3930K to 5930K

[Carfax83]

Is anyone else with a 3930K setup looking to make the jump to Haswell-e later this year?

I'm seriously thinking about it. Sandy Bridge core still has plenty of life left in it, but the Haswell core is just much better imo. The only thing that ticks me off is that I will have to buy some new DDR4 memory, which will undoubtedly use higher timings yet run at the same speeds as current high end DDR3 modules but cost a lot more... It will be sometime before we see DDR4 will ramp up to speeds that will justify the price premium over DDR3.

As a platform though, x99 looks like it's going to be very strong..

 

[ShintaiDK]

If you have to ask, then no.

 

[BrightCandle]

No. The 8 core is too expensive to be worth the cost and the 6 core is going to be only marginally faster for a lot of cost of upgrade. Its ridiculous that a CPU this old is still as close to the top speed as it is but its the new world of almost zero improvement in CPUs.

Until we see the reviews we can't be certain but its unlikely its going to be substantially better than SB -> Haswell which isn't exactly a big upgrade either.

 

[Carfax83]

No. The 8 core is too expensive to be worth the cost and the 6 core is going to be only marginally faster for a lot of cost of upgrade. Its ridiculous that a CPU this old is still as close to the top speed as it is but its the new world of almost zero improvement in CPUs.

The 8 core option seems to be targeted to those entry level workstation guys who need more CPU power, but are unwilling to shell out for the Xeons. Kind of like NVidia's Titan line..

I don't think Haswell is a marginal upgrade over Sandy Bridge either. Hulk listed some clock normalized benchmarks last year comparing core performance between Nehalem, Sandy Bridge and Haswell and the performance increase from Sandy Bridge to Haswell is actually greater than Nehalem to Sandy Bridge.

And that's not even taking into account AVX2. We always knew Haswell was going to pick up steam as software became more optimized for AVX2, and that takes time.

It's just too bad that the x99 platform is using DDR4, and that the introductory speeds will be so low. DDR4 would start at the absolute minimum clock speed of 2600 if it were up to me.

 

[frozentundra123456]

The problem with that comparison is that ipc is still increasing, but clockspeed and overclocking headroom are not increasing, and overclocking becoming worse. So the total gain from tock to tock is clearly less.

 

[ruhtraeel]

Depends on what you're using it for, I would say. If the difference in performance is similar to this:
http://www.ocaholic.ch/modules/smartsection/item.php?itemid=1158&page=13

Then I would say it is silly and an absolute waste of money. In real world scenarios, I would say an average gain of 2-3% at the expense of switching processors and chipsets is definitely not worth it. Might be different on the >4 core side, I dunno.

 

[Arachnotronic]

I'd only upgrade if I needed a substantially better platform.

I'm happy with my X79, so I personally wouldn't do that upgrade.

 

[crashtech]

Is anyone else with a 3930K setup looking to make the jump to Haswell-e later this year?

I'm seriously thinking about it. Sandy Bridge core still has plenty of life left in it, but the Haswell core is just much better imo. The only thing that ticks me off is that I will have to buy some new DDR4 memory, which will undoubtedly use higher timings yet run at the same speeds as current high end DDR3 modules but cost a lot more... It will be sometime before we see DDR4 will ramp up to speeds that will justify the price premium over DDR3.

As a platform though, x99 looks like it's going to be very strong..

Are you viewing this solely from a performance increase aspect, or do you sometimes think of upgrading just to gain experience with new features and have something different to tinker with?

I fall into the latter group, so from my perspective, if you are asking, then the answer is "yes," and you don't need validation to do so, all you need is disposable income.

 

[wabbitslayer]

I'd only upgrade if I needed a substantially larger e-penis.

 

[RussianSensation]

With your CPU, I would rather upgrade the GPUs and wait until Skylake-E/Canonlake-e for a platform upgrade. At least those will be on 14/10nm and should have another 10% IPC increase and a shot of hitting 5Ghz on air. Given that 4790K generally hits 4.7-4.8, I don't see 5920/5930 hitting 5Ghz.

 

[Homeles]

If you have to ask, then no.

I'm going to have to agree with you on this one.

Is anyone else with a 3930K setup looking to make the jump to Haswell-e later this year?

I'm seriously thinking about it. Sandy Bridge core still has plenty of life left in it, but the Haswell core is just much better imo. The only thing that ticks me off is that I will have to buy some new DDR4 memory, which will undoubtedly use higher timings yet run at the same speeds as current high end DDR3 modules but cost a lot more... It will be sometime before we see DDR4 will ramp up to speeds that will justify the price premium over DDR3.

As a platform though, x99 looks like it's going to be very strong..

The HEDT platform isn't something you should really be using unless you have a legitimate reason. If you just need a fast computer, mainstream platforms are the only thing you should be looking at. Otherwise, you're just throwing money down the drain.

The more endowed LGA2011 processors are well worth their added cost, but only if you have something that truly will leverage the multithreaded performance, or other features offered by the platform. They really should be viewed as workstations. Since they should be viewed as such, you should be weighing the boost in productivity to your needs against the the cost of upgrading. If you don't use your computer to earn a living -- that is, if you don't use your computer as a workstation -- you have no business being on LGA2011. Even then, you should consider whether or not LGA1150 would suit your needs.

If you're just interested in LGA2011 because of a desire to fill some void in you... well... honestly you should convince yourself thoroughly that you don't need it, otherwise that itch is never going to go away. If money doesn't matter, and your part of your hobby is to have the most expensive machine feasible... then you shouldn't be asking us.

I suppose there are worse vices. Go ahead and buy it, if it'll make you happy, but you might want to kick your smoking habit or whatever first.

Keep in mind that the 5930K and other Haswell-E processors are on a different LGA2011 socket. You'd need a new, expensive board as well.

 

[BrightCandle]

The more endowed LGA2011 processors are well worth their added cost, but only if you have something that truly will leverage the multithreaded performance, or other features offered by the platform. They really should be viewed as workstations. Since they should be viewed as such, you should be weighing the boost in productivity to your needs against the the cost of upgrading. If you don't use your computer to earn a living -- that is, if you don't use your computer as a workstation -- you have no business being on LGA2011. Even then, you should consider whether or not LGA1150 would suit your needs.

I don't fully agree with this. For all the accumulated data from gamegpu.ru on the impact of CPUs its pretty apparent the 6 core CPUs are faster in certain games and sometimes quite significantly. Even now the 3930k averages even with a 4770k, a CPU 2 generations later and its improved by those additional cores. Its only about 10% of games but many of those games are the big releases of the year, the battlefield 4's the Crysis' etc. Those games do benefit, sometimes up to 50% over and above the normal mainstream processor. The data says they aren't just for workstations, for high end gaming rigs they are beneficial to, just not always or even often.

 

[guskline]

No. Happy as a clam with my WC'd 3930k @ 4.6ghz.

 

[Carfax83]

Are you viewing this solely from a performance increase aspect, or do you sometimes think of upgrading just to gain experience with new features and have something different to tinker with?

All three really. As I mentioned earlier, Haswell constitutes a bigger upgrade in core performance over Sandy Bridge, than Sandy Bridge did over Nehalem. And the performance delta is going to widen as AVX2 adoption increases and becomes more prevalent..

Have any games started using AVX2 yet?

 

[Carfax83]

With your CPU, I would rather upgrade the GPUs and wait until Skylake-E/Canonlake-e for a platform upgrade. At least those will be on 14/10nm and should have another 10% IPC increase and a shot of hitting 5Ghz on air. Given that 4790K generally hits 4.7-4.8, I don't see 5920/5930 hitting 5Ghz.

Broadwell-E will be on 14nm as well. Skylake-E is too far down the road to wait that long, at least for me.

It's either I upgrade to Haswell-E, or wait for Broadwell-E late next year.

 

[Carfax83]

If you don't use your computer to earn a living -- that is, if you don't use your computer as a workstation -- you have no business being on LGA2011. Even then, you should consider whether or not LGA1150 would suit your needs.

I disagree with this. Before I upgraded to my 3930K setup, I predicted that with the coming of the PS4 and Xbox One, both of which use 8 core CPUs, that games would become more multithreaded by necessity as the individual cores on the PS4 and Xbox One are low speed and have relatively weak IPC compared to desktop CPUs.

And it's definitely happening, as ALL the next gen 3D engines use multithreaded renderers and scale to at least 6 threads or more.. The 3930K/3960x is faster than the 4770K in the latest games due to the extra cores, despite having a less advanced core microarchitecture.

So hex core CPUs are worth the investment as a high end gaming CPU, especially if you are running with multi GPUs like I am, and shouldn't be confined to just workstation type applications.

They are very relevant for high end gaming PCs as well.

 

[crashtech]

It's pretty early to say, but there should be little doubt that Haswell-E is going to be the dominant CPU until late 2015, unless we are expecting some magic with Broadwell-K.

 

[Arachnotronic]

I'm going to have to agree with you on this one.
The more endowed LGA2011 processors are well worth their added cost, but only if you have something that truly will leverage the multithreaded performance, or other features offered by the platform. They really should be viewed as workstations. Since they should be viewed as such, you should be weighing the boost in productivity to your needs against the the cost of upgrading. If you don't use your computer to earn a living -- that is, if you don't use your computer as a workstation -- you have no business being on LGA2011. Even then, you should consider whether or not LGA1150 would suit your needs.

Honestly, I don't like that Intel makes it even a choice between both platforms. HEDT should be unequivocally better than mainstream, but for singlethreaded performance out of the box, LGA 1150 + 4790K is a superior choice to any LGA 2011 part.

 

[Grooveriding]

Maybe 15-20% more performance at the same clocks, but it will almost surely not overclock as well as SB-E does.

I'm getting an 8 core just because it's fun. GPUs have been incredibly boring since Titan came out and then Hawaii and it looks like it's going to stay that way with the next GPUs coming out on 28nm again. HW-E is the most exciting release CPU wise since SB-E.

You'd probably get a lot more out of going to dual 780ti from your current cards though. The cost would work out about the same.

 

[RussianSensation]

Broadwell-E will be on 14nm as well. Skylake-E is too far down the road to wait that long, at least for me.

It's either I upgrade to Haswell-E, or wait for Broadwell-E late next year.

Ya you sound like you just want to upgrade for the sake of upgrading. The point you made that the move from Conroe to Nehalem is less than SB to HW is incorrect. When did AT do a test of Q9550@ 4.0 ghz vs. i7 920 @ 4.0ghz vs. 2600k @ 4.0 vs. 4770k @ 4.0? They never did. You can look at games like FC3 where 3770k 4.5 Ghz OC can't even beat i7 965 3.75ghz OC:

http://www.guru3d.com/articles_pages/far_cry_3_graphics_performance_review_benchmark,7.html

In most games, the move from SB to HW made almost no performance increase once both are overclocked to the max. You will be 99% GPU limited with your 770s. In cases where CPU speed might help, it will be in 1-2 threaded games such as skyrim or Blizzard games where performance is through the roof as it is. A lot of sites do testing when they remove MSAA or Ultra settings when testing CPUs which is pretty much worthless and creates incorrect information for IPC scaling in games. These types of tests are only useful for people who sacrifice IQ for 120-144hz gaming. For the rest, you keep cranking visuals, even as far as enabling super-sampling or at least 4xSMAA/4xMSAA. People with $1k GPUs don't generally want to lower in-game settings which means the CPU bottleneck is usually hidden.

Let's say your 5930 hits 4.7ghz, 12% IPC will translate as 5-6% at best on average since most games are still GPU limited. That amounts to an equivalent SB speed of 4.98Ghz or just 13% faster over yours. A pointless upgrade for games. You are better off getting 3 fastest Maxwell GPUs. I can't think of any modern games from now that will run faster on i7 5930 @4.7ghz + 2x flagship Maxwell over 3930@4.4ghz + 3 flagship Maxwell.

And well since you don't seem to buy GPUs that cost more than $1k a pair, you will be FAR better off spending the money on 2 flagship maxwells, meaning you should spend everything on GPUs before even considering a CPU upgrade if all you are doing it for is gaming. This is why waiting for Skylake-E or Canonlake-E makes logical sense.

Unless you have absolutely a very fast GPU setup (Tri-290s or at least 780Ti SLI), there will be hardly much difference in performance between an I7 920@4.2ghz and 4790k@4.8ghz. This point is even stronger once you go 1440p or higher and add SMAA or MSAA.

If 880 is 30% faster than 780Ti, then it will be faster by nearly 80% vs. a single 770. Also, those 770 4GBs still have some value in them, and should be sold once 880s come out before the market realizes that a 770 4GB is not worth more than $250.

 

[escrow4]

What games actually use a hexa core that isn't due to poor optimization? An i7 4770 can drive 60FPS in the vast bulk of games with a single GPU. We are not at the point where a 4770 say hits 40FPS, and a 5930K hits over 60FPS. The gains are minimal to irrelevant from quad to hexa.

 

[cytg111]

if I was to make that jump it'd be 8 cores or bust. Even then I'd have an eye on skylakes ST performance. No, wouldnt jump it.

 

[Carfax83]

Maybe 15-20% more performance at the same clocks, but it will almost surely not overclock as well as SB-E does.

And as more software becomes AVX2 optimized, that number will grow. Aren't you concerned about having to buy new memory though?

The DDR4 will cost an arm and a leg upon release, and it will run slower than comparable DDR3 until they start ramping up the clock speeds.

I'm getting an 8 core just because it's fun. GPUs have been incredibly boring since Titan came out and then Hawaii and it looks like it's going to stay that way with the next GPUs coming out on 28nm [g]again[/b]. HW-E is the most exciting release CPU wise since SB-E.

To me, tweaking the CPU and RAM performance was always more fun than messing with the GPU..

You'd probably get a lot more out of going to dual 780ti from your current cards though. The cost would work out about the same.

I usually don't upgrade my GPUs until they can no longer deliver high performance and IQ that I expect in games. So far, these GTX 770 4GBs have held out very well in that respect @ 1440p and I don't expect any games on the horizon to change that.

The next games I will purchase this year are Dragon Age Inquisition and Assassin's Creed Unity, and neither game seems like it would be problematic.

I'm actually hoping these GTX 770 4GBs will last until 22nm GPUs become available, or until I get a 4K monitor..

 

[Grooveriding]

The DDR4 prices are not going to be the complete wallet buster. My thought from past releases is perhaps a 30% premium over DDR3 pricing.

You'll get plenty of bandwidth and the same speeds you have on DDR3 right now, likely just looser timings. RAM performance is not crucial once you're beyond a certain performance level and you'll be past that on DDR4.

It's hard to know how much more pervasive the benefit of lots of cores in gaming will continue to be. We're starting to see more and more games that are making use of it, but it's not going to be every new game that comes out. I feel the enthusiast platform is superior and really lasts a lot longer than the mainstream platform does - if - you get onto the new platform on release day.

I don't really over-think it too much. It's fun to get some new stuff, and CPU wise, the -E platforms are awesome. If you're trying to find some vast and stark value metric in X99 over mainstream, it's going to be tough, but there is just the fact that it's the coolest platform if you're an enthusiast about PC hardware.

 

[Carfax83]

Ya you sound like you just want to upgrade for the sake of upgrading. The point you made that the move from Conroe to Nehalem is less than SB to HW is incorrect. When did AT do a test of Q9550@ 4.0 ghz vs. i7 920 @ 4.0ghz vs. 2600k @ 4.0 vs. 4770k @ 4.0? They never did. You can look at games like FC3 where 3770k 4.5 Ghz OC can't even beat i7 965 3.75ghz OC:

I didn't say anything about the move from Conroe to Nehalem. I said that the jump from Sandy Bridge to Haswell was bigger than the jump from Nehalem to Sandy Bridge..

Conroe (or Penryn/Yorkshire to be more accurate) to Nehalem was a big jump, as the latter put everything on die and gave you SMT.

In most games, the move from SB to HW made almost no performance increase once both are overclocked to the max. You will be 99% GPU limited with your 770s. In cases where CPU speed might help, it will be in 1-2 threaded games such as skyrim or Blizzard games where performance is through the roof as it is. A lot of sites do testing when they remove MSAA or Ultra settings when testing CPUs which is pretty much worthless and creates incorrect information for IPC scaling in games. These types of tests are only useful for people who sacrifice IQ for 120-144hz gaming. For the rest, you keep cranking visuals, even as far as enabling super-sampling or at least 4xSMAA/4xMSAA. People with $1k GPUs don't generally want to lower in-game settings which means the CPU bottleneck is usually hidden.

This is a very salient point. If I stayed on the GTX 770 4GBs, I would be GPU bottlenecked regardless.. As I mentioned previously, I don't really have any intention of upgrading my GPUs unless they become an obstacle to getting the frame rates and IQ that I want, and so far, I don't see any game on the horizon that can cause this to happen..

And well since you don't seem to buy GPUs that cost more than $1k a pair, you will be FAR better off spending the money on 2 flagship maxwells, meaning you should spend everything on GPUs before even considering a CPU upgrade if all you are doing it for is gaming. This is why waiting for Skylake-E or Canonlake-E makes logical sense.

I would drop 1K on GPUs if I thought they were worth the money. The GTX 780s last year to me were not worth the 1200 dollars I would have paid for them if I had bought them as the performance increase over the GTX 770 wasn't big enough.

If 880 is 30% faster than 780Ti, then it will be faster by nearly 80% vs. a single 770. Also, those 770 4GBs still have some value in them, and should be sold once 880s come out before the market realizes that a 770 4GB is not worth more than $250.

I thought the GTX 880 was going to be 256 bit bus width, so I'm pretty sure it won't be 30% faster than the 780 Ti..