Is the 3930K really worth £200 more than the 3820?

[mmj_uk]

The more cores you add to a CPU, the slower the stable clock speeds become (at concept and build stage) and you can't just ignore that fact and claim the 2 cores equals 50% more power. Each of the cores on the 3820 is faster than each of the cores on the 3960K even if there are two less.

but for every 100mhz you overclock the 3930K sees a greater overall percentage increase than a 3820 due to the extra cores, it's theoretically 50% faster clock for clock and gets very close to that in things like Cinebench that use the additional hardware.

If we're talking real world performance then frankly I think you've wasted just as much money as 3930K owners have on having an extra SSD for RAID and buying a GTX680 whilst they're heavily overpriced.

 

[andybird123]

this is a brilliant thread, seemingly designed to provoke the most ire possible except to other 3820 purchasers

on the one side you are saying that anyone who bought 1155 was wasting money and citing the benefits of X79 even though most of those aren't used by the majority of people and on the other end you are telling 3930k users that they wasted their money because in games your 3820 gets most of the performance for a good saving, completely ignoring other benefits of the 3930k in the same way that accuse other people of ignoring the benefits of your 3820

how to invite attacks from the maximum amount of people in one go

well done

 

[estebanrey]

http://www.anandtech.com/bench/Product/552?vs=523

look at the cinebench 11.5 benchmark 3930k shows a 41.5% increase over the 3820, stock clocks aswell, not sure what it would be if the 3930k was running at 3.6ghz

They're just numbers based on sythetic testing, I want real world examples. People don't build PCs to run Cinebench.

On paper <> Real performance

 

[ChileanLlama]

this is a brilliant thread, seemingly designed to provoke the most ire possible except to other 3820 purchasers

on the one side you are saying that anyone who bought 1155 was wasting money and citing the benefits of X79 even though most of those aren't used by the majority of people and on the other end you are telling 3930k users that they wasted their money because in games your 3820 gets most of the performance for a good saving, completely ignoring other benefits of the 3930k in the same way that accuse other people of ignoring the benefits of your 3820

how to invite attacks from the maximum amount of people in one go

well done

Good summary.

The only bit you missed is that if he could afford it he'd buy one (3960X)

 

[ChileanLlama]

They're just numbers based on sythetic testing, I want real world examples. People don't build PCs to run Cinebench.

On paper <> Real performance

You started with posting benchmarks :s and I've given you a real world example with both chips at 4.5Ghz, same render, same settings.

I think to make yourself feel better about this whole thing, specify exactly what software you use and what %age of the time you use it. Also how much idle time your PC has. Clearly you don't see the benefit of a 3930K because you probably have nothing intensive enough that would demand one...

You'll probably find that a 3820 more than adequately fills your needs as you will have lots of idle time, and if a job takes longer due to 2 less cores you won't really notice it.

Which brings us back to an i7 2600/2700

 

[estebanrey]

But practically everyone has already said it relies on the software utilising the extra cores. Now overclock both chips to the same level, and the 3930K is faster.

What do you mean 'to the same level'. Do you mean overclock the 3960K more so it matches the clock speed of the 3820? That's no longer a fair comparison as you're comparing a more overclocked CPU against another.

If we're talking real world performance then frankly I think you've wasted just as much money as 3930K owners have on having an extra SSD for RAID and buying a GTX680 whilst they're heavily overpriced.

I experience noticeable benefits from my RAID system everyday. In fact in terms of video encoding it doesn't matter if your processor can encode faster than mine if it can't write the data to the disk fast enough. My 1GB write speed is going to improve the speed of creating a full HD high bit rate video more than a faster processor would.

I bought the 680 because I wanted a single GFX card capable of running BF3 at ultra settings and 1080p at over 60fps. From the benchmarks I've seen the 680 is the only one that fitted that criteria. The other less reasonable point was the fact I've always owned ATis and I fancied a change.

I could have bought 2 560s for the same price and SLI'd them but then it would leave a more expensive upgrade later on (this way I can simply purchase another 680 when they drop in price rather than having to replace my whole GPU set up.

this is a brilliant thread, seemingly designed to provoke the most ire possible except to other 3820 purchasers

No it's not

on the one side you are saying that anyone who bought 1155 was wasting money

Where did I say that

I answered the question as to why I chose X79 over P67 which was a decision I had to make just two weeks ago. X79 will be supported through to IB-E whereas P67 effectively becomes a defunct chip at the end of this month.

That was a decision I made based on timing, how does that translate into me saying anyone who's ever bought 1155 has wasted their money? In fact I advised a friend about a year ago to build a 1155 when he was going to build a 1366 system.

and citing the benefits of X79 even though most of those aren't used by the majority of people

Again the only advantage I cited was if you're buying right now, the X79 will have more CPUs to come whereas the 1155 won't. Nothing to do with the features of each board (have I even compared them?).

and on the other end you are telling 3930k users that they wasted their money because in games your 3820 gets most of the performance for a good saving, completely ignoring other benefits of the 3930k in the same way that accuse other people of ignoring the benefits of your 3820

I believe that MOST people won't benefit to the tune of £200 by going for the 3930K. I'm not telling anyone personally anything.

That doesn't mean there aren't people who should buy the 3960K, it seems we have a few in here but 3D model rendering is a very niche task.

how to invite attacks from the maximum amount of people in one go

well done

I would say ignoring what people have actually said and completely misrepresenting them is a better way to invite attacks myself.

The only bit you missed is that if he could afford it he'd buy one (3960X)

No i said 'if money was no object' I would buy one, small but significant difference.

 

[ChileanLlama]

What do you mean 'to the same level'. Do you mean overclock the 3960K more so it matches the clock speed of the 3820? That's no longer a fair comparison as you're comparing a more overclocked CPU against another.

Why is it not fair? Intel market it as an unlocked chip. You're posting this on "overclockers.co.uk" forum. Your own 3820 is overclocked, and most people responding have overclocked CPUs too.

That IS the real world. Stop moving the goalposts to suit your own agenda.

 

[cmndr_andi]

Cost and worth are two different things.

Worth is subjective.

If it's not worth the extra cost to you then you have your answer.

This was the first reply in the thread, and in my mind it answered the OP's question succinctly.

To the VAST majority of people there is no benefit going for an i7 3930K instead of a i7 3820 - as they won't make use of the extra two cores or if they did then the extra performance isn't worth the extra money.

However, there is a relatively small number of people where the extra performance offered by the hex core is both made use of and worth the extra cost to them.

Obviously, if someone buys an i7 3930K and the extra performance isn't either utilised or worthwhile then they didn't make a cost-effective CPU choice, however that is their prerogative.

 

[Zuban]

They're just numbers based on sythetic testing, I want real world examples. People don't build PCs to run Cinebench.

On paper <> Real performance

so no one buys a pc to use cinema 4d? we could pick any 3d software and create a scene and compare how two different cpu's render that same scene, would you accept that as being a real world example?, it would be no different to using the cinebench test because thats what it is, which is why its a good benchmark.

 

[estebanrey]

Why is it not fair? Intel market it as an unlocked chip. You're posting this on "overclockers.co.uk" forum. Your own 3820 is overclocked, and most people responding have overclocked CPUs too.

That IS the real world. Stop moving the goalposts to suit your own agenda.

My point was comparing a 3.6GHz 3960k which is overclocked to the tune of 12% is unfair to compare to a 3.6GHz 3820 at no overclock. If you want to compare them OC'd then you should up the 3820 12% too.

It's nothing to do with whether they are overclocked or not, but what is a fair comparison.

I also don't buy the argument that the two will generally overclock to the same levels. Logic would dictate that a CPU that runs 3.6GHz standard would OC to 4.0GHz a lot easier than one at 3.2 would.

Yes it's true if you look at record breaking overclocks you'll see 3960X's on Liquid Nitrogen that made over 6Ghz etc but I would bet if looked at top 100 standard OC's on the 3820 and 3960K, the average 24/7 stable OC would be higher on the 3820.

Again this is only from experience from various boards I'm reading but it seems the 3960K has a lot of "Can't get over 4Ghz" or "High temps at 4.2" type subjects whereas I've yet to see anyone struggle to get 4.5 with the 3820.

But ultimately it's basic physics and there is a reason why the standard clock on the 6C is "only" 3.2GHz when the 3820 is 3.6, because mores cores crammed into one space produces more heat and needs more voltage. Yes they both OC, but I would argue that in general the 3820 will give higher stable clocks more often.

 

[estebanrey]

so no one buys a pc to use cinema 4d? we could pick any 3d software and create a scene and compare how two different cpu's render that same scene, would you accept that as being a real world example?, it would be no different to using the cinebench test because thats what it is, which is why its a good benchmark.

Yes I would because that would provide something that actual meant something (like encoding times).

The Cinebench score is just that, a score or a guide, it doesn't convert to anything meaningful other than being a number.

 

[cmndr_andi]

Here are some tests which compare the i7 3900 series hex cores with i7 3820 quad core, these are tests using real-world applications.

http://www.techspot.com/review/492-intel-core-i7-3820/page6.html
http://www.tomshardware.co.uk/core-i7-3930k-3820-test-benchmark,review-32336-9.html
http://www.hardwarecanucks.com/foru...core-i7-3820-sandy-bridge-e-cpu-review-9.html

Also, please bear in mind that Cinebench may just be a score, but is based on the actual performance using the Cinema 4D software (which is used in industry) - so it does represent real-world performance and hence why it is used as a benchmark test so often.

I'm not making a judgement whether this performance difference is worth the extra money (since I personally don't need to carry out these tasks). But for someone who does a lot of this stuff and/or makes money from it then it may well be worth the premium.

As for overclocking, I think it's fair to say that the i7 3900 series pretty reliably gets to 4.4GHz, while the i7 3820 pretty reliably gets to 4.6GHz. This means a 200Mhz clockspeed advantage over the hex cores when overclocked. However, 200MHz is only 4.6% of 4.4GHz while (at least in theory) the i7 3900s do have 50% more cores.

 

[mmj_uk]

Yes I would because that would provide something that actual meant something (like encoding times).

The Cinebench score is just that, a score or a guide, it doesn't convert to anything meaningful other than being a number.

The engine Cinebench uses is the same as Cinema 4D it doesn't get anymore real than that, the faster the image gets processed the bigger the score you get how is that not meaningful?

I may not really use my 3930K to its fullest but I'm delighted with it, [email protected], 13.4 score in Cinebench... the only thing I regret is not getting one sooner.

Like I said before SSD's in RAID is a waste of money nobody realistically needs 1GBs read/write speeds and GTX680 are massively overpriced atm so you're as wasteful at spending as we are, you just come across like you are bitterly regretting your choice and trying to kid yourself otherwise. Please do us all a favour and just buy one.

 

[ChileanLlama]

My point was comparing a 3.6GHz 3960k which is overclocked to the tune of 12% is unfair to compare to a 3.6GHz 3820 at no overclock. If you want to compare them OC'd then you should up the 3820 12% too.

It's nothing to do with whether they are overclocked or not, but what is a fair comparison.

I also don't buy the argument that the two will generally overclock to the same levels. Logic would dictate that a CPU that runs 3.6GHz standard would OC to 4.0GHz a lot easier than one at 3.2 would.

Yes it's true if you look at record breaking overclocks you'll see 3960X's on Liquid Nitrogen that made over 6Ghz etc but I would bet if looked at top 100 standard OC's on the 3820 and 3960K, the average 24/7 stable OC would be higher on the 3820.

Again this is only from experience from various boards I'm reading but it seems the 3960K has a lot of "Can't get over 4Ghz" or "High temps at 4.2" type subjects whereas I've yet to see anyone struggle to get 4.5 with the 3820.

But ultimately it's basic physics and there is a reason why the standard clock on the 6C is "only" 3.2GHz when the 3820 is 3.6, because mores cores crammed into one space produces more heat and needs more voltage. Yes they both OC, but I would argue that in general the 3820 will give higher stable clocks more often.

Leave them at stock, overclock them by the same amount, overclock each CPU to it's maximum - it probably won't make a blind bit of difference. The 3930K has more cores and with properly threaded software is faster than the 3820.

And for clarity as you seem to be referring interchangeably to the 6 core CPUs; it's the 3820, 3930K and 3960X

I really don't know what you're trying to prove to yourself...

If I were to say you've unarguably got the best, fastest, CPU ever produced and will never be obsolete would you be happy?

 

[Zuban]

Yes I would because that would provide something that actual meant something (like encoding times).

The Cinebench score is just that, a score or a guide, it doesn't convert to anything meaningful other than being a number.

as others have pointed out, its a real world test, and a very good one at that for comparing cpu's as its able to max out all the cores in a cpu without being affected by other bottlenecks in the system.

 

[estebanrey]

The engine Cinebench uses is the same as Cinema 4D it doesn't get anymore real than that, the faster the image gets processed the bigger the score you get how is that not meaningful?

OK let me put it this way, does a 5% higher CB score mean a 5% increase in Cinema 4D production times? If it does then (for that title at least) it means something, otherwise it's just a guide to tell which should perform better no?

I'm not debating Cinebench as a guide to how your PC would perform and it's a perfect tool for CPU reviewers to use to guide folks who do that kind of work on performance but the original argument implied that building a PC yourself to get just to get a high Cinebench score was something someone would do (They may build it based on scores they seen to run Cinema 4D but not just to run a benchmarking tool), that's all I'm saying.

I may not really use my 3930K to its fullest but I'm delighted with it, [email protected], 13.4 score in Cinebench... the only thing I regret is not getting one sooner.

Good, I don't know why you're so defensive though. You talk as if I'm having a pop at all 3930k owners when all I'm asking is for MOST PEOPLE (i.e. not just you and your use of Cinema 4D) is the extra £200 worth it that's all...chill out man.

Like I said before SSD's in RAID is a waste of money nobody realistically needs 1GBs read/write speeds

No one?

Brilliant hypocrisy there my friend. You defend the need for 6 cores on the basis that a few high end specialist applications will take advantage of them but then can't think why anyone would need 1GB write speeds (and presumably the 1GB read speeds you get too)

So as a video encoder I would presume you work with rather large files? Do you not thinking opening a 2GB file in 2 seconds would be better than 4 seconds without RAID or 15 seconds with a Barracuda HD?

So you will pay £200 to [maybe] get 50% increase on rendering times but think I'm an idiot for spending £100 (the cost of the extra SDD) on something that would at least double the times it takes you to open and save large files?

The fact my PC boots up to a ready state in 50 seconds and shuts down in 5 (inc all boot screens) now means I can turn it off when I go to make something to eat, for example, whereas I'd leave it on before isn't an advantage?

Finally on the SSD, had I not bought the second one, I'd only have 120GB of primary drive space and I've gone past that mark already with the programmes I've installed which would have meant having to buy the more expensive 240GB version which is over £100 more anyway. So I could either pay £209 for an SSD that has enough space to be a primary hard drive or £200 on 2 of 120GB versions and get double the read/write times...what would you do?

But if you think I'd have been better with 1 120GB SSD and no more free space then I'll have to disagree.

Claiming people don't need super quick read/write times is stupid, haven't you ever thought what RAM was invented for? To remove the bottleneck between the CPU and the storage drive. Ideally you wouldn't have RAM at all and just one method of storage that could retain it's memory without a current but provide quick enough speeds to work with the CPU on the fly (MRAM was supposed to eventually replace RAM and the storage drive).

and GTX680 are massively overpriced atm so you're as wasteful at spending as we are, you just come across like you are bitterly regretting your choice and trying to kid yourself otherwise. Please do us all a favour and just buy one.

Again I've explained why I bought it, in other words I can justify it. You, it seems, can justify why you bought the 3930K which is fine, I have no issue with it as you obviously do a lot of rendering and movie work.

But unlike me, who has stated since the first post that some people can justify buying certain items, you seem to think that no one can justify using RAID SSDs or buying a GTX 680. You're the one being unreasonable here not me.

Back to the GTX 680, if you can show me a cheaper Nvidia GFX card that can run BF3 @ Ultra settings @ 1080p @ >60fps, I'll agree with you, otherwise you are ignoring my reasons for buying it and just applying personal opinion and are ironically sounding 'bitter' yourself my friend.

 

[estebanrey]

Leave them at stock, overclock them by the same amount, overclock each CPU to it's maximum - it probably won't make a blind bit of difference. The 3930K has more cores and with properly threaded software is faster than the 3820.

You say that as if I claimed it wouldn't be?

£200 faster for most people though (getting annoyed at having to remind people what the actual question is all the time!!!)

If I were to say you've unarguably got the best, fastest, CPU ever produced and will never be obsolete would you be happy?

No, because that completely misses the point of the original question and I've never said anything that could remotely imply I was looking for that kind of assurance.

 

[ChileanLlama]

You say that as if I claimed it wouldn't be?

£200 faster for most people though (getting annoyed at having to remind people what the actual question is all the time!!!)



No, because that completely misses the point of the original question and I've never said anything that could remotely imply I was looking for that kind of assurance.

You'll need to do a survey of most people then, or at least define who you mean by "most people".

Personally I don't find £200 a lot of money, and spending another £200 on a CPU wouldn't force me to skimp on other areas. But am I representative of most people, or are you, or is it something else?

But as I've said numerous times, even without factoring in how much it costs, "most people" won't use or need the extra cores of a 3930K - I think that has already been clearly established.

 

[andybird123]

My point was comparing a 3.6GHz 3960k which is overclocked to the tune of 12% is unfair to compare to a 3.6GHz 3820 at no overclock. If you want to compare them OC'd then you should up the 3820 12% too.

It's nothing to do with whether they are overclocked or not, but what is a fair comparison.

I also don't buy the argument that the two will generally overclock to the same levels. Logic would dictate that a CPU that runs 3.6GHz standard would OC to 4.0GHz a lot easier than one at 3.2 would.

erm... the K series processors come multiplier unlocked... meaning that Intel fully expect you to go overclocking them... they even offer additional warranty that will cover you even if it fails while OC'd

so having paid that extra money for what intel deem to be an overclocking processor, why is it suddenly "unfair" to make use of it and use that as a comparison

my motherboard doesn't handle BCLK overclocking very well if at all (gigabyte), so in my motherboard the 3930k OC's really well and the 3820 wouldn't be able to... so is that unfair?

my motherboard was the cheapest X79 on sale and it also had all the features I wanted, so that saved me over £100 on buying a motherboard that does support BCLK that would allow the 3820 to stretch it's legs

so in my case your rather wonky to begin with logic goes straight out the window, as the 3820 wouldn't overclock at all, where as my 3930K goes to 4.5ghz easily with just a multiplier change and a small voltage tweak

considering I also got a good price on the 3930k, I actually paid the same if not less than you for my system and yet you are trying to tell me that mine is not good value and isn't a goodly % faster than yours...

every comparison I've seen of overclocking the 2 show that they oc to within 1-200mhz of each other on the same system/cooler... 200mhz is pretty negligible at 4.5-4.8ghz so then it does come back to 50% extra core...

am I making full use of all 6 cores currently? maybe not, but considering I went X79 purely on the basis of extra PCIe 3.0 lanes with a view to SLI performance, and given that 2500k users do seem to be hitting CPU bottlenecks with a pair of GTX680's, it may not be all that long before my 3930k shows it's true face

you also keep saying £200 as if it's a hard and fast number, but depending on where you go looking the price differential between a 3820 and 3930k can be as little as £140... knock £100 off for not needing an uber overclocking board to be able to do BCLK clocking and all of a sudden you're in the realms of £50 price difference between a 3820 system @ 4.8Ghz and a 3930K system @ 4.5Ghz... I know which I'd rather have

or you can just go ahead and keep sticking on more and more provisos to limit the scope of your comparison so that it makes it basically meaningless
If you buy the hardware from supplier X and you MUST buy this motherboard and THIS cooler and you only run these 3 apps and the maximum OC you can do is Z%, THEN the 3820 is better value "in my opinion"

 

[Zuban]

OK let me put it this way, does a 5% higher CB score mean a 5% increase in Cinema 4D production times? If it does then (for that title at least) it means something, otherwise it's just a guide to tell which should perform better no?

I'm not debating Cinebench as a guide to how your PC would perform and it's a perfect tool for CPU reviewers to use to guide folks who do that kind of work on performance but the original argument implied that building a PC yourself to get just to get a high Cinebench score was something someone would do (They may build it based on scores they seen to run Cinema 4D but not just to run a benchmarking tool), that's all I'm saying.

it does yes, they simply convert the time to a number to make it easier to compare and sort tables of results etc. Its a very valuable benchmark for anyone who does this kind of work.

what is the original question though is it, is the 3930k a waste of money for most people?, which is a bit pointless because most poeple dont buy 3930k's.

or is the question was it a waste of money for many of those who bought a 3930k because they wont use it, in which case its an unanswerable question, everyones reasons will be different, a good amount of the time my processor is sat idling, or being barely tasked, so half the time its wasted, the other half its got all its cores maxxed rendering or whatever, could i have got away with a 3820, yes i could, am i glad i bought a 3930k yes i'm delighted i think its well worth it. Its like people buying expensive audi's to get to work when a fiat panda will do the job just fine, if you asked them all why you'd be there forever and never find one definitive answer. you will never find an answer to this, and does it matter if someone wants a 3930k just because its a fast processor and it makes them happy even if they dont make full use of it? whats wrong with that, people like nice cars, huge tv's they dont really need, doesnt matter.