by about 14%).

I would also agree that one of the issues we are having are programs not specifically written to take advantage of the newest architecture and power of these CPU's.

AMD is severely behind Intel in the mobile sector, but that's mostly because their 'quad core A10' is not really a quad chip, instead has 2 modules (which seems to put it more or less on par with Intel mobile dual core i3).
On another hand, in numerous tasks it does approach mobile i5 (Ivy Bridge) - but it definitely needs to shake things up and bring a 'proper' quad core that could actually offer a viable alternative to Intel's i7 (even if it's 15% slower).

Finally, one other issue is that Intel is severely dominating the market share, and people automatically associate AMD as being 'inadequate' for their needs (which is not the case).
That aside, laptop manufacturers for example have also been packing the AMD mobile APU's in stupid form factors (the A4 and A6 into 17" chasis for example - whereas even the A10 could be easily put into 14" form factor or even 13" seeing its thermal envelope is within 35W - it wasn't until just recently with latest releases that manufacturers decided to put APU's into smaller form factors [which incidentally seems to be their main purpose]).

My point is that AMD can be a viable purchase - though the only area where Intel beats it in the mobile department is performance, while in the desktop segment, thermal envelope seems to beat AMD for same performance (but its also more expensive than AMD) and of course has extreme versions which ARE faster, whereas otherwise, the 8350 seems to be comparable to the 3770K performance-wise.

AMD definitely needs more exposure if it wants to be more 'competitive' with Intel, but it also needs to bring better/faster CPU improvements than mere 10-15%.

 

[neon neophyte]

I don't think AMD is so far behind Intel as people seem to think in the desktop segment (at least).
The FX-8350 seems to be faster than i7-3770K when it comes to multi-threaded programs (3d studio Max comes to mind - I've seen the benchmarks and the Intel quad is either on par or slower than FX-8350).
In most games, its more or less comparable to i5 and i7, but certain games such as F1 and some flight simulators seem to give Intel the edge by at least 10 to 25 FPS (making the 8350 appear to be on the level of i3 in those specific games).

Personally, I don't play any of the above games (and the ones I do play, the AMD can seemingly handle with ease, but I do use 3d Studio Max... so on that front, the FX-8350 seems to be more than enough (though granted, its power consumption is greater than Intel's - which can be undervolted on stock clocks and reduced by about 14%).

I would also agree that one of the issues we are having are programs not specifically written to take advantage of the newest architecture and power of these CPU's.

AMD is severely behind Intel in the mobile sector, but that's mostly because their 'quad core A10' is not really a quad chip, instead has 2 modules (which seems to put it more or less on par with Intel mobile dual core i3).
On another hand, in numerous tasks it does approach mobile i5 (Ivy Bridge) - but it definitely needs to shake things up and bring a 'proper' quad core that could actually offer a viable alternative to Intel's i7 (even if it's 15% slower).

Finally, one other issue is that Intel is severely dominating the market share, and people automatically associate AMD as being 'inadequate' for their needs (which is not the case).
That aside, laptop manufacturers for example have also been packing the AMD mobile APU's in stupid form factors (the A4 and A6 into 17" chasis for example - whereas even the A10 could be easily put into 14" form factor or even 13" seeing its thermal envelope is within 35W - it wasn't until just recently with latest releases that manufacturers decided to put APU's into smaller form factors [which incidentally seems to be their main purpose]).

My point is that AMD can be a viable purchase - though the only area where Intel beats it in the mobile department is performance, while in the desktop segment, thermal envelope seems to beat AMD for same performance (but its also more expensive than AMD) and of course has extreme versions which ARE faster, whereas otherwise, the 8350 seems to be comparable to the 3770K performance-wise.

AMD definitely needs more exposure if it wants to be more 'competitive' with Intel, but it also needs to bring better/faster CPU improvements than mere 10-15%.

http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328-8.html

a 4ghz 8350 isnt even keeping up with a 3.5ghz 3770k... and it is using quite a bit more electricity to do it

 

[blazorthon]

[citation][nom]deksman[/nom]I don't think AMD is so far behind Intel as people seem to think in the desktop segment (at least).The FX-8350 seems to be faster than i7-3770K when it comes to multi-threaded programs (3d studio Max comes to mind - I've seen the benchmarks and the Intel quad is either on par or slower than FX-8350).In most games, its more or less comparable to i5 and i7, but certain games such as F1 and some flight simulators seem to give Intel the edge by at least 10 to 25 FPS (making the 8350 appear to be on the level of i3 in those specific games).Personally, I don't play any of the above games (and the ones I do play, the AMD can seemingly handle with ease, but I do use 3d Studio Max... so on that front, the FX-8350 seems to be more than enough (though granted, its power consumption is greater than Intel's - which can be undervolted on stock clocks and reduced by about 14%).I would also agree that one of the issues we are having are programs not specifically written to take advantage of the newest architecture and power of these CPU's.AMD is severely behind Intel in the mobile sector, but that's mostly because their 'quad core A10' is not really a quad chip, instead has 2 modules (which seems to put it more or less on par with Intel mobile dual core i3).On another hand, in numerous tasks it does approach mobile i5 (Ivy Bridge) - but it definitely needs to shake things up and bring a 'proper' quad core that could actually offer a viable alternative to Intel's i7 (even if it's 15% slower).Finally, one other issue is that Intel is severely dominating the market share, and people automatically associate AMD as being 'inadequate' for their needs (which is not the case).That aside, laptop manufacturers for example have also been packing the AMD mobile APU's in stupid form factors (the A4 and A6 into 17" chasis for example - whereas even the A10 could be easily put into 14" form factor or even 13" seeing its thermal envelope is within 35W - it wasn't until just recently with latest releases that manufacturers decided to put APU's into smaller form factors [which incidentally seems to be their main purpose]).My point is that AMD can be a viable purchase - though the only area where Intel beats it in the mobile department is performance, while in the desktop segment, thermal envelope seems to beat AMD for same performance (but its also more expensive than AMD) and of course has extreme versions which ARE faster, whereas otherwise, the 8350 seems to be comparable to the 3770K performance-wise.AMD definitely needs more exposure if it wants to be more 'competitive' with Intel, but it also needs to bring better/faster CPU improvements than mere 10-15%.[/citation]

FX-8350 usually doesn't beat the quad-core i7s in highly threaded performance. They usually tie or the FX-8350 is a little behind them, but the FX-8350 does usually best the i5s when it comes to highly threaded performance. There are only a few workloads where the FX-8350 will beat the quad core i7s regularly.

The FX-8350 is usually closer to the i3s than it is to the i5s and i7s in gaming.

Being modular doesn't mean it's not a true quad core. They are truly quad core CPUs. They are also very lowly clocked CPUs with poor performance per Hz going up against higher clocked CPUs with much higher performance per Hz. That's easily enough to counteract the core count difference compared to the mobile i5s in workloads that don't scale nearly perfectly across four or more threads.

It is true that there are some bottle-necks (such as the insufficient front end) in AMD's modular CPUs, but these are not issues inherent in modular architectures. These are merely problems in AMD's current designs and they're problems that are to be resolved in the next AMD architectural tweak, Steamroller. Steamroller is also likely to make a much more than 10-15% difference, if not in performance all-over, then at least in performance per watt and giving enough overclocking headroom as a result to get a more significant performance improvement if desired.

AMD most certainly could use a lot more exposure, I'll give you that one.