Caporegime
The 3820 is X79.
Read point a)
He wasn't using c) as a justification over the 3820, he was using it over the 2600k/1155 platform.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Is the 3930K really worth £200 more than the 3820?
- Thread starter estebanrey
- Start date
Read point a)
He wasn't using c) as a justification over the 3820, he was using it over the 2600k/1155 platform.
...and an encoder written to utilise all cores
Which is why I said in a perfect situation
Buchanan0204
You got your maths wrong i'm afraid.
A 6 core CPU has 50% more cores than a 4 core.
If all cores are kept busy and clock speeds are equal then it will be 50% faster than a 4 core.
However you have by accident come up with the increase one normally sees when going from a 4 core to a 6 core for tasks such as encoding.
33% is about right, seems most software doesn't fully utilise the extra cores as well as it could.
Certainly that is what i see when using handbrake, it uses 4 cores really well but doesn't manage to fully utilise a 6 core CPU.
When not encoding my 3930k is used for number crunching.
The software i use for that is single threaded but i run 12 clients in parallel.
In this scenario i do see the full 50% increase in work done over the 4 core.
estebanrey
You need to be aware of the effect HT has on reported Gflops in IBT.
For a 3820 to get 120+ Gflops at 4.5Ghz HT has to be disabled.
A 3930K also running at 4.5Ghz with HT disabled scores 180+ Gflops
I did a quick test see image below.
3930k ibt.png
As to the 3820 i use one in my gaming PC.
The X79 platform is excellent for high performance gaming setups,
At the present time a hex core isn't really required for gaming so the 4 core 3820 is a perfect fit.
I'm already seeing benefits of the X79 platform with the SLi'd 680's i have in it.
X79 gives you a high performance long life platform.
BTW
A 3820 with an asrock extreme 4 isn't overly expensive and makes a great gaming platform.
As to the question of is it worth paying a lot more for the 6 core over the 4, it depends on your intended usage,
For gaming no, unless you want to bench QUAD SLI/crossfire setups.
For encoding again no, unless you do a lot of it.
I do a lot of encoding both for my own pleasure and for work, so the decision to purchase a six core SBe was an easy one.
The 3820 and the 680's however were a pure indulgence.
You got your maths wrong i'm afraid.
A 6 core CPU has 50% more cores than a 4 core.
If all cores are kept busy and clock speeds are equal then it will be 50% faster than a 4 core.
However you have by accident come up with the increase one normally sees when going from a 4 core to a 6 core for tasks such as encoding.
33% is about right, seems most software doesn't fully utilise the extra cores as well as it could.
Certainly that is what i see when using handbrake, it uses 4 cores really well but doesn't manage to fully utilise a 6 core CPU.
When not encoding my 3930k is used for number crunching.
The software i use for that is single threaded but i run 12 clients in parallel.
In this scenario i do see the full 50% increase in work done over the 4 core.
estebanrey
You need to be aware of the effect HT has on reported Gflops in IBT.
For a 3820 to get 120+ Gflops at 4.5Ghz HT has to be disabled.
A 3930K also running at 4.5Ghz with HT disabled scores 180+ Gflops
I did a quick test see image below.
3930k ibt.png
As to the 3820 i use one in my gaming PC.
The X79 platform is excellent for high performance gaming setups,
At the present time a hex core isn't really required for gaming so the 4 core 3820 is a perfect fit.
I'm already seeing benefits of the X79 platform with the SLi'd 680's i have in it.
X79 gives you a high performance long life platform.
BTW
A 3820 with an asrock extreme 4 isn't overly expensive and makes a great gaming platform.
As to the question of is it worth paying a lot more for the 6 core over the 4, it depends on your intended usage,
For gaming no, unless you want to bench QUAD SLI/crossfire setups.
For encoding again no, unless you do a lot of it.
I do a lot of encoding both for my own pleasure and for work, so the decision to purchase a six core SBe was an easy one.
The 3820 and the 680's however were a pure indulgence.
Last edited:
i skimped on other things in my rig, ssd/superfast memory etc to get me a 3930k, at the time my other option was an 2600k, i do 3d rendering though so processer power is king, the only benchmarks i'm interested in are things like the cinebench 11.5, and in that instance i'm able to pull around 40% extra performance from having 50% extra cores, its diminishing returns yes but well worth it to me as its speeded up my work flow tremendously.
I've since added an ssd, and its made absolutly chuff all difference to my output, and i'd never reccomend anyone doing similar work to me to go for a lesser cpu in order to include one, its also something thats easy to add later, my cheap memory makes no difference aswell as my benchmark figures for the types of software i use give the same results as people using memory up to four times as expensive as mine.
I've been using properly mutlithreaded software since long before intel or amd had dual core processors on the market, there were plenty of machines available with multiple sockets, so multithreading was well established, you would probably find that the "When dual cores first appeared almost no software could utilise them" is not very true, more like the software you were aware of wasnt able to do it, my particular 3d software has been multithreaded since i started using it in the mid 90's, it had the option of up to 16 threads at that point.
I've since added an ssd, and its made absolutly chuff all difference to my output, and i'd never reccomend anyone doing similar work to me to go for a lesser cpu in order to include one, its also something thats easy to add later, my cheap memory makes no difference aswell as my benchmark figures for the types of software i use give the same results as people using memory up to four times as expensive as mine.
I've been using properly mutlithreaded software since long before intel or amd had dual core processors on the market, there were plenty of machines available with multiple sockets, so multithreading was well established, you would probably find that the "When dual cores first appeared almost no software could utilise them" is not very true, more like the software you were aware of wasnt able to do it, my particular 3d software has been multithreaded since i started using it in the mid 90's, it had the option of up to 16 threads at that point.
Last edited:
- Joined
- 8 Mar 2007
- Posts
- 10,938
I never said it was
Firstly I'm not only interested in gaming, I don't know what gave you that impression. The reason I mentioned gaming in my OP is because I suspect a lot of 3930K users are mainly gamers and such (in my opinion) have wasted £200.
I game yes, but I also do a moderate amount of video encoding (My Samsung SII films in 1080p and I use the PC for editing those videos) but not my main task. I probably use my PC most for programming via Visual Studio and as an SQL server for testing my software.
The reason I said "If I won the lottery I'd get the 6960X" is because my OP is about what affordable and important to the 99% of us who have some kind of budget, if you have no budget then there's no problem in what you spend so you might as well.
As I said, just seen a few prospective builds where people are opting for the 3960K then buying a mid-range PSU, RAM and sticking with hard drives etc because they'd blown most of their budget on the CPU.
Obviously if you REALLY are going to benefit from it, like PCZ said then fair enough but I suspect that most people don't.
As I said, a 3960X owner would probably disagree and be able to justify paying twice the price for the X model.
A) Longevity - X79 still has Ivybridge-E to come whereas z68 is about to be replaced at the end of the month. A better question would be why I didn't wait for Z77 really.
B) 1155 being cheaper is a myth, at least in my case...
3820 = £229
2700k = £265 (even the 2600 non-K is more expensive than the 3820!)
The motherboards have a higher general range of prices (although were higher before 2011 arrived) but hypothetical the P67 build (shopping list) I did before I finally splashed out on the X79 had a £200 mobo on anyway.
So I've probably spent an extra £20 on the rig I was planning before in which reason A was enough to justify it.
Associate
I never said it was
Firstly I'm not only interested in gaming, I don't know what gave you that impression. The reason I mentioned gaming in my OP is because I suspect a lot of 3930K users are mainly gamers and such (in my opinion) have wasted £200.
I game yes, but I also do a moderate amount of video encoding (My Samsung SII films in 1080p and I use the PC for editing those videos) but not my main task. I probably use my PC most for programming via Visual Studio and as an SQL server for testing my software.
But you only mentioned gaming in your original post, so that was the only reference point you gave us. Now you mention your other uses, then quite possibly the extra £200 is worth it. But only you can make that judgement based on how much you do of each activity and how well it is multithreaded.
That's not what your OP came across as. Everyone will have a different budget from doing it on the cheap, right up to money no object. The relative value of the different CPUs can only then be judged by the individual, the amount of money they have available and the tasks they want to perform.
Fools (not PCZ). Like I said, you've opted for a well balanced system within your budget. Be happy.
They probably can to some degree, but the point was there is less of a differential between 3960 and 3930 than 3930 and 3820 purely due to the number of cores.
But you could have built a cheaper z68 system with an OC 2600K, that has a similar spec/capability to the one in your sig. You chose not to with very valid reasons to yourself, no doubt the people who have gone for 6 core CPUs had very valid reasons for themselves. That's what I was aiming to demonstrate to you.
- Joined
- 8 Mar 2007
- Posts
- 10,938
When I mentioned gaming I wasn't referring to me though, just the concept of it and how the 3960K probably wouldn't be an advantage (not considering the extra GPU power you can use the £200 saved on).
I don't think my OP implies I was mentioning gaming for personal reasons, I was doing it because it probably is the main reason for new builds if you take the whole of the this forum into account.
Yes but £200 isn't worth the same as it is to a millionaire as it is to the rest of us so losing it doesn't have the same effect. Also the millionaire, won't then downgrade the other parts which is more probable for the 99%.
I am, I guess I'm hoping this thread may make some people who are building new systems think a little I guess rather than have a pop at current 3930K owners.
Hence the title being "is the 3930K really worth £200 more than the 3820" and not "ZOMG peeps who bought the 3930K wasted 200 hundred squid SUCKERS!!!!
But what about the differential between the clock speeds, downgraded to 3.2 to accomodate those cores and what about the software you're using?
The 50% better thing is only true under specific circumstances and you have to OC the 3960K to 3.6GHz as well to compare to a stock 3820.
It's simply a myth to imply that your PC will generally be 50% faster using the 3930K. I'm not saying that is what you're claiming but there are a lot of caveats to the 50% more power figure that aren't being mentioned here.
Of course, there are people who obviously use specialist software and use their PCs for non-mainstream reasons who can justify it. I'm not claiming that under no conditions can you justify buying a 3930K, I am asking do most people need it.
Notice 'most people' not the small number of people who do 3D modelling or the tiny number of people who encode videos constantly despite not having a job that requires it (as they'd use industry based equipment and not a home PC anyway).
Last edited:
Associate
I'm in the middle of doing some reviews and part of that is comparing the SB and SB-E. Currently I have looked at the 2600K (1155) and the 3930K (2011).
There are two main areas that the 2011 socket x79 chipset gives with the 3930K. These are the additional cores and the quad channel memory. Clock for clock i'm seeing about a 12% improvement in my media benchmarks for the 6 core 3930K. Games are very similar with about a 1% improvement which is probably down to the Quad channel memory.
Theoretical tests show the ram as over 100% faster and arithmetic tasks as 35-40% faster.
So personally i'm not seeing a significant improvement in real life, so would have to say that the 1155 SB / IB represents the best value for money because the motherboards are so much cheaper.
There are two main areas that the 2011 socket x79 chipset gives with the 3930K. These are the additional cores and the quad channel memory. Clock for clock i'm seeing about a 12% improvement in my media benchmarks for the 6 core 3930K. Games are very similar with about a 1% improvement which is probably down to the Quad channel memory.
Theoretical tests show the ram as over 100% faster and arithmetic tasks as 35-40% faster.
So personally i'm not seeing a significant improvement in real life, so would have to say that the 1155 SB / IB represents the best value for money because the motherboards are so much cheaper.
Simon
You might wan't to investigate why the X79 platform has higher performance with multi GPU configs than 1155.
I am seeing a fairly significant improvement in GPU performance on the X79 over my older P67 machine.
This isn't down to CPU performance differences as the 3820 and 2500/2600
have near identical IPC.
It may be PCIe latency is lower on the X79 or that the latest cards. 680's and 7970's are now being held back a bit by the 8x PCIe on P67/Z68.
Maybe the quad chan memory is helping ?
You might wan't to investigate why the X79 platform has higher performance with multi GPU configs than 1155.
I am seeing a fairly significant improvement in GPU performance on the X79 over my older P67 machine.
This isn't down to CPU performance differences as the 3820 and 2500/2600
have near identical IPC.
It may be PCIe latency is lower on the X79 or that the latest cards. 680's and 7970's are now being held back a bit by the 8x PCIe on P67/Z68.
Maybe the quad chan memory is helping ?
I think even 3960X owners would admit they've payed a massive premium for mainly just the overclocking potential.
3930K v 3820
Fully unlocked
2 additional cores
50% more performance
100% extra cost
3960X v 3930K
3MB extra L3 cache
1-5% more performance (?)
Supposedly the best silicon
100% extra cost
3930K is clearly a great value product compared to 3960X but there is still a slight premium (£100?) on it compared to 3820 because it's 6 core and like 3960X has unrivalled performance.
Last edited:
Associate
Your original post was you trying to justify your purchase...the only reference point you gave was gaming. You now seem to be looking at the discussion as a way of informing others, that's good, but not how you set it out originally.
Exactly - but who are the rest of us? Even the rest of us will have different incomes and budgets. If you'd posed the original question based on a specific budget it may have been different.
I am, I guess I'm hoping this thread may make some people who are building new systems think a little I guess rather than have a pop at current 3930K owners.
Hence the title being "is the 3930K really worth £200 more than the 3820" and not "ZOMG peeps who bought the 3930K wasted 200 hundred squid SUCKERS!!!!
But as already said, you didn't set that out in the OP, it was about justifying your purchase to yourself...
For 24/7 overclocks the 3820, 3930 and 3960 all appear to be able to hit the same speeds give or take. The 3930 and 3960 are harder to get there. But for proper multithreaded software more cores and a hit on raw clockspeed is probably a better position to be in then less cores and slightly higher clockspeed.
Who's to say the large majority of 3930 owners aren't making full use of them? But you are right most people don't need a 3930, but then I don't see "most people" considering them or buying them. Just looking at this forum alone, the clear favourite for most people is the i5 2500K.
Associate
My tests between Z68 and X79 with a single EVGA 680 GPU show only marginal improvements in 3D Mark tests and as I say no real improvement in Game benchmarks (Dirt3, Crysis2 and F1-2010). I worry that the 3D mark tests use the CPU too much to be a fair comparison.
I do not have the resorces for a second 680 GPU co will be unable to comment on multiple GPU, but in theory the X79 chipset is promoted as a better solution for using multiple PCI-E slots.
Caporegime
The 3930K is 204% of the cost of the 3820 (£299.99 vs £469.99).
- Joined
- 8 Mar 2007
- Posts
- 10,938
Arggghhh, can people stop making this claim. It's not 50% more performance at all.
Firstly, it has 12% slower clock speed (3.2 vs the 3.6 on the 3820) so overall it's 33% faster on paper (so actually Buchanan0204 was right in the first place).
3.2 x 6 = 19.2 GHz overall CPU Power
3.6 x 4 = 14.4 GHz
Difference = 33%
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.
Even then you still need certain software (the vast majority of software that most people use isn't) to see that benefit and even then there's another factor...how much the software's performance is tied to the CPU.
Only if you had something that was 100% CPU intensive (so it would not use the RAM, GPU, Hard drives etcs at all) would the 33% figure be accurate..but no sofware ONLY uses the CPU for performance.
So for example if a piece of sofware's performance was afftected by 50% based on the 'power' of your CPU, then the 3960K would only out-perform the 3820 by 16.5%
So the reality is, the 3960K can be around 30% faster for CPU intensive pieces of software that were coded specifically for muiti-cored PCs; but for most people/applications it will provide you will a couple of percent extra power...in fact in a lot of applications it would be slower due the lower clock speed.
If anyone can find me a benchmarking test that shows the 3960K outperforming the 3820 by 50% in any field I'd love to see it. In fact, here's some benchmarks comparing the two.
Photoshop = 0.8% increase
H264 encoding = 7% faster
Music encoding = 38% faster
Gaming = 4% faster
The only area the 3960K really exceeds is music encoding and even then it still a way off being 50% better.
Last edited:
Yup. I'm full up with the cold that is going around at the moment...
Not according to: http://www.anandtech.com/bench/Product/434?vs=287
plus I have my i7-2600K o/c to 4.3GHz....
I simply don't see any compelling reason to buy Bulldozer unless price is the overriding factor....
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,376
the OP talks about OCing the processor - so considering they both OC to roughly the same OC, the difference in clock speed is negligible, so then it is purely down to the extra cores
IBT scores do show a significantly greater score using the 3930k vs 3820 when they are both clocked the same, how that translates in to speed / performance whatever in various applications is down to the fact that those apps aren't yet modified to properly take advantage of the raw power available from the 3930k
you claim that the 3930k is only 33% faster than a 3820 at stock clocks, yet your benchmarks show in one application that it gets 38% faster performance... so when OC'd to roughly the same clocks - and therefore 50% more raw performance is available from the 3930k, it's not unreasonable to assume that there would be greater than 50% improvement between the two
Last edited:
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
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
Last edited:
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,376
the divx encode also shows something like 60% improvement, even at stock clocks
Associate
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. If you're not using the software that gives you the benefit then clearly the 3820 is all that is needed.
For example clock for clock I can render 6 frames of an animation on a 3930K in the same time it takes a 3820 to do 4 frames. Or for every 2 3930K equipped machines I'd need 3 3820 machines, and the £200 saving on the CPU is more than saved. Now keep scaling that up...
There's a place for 3930 and 3960's, but it's a very niche place, and the 3820 is clearly inferior in those circumstances.
Last edited:
I said EXTRA cost, 100% of £225 is £225 so 100% extra would mean £225.