Waiting for Haswell?

[whitewater]

Then you don't need an awesome CPU in 99% of cases. Unless you're one of the 0.01% that actually bothers encoding video/audio AND waiting an extra 30 seconds on a 10 minute operation matters to you, sure. go nuts. Even then most applications like this are actually becoming GPU accelerated, because GPUs are easily 10-20x the raw power of a CPU, so the one thing you care enough about to spend twice as much for that 20% gain would actually be 500% faster if you spent that extra money on a GPU.

And by don't need I mean there would literally be no discernible difference by having a slightly slower CPU, but there would be a huge performance difference by having that GPU. It's basically like buying a Ferrari and not taking it to the race track, it's just something to show off, not to actually use. But what is worse is that you want to travel from New York to Brazil, and instead of taking the plane which is much quicker, cheaper and practical, you drive all the way there, then when you get there continue showing off how awesome your Ferrari is. Meanwhile the sensible person who did the same trip has a normal car, but he flew, he has been there a week already, hired a car and had a much better experience.

 

[Raghar]

I really don't understand why people pair awesome CPUs with shitty GPUs. The only time CPUs are remotely stressed nowadays is basically just in games, and that's only IF you have an awesome GPU to pair it with. Really...why spend twice as much on a CPU that's twice as powerful when only ~1/4 of it will ever be used in even the least GPU bound games around...people should buy a balanced machine for the same price that runs games at 5x the framerate at the same settings..

I never understood why are people willing to pay for a high end GFX card, and use a slow few core CPU. And I have more than 10 000 hours behind me as game developer. I never had enough CPU time for AI, even when I'd ran it on three threads, and the rest of the game on the remaining thread. (And that's before you'd do something as computationally intensive as molecular matter, and other funny stuff.)

Majority of my stuff was framerate decoupled. Thus 200 FPS will not prevent game to run slow on a slow CPU. The engine would simply interpolate next frame from existing data.

Considering I tend to model the rest of the world, not only player's location, 60 - 200 cores sound something that could reasonably work.

BTW have you heard about CPU intensive games like games emulated on PCSX2 emulator, HOI3, and some other strategies? It's PITA to run them at slow CPUs.

 

[phrozenfayte]

So this was apparently on youtube a few hours ago, then the user got banned.

 

[wadec22]

So this was apparently on youtube a few hours ago, then the user got banned.

4770k at 8ghz

if these things actually hit 6-8ghz clocks every member on [H] will get one. with the focus on the on die graphics, I just don't see it happening.

 

[whitewater]

Emulators are a pain, but even on high end i7 you can't play ps2 games for shit, so it doesn't matter. It's much rare to find high end GFX with slow CPU than the other way around.

Of course modern engines decouple things quite well from the framerate, Updating the physics world can be a pain on the CPU with complex geometries, AI really depends on the problem you're trying to solve, generally you can always write more efficient (but slightly less accurate) AI if you're having trouble, the fact that 3/4 of your games CPU resources is AI and even that is starved is....worrying to say the least about your algorithms.

Really the difference between 4 cores 5 years ago and 4 cores now isn't that huge though, 100% gain at best for CPUs vs 1000% gain for GPUs. The real big differences for CPUs will happen when intel goes to 8 core, but beyond that isn't going to happen for 5yrs in consumer land unfortunately.

 

[drescherjm]

I really don't understand why people pair awesome CPUs with shitty GPUs. The only time CPUs are remotely stressed nowadays is basically just in games, and that's only IF you have an awesome GPU to pair it with. Really...why spend twice as much on a CPU that's twice as powerful when only ~1/4 of it will ever be used in even the least GPU bound games around...people should buy a balanced machine for the same price that runs games at 5x the framerate at the same settings..

Depends on what you use your PC for. I routinely stress my 6 core 12 threaded processor at 100% CPU usage for days on end doing my research while my fanless nVidia GTX 220 GPU is rarely ever stressed. I know what I do is not a common task for many users however if you have an HTPC and encode video you probably do not need a beefy GPU either but may need a beefy CPU.

 

[teiresias]

Emulators are a pain, but even on high end i7 you can't play ps2 games for shit, so it doesn't matter. It's much rare to find high end GFX with slow CPU than the other way around.

Not to go completely off topic, but playing PS2 games at 1080p+ on PCSX2 has been feasible at 60fps+ framerates for more than a few years, so this statement is hardly true at all.

 

[PGHammer]

Not to go completely off topic, but playing PS2 games at 1080p+ on PCSX2 has been feasible at 60fps+ framerates for more than a few years, so this statement is hardly true at all.

True. In fact, you can hit 60+ fps (at 720p) using (surprisingly) Kentsfield (Q6600) - the original Intel quad core - with background applications running and the CPU itself bone-stock. I have, in fact, done so. (And Q6600 has been EOL how long?)

Still, such CPU usage remains an outlier. Even desktop virtualization remains largely an outlier - despite better support for it, in both software AND hardware terms. (The major driver for BridgeWalker is, in point of fact, desktop virtualization, not gaming so much.)

 

[Araxie]

if these things actually hit 6-8ghz clocks every member on [H] will get one. with the focus on the on die graphics, I just don't see it happening.

In fact, 6GHz will be the minium OC for most OC'er Users, and not only with a 80x multiplier, i think the main diference as i stated several times on this same topic comes with the 3 bases of blck, 100mhz stock, 125mhz ans 167mhz. That only affect the processor and not the pci and dmi lines, free upgrade without impact the voltage, that as the image posted at 2.29VID the last news about haswell said it will perform optimum even at 2.3v stable.. So will be mucho easier to reach 6-8ghz without touch the max multiplier and voltages...

 

[wadec22]

In fact, 6GHz will be the minium OC for most OC'er Users, and not only with a 80x multiplier, i think the main diference as i stated several times on this same topic comes with the 3 bases of blck, 100mhz stock, 125mhz ans 167mhz. That only affect the processor and not the pci and dmi lines, free upgrade without impact the voltage, that as the image posted at 2.29VID the last news about haswell said it will perform optimum even at 2.3v stable.. So will be mucho easier to reach 6-8ghz without touch the max multiplier and voltages...

hoping you are right, but i'm more than skeptical. we'll know soon enough.

-on a side note, 2.2+v and 6ghz+, I wonder what kind of power draw would we be talking...

 

[Parja]

In fact, 6GHz will be the minium OC for most OC'er Users, and not only with a 80x multiplier, i think the main diference as i stated several times on this same topic comes with the 3 bases of blck, 100mhz stock, 125mhz ans 167mhz. That only affect the processor and not the pci and dmi lines, free upgrade without impact the voltage, that as the image posted at 2.29VID the last news about haswell said it will perform optimum even at 2.3v stable.. So will be mucho easier to reach 6-8ghz without touch the max multiplier and voltages...

Are you high or just insane? People aren't going to be hitting 6GHz clocks regularly. Clocks like that are only going to be achievable with insane voltages and subzero cooling.

 

[NovaCloud]

hmm im really hoping all of these high OC's on these leaked haswell processors turn out true.. would definitely make me wanna build another rig based on haswell... i hear even memory overclocks could now play a factor as well..

 

[Parja]

hmm im really hoping all of these high OC's on these leaked haswell processors turn out true..

I'm sure they are, but not using normal cooling methods and not for every day use. They're going to be LN cooled with HT disabled and are probably just barely stable enough to get CPU-Z to stay running long enough to get a screenshot. Ya know, just like every new processor generation release.

 

[Raghar]

Of course modern engines decouple things quite well from the framerate

I talked about forward interpolation.

AI really depends on the problem you're trying to solve, generally you can always write more efficient (but slightly less accurate) AI if you're having trouble,

Not always. Didn't Einstein said: "Simple, but not too much simple."?

the fact that 3/4 of your games CPU resources is AI and even that is starved is....worrying to say the least about your algorithms.

I do stuff like psycho-modeling and some more complex stuff. But of course, when I have free cores I tend to do... Have you heard about speculative computation? When I have these spare cores, why shouldn't I use them for machine learning... ?

Otherwise, lets say you are simulating country A, and decide to simulate country B. Most of stuff in A and B is completely independent, thus it can use two different cores. In addition, from my experience I found I should calculate a lot of stuff that "isn't exactly needed" when you look at only one simulated behavior, but when you kill majority of these not needed stuff, the result would look bland, and on the other hand when that "unneeded" stuff is calculated, it creates important emergent behavior. (But of course, majority of that stuff is only what I program/develop for my entertainment...)

The real big differences for CPUs will happen when intel goes to 8 core, but beyond that isn't going to happen for 5yrs in consumer land unfortunately.

Intel is killing its market. It concentrates on a single core performance, and spends effort on stuff like boost and automatic voltage increase. I also wonder what happened to all these large L2 caches, which were that important during Q6600 days.

 

[TheOrangeOne]

I highly doubt anyone will be hitting 6ghz on these things with typical air and water coolers, maaaybe, but I doubt it.

I would be perfectly happy with 5ghz, coming from a 3.6 920.

 

[Araxie]

Are you high or just insane? People aren't going to be hitting 6GHz clocks regularly. Clocks like that are only going to be achievable with insane voltages and subzero cooling.

Well you are wrong, using the 167mhz blck base profile, does not need any voltage bump, does not even touch the pcie lines, or dmi lines, all keep at 100mhz while processor blck remain at 167 (without the 5-7% extra margin as sandy/ivy was possible to jump in the blck) so if you have 167mhz with the stock 35 multiplier 167mhzx35=5.8ghz instant jump.. WITHOUT do anything more than increase the blck base.. (That mean in a lower increase on the heat... Cuz the chip are under stock multiplier and voltages... Most people will like to use the 167mhz blck base with a 40 multiplier for a 6.6ghz or increase the blck to 170mhz and same 40 multiplier to get 6.8ghz. Its not anything insane... That test was just stated by intel.. Nothing crazy... They said that even using 125mhz blck base allow for even higher stability on increased voltage for enthusiast ppl..

 

[Nauseous]

i plan on making the jump... see sig i expect to feel a differnce... probably wont.

 

[TheOrangeOne]

i plan on making the jump... see sig i expect to feel a differnce... probably wont.

I'm pretty sure you will see a massive difference.

 

[Tsumi]

Well you are wrong, using the 167mhz blck base profile, does not need any voltage bump, does not even touch the pcie lines, or dmi lines, all keep at 100mhz while processor blck remain at 167 (without the 5-7% extra margin as sandy/ivy was possible to jump in the blck) so if you have 167mhz with the stock 35 multiplier 167mhzx35=5.8ghz instant jump.. WITHOUT do anything more than increase the blck base.. (That mean in a lower increase on the heat... Cuz the chip are under stock multiplier and voltages... Most people will like to use the 167mhz blck base with a 40 multiplier for a 6.6ghz or increase the blck to 170mhz and same 40 multiplier to get 6.8ghz. Its not anything insane... That test was just stated by intel.. Nothing crazy... They said that even using 125mhz blck base allow for even higher stability on increased voltage for enthusiast ppl..

/facepalm

Regardless of whether you're overclocking by BCLK or multiplier, the result is still the same. Increased heat output and increased voltage needed to sustain the overclock.

 

[tonyftw]

i plan on making the jump... see sig i expect to feel a differnce... probably wont.

The good ol' athlon. Trust me, you will notice the difference.

 

[acefightr]

I hope this is worth the way.

 

[defaultluser]

I also wonder what happened to all these large L2 caches, which were that important during Q6600 days.

Reason 1: decreasing cache size lowers the access latency. Nehalem reduces the latency from 15 cycles to 10 cycles.

http://www.anandtech.com/show/2658/4

Reason 2: as feature sizes shrink and core voltages get lower, the reliability of reads from 6T SRAM fall precipitously due to increasing Static Noise Margin (SNM). This is due to the unstable nature of the cell: you get access noise when you read/destroy the stored bit in a cell. This means that you can keep your high density 6T cells, but you need to either (1) increase the voltage (bad for low-power parts) or (2) add support circuits (which increases access times). The other option is to go with 8T SRAM cells, which operate at a lower voltage and have no read penalty (because they're stable), but they significantly increase the size of your cache.

http://www.electroiq.com/articles/sst/2013/03/dimensional-scaling-and-the-sram-bit-cell.html

http://www.google.com/url?sa=t&rct=...=5_zb_O9cEQN2hod6ydr3cQ&bvm=bv.45960087,d.dmg

Core 2 had large L2 cache because it was shared by two CPUs, and because it was originally designed for 65nm and higher voltages it could use more dense 6T SRAM cells and not pay a serious performance/power penalty.

But Nehalem was designed for smaller process nodes (45, 32, 22nm) and sub 1v threshold, and power was a primary concern. Thus, Intel optimized the design by using slow 6T for the L3 cache, and fast (but physically larger) 8T for the L1 and L2:

One new thing we learned about Nehalem at IDF was that Intel actually moved to a 8T (8 transistor) SRAM cell design for all of the core cache memory (L1 and L2
caches, but not L3 cache). By moving to a 8T design Intel was able to actually reduce the operating voltage of Nehalem, which in turn reduces power consumption.
You may remember that Intel’s Atom team did something similar with its L1 cache

http://www.google.com/url?sa=t&rct=...1f-mjsALhgr-3EQ&bvm=bv.45960087,d.dmQ&cad=rja

This balanced design means fast L1 and L2, and the small size of these caches means 8T doesn't seriously impact die space (L3 is high-density 6T). Of course this has the downside of making the L3 cache very slow (39 cycles), but since the L2 is both FASTER and NOT SHARED it's enough to make up for the L3.

 

[Blackbird]

In fact, 6GHz will be the minium OC for most OC'er Users, and not only with a 80x multiplier, i think the main diference as i stated several times on this same topic comes with the 3 bases of blck, 100mhz stock, 125mhz ans 167mhz. That only affect the processor and not the pci and dmi lines, free upgrade without impact the voltage, that as the image posted at 2.29VID the last news about haswell said it will perform optimum even at 2.3v stable.. So will be mucho easier to reach 6-8ghz without touch the max multiplier and voltages...

Well you are wrong, using the 167mhz blck base profile, does not need any voltage bump, does not even touch the pcie lines, or dmi lines, all keep at 100mhz while processor blck remain at 167 (without the 5-7% extra margin as sandy/ivy was possible to jump in the blck) so if you have 167mhz with the stock 35 multiplier 167mhzx35=5.8ghz instant jump.. WITHOUT do anything more than increase the blck base.. (That mean in a lower increase on the heat... Cuz the chip are under stock multiplier and voltages... Most people will like to use the 167mhz blck base with a 40 multiplier for a 6.6ghz or increase the blck to 170mhz and same 40 multiplier to get 6.8ghz. Its not anything insane... That test was just stated by intel.. Nothing crazy... They said that even using 125mhz blck base allow for even higher stability on increased voltage for enthusiast ppl..

lol, seriously.. what is wrong with you?

A. you are insane: seek professional help
B. you are trolling, can a moderator please do something.

I think it's B.

 

[Fahrenheit]

While my i7950 is still good for most of what I do my mobo is starting to give me some issues with some ports being flakey. Also the 3rd part USB3 and Sata3 never ever worked right on my mobo.


Wish we had more info on the motherboards. Dunno witch way to go. Gonna need a new PSU and RAM as well at some point.


Nothing is gonna change with ram right? Gonna start looking for deals on a 16gb kit now.

 

[KENNYB]

While my i7950 is still good for most of what I do my mobo is starting to give me some issues with some ports being flakey. Also the 3rd part USB3 and Sata3 never ever worked right on my mobo.

How are the ports being flakey? I think the 2nd SATA 6Gb port on my motherboard is killing drives. Haswell will finally bring motherboards that bring more than 2 Intel SATA 6Gb ports.

 

[Rendon]

It isn't uncommon to see people these days playing less gpu intensive games like Starcraft 2 while also streaming in 1080p with good audio quality. A decent internet connection and an i7 (possibly socket 2011) is more important than GPU in that use case.

 

[Raghar]

Reason 1: decreasing cache size lowers the access latency. Nehalem reduces the latency from 15 cycles to 10 cycles.

Well I have 3 cycles L1, and 17 cycles (3MB) L3 equivalent, which is quite a difference when compared to 39 cycles L3 on i7-3770K.

Reason 2: as feature sizes shrink and core voltages get lower, the reliability of reads from 6T SRAM fall precipitously due to increasing Static Noise Margin (SNM). This is due to the unstable nature of the cell: you get access noise when you read/destroy the stored bit in a cell. This means that you can keep your high density 6T cells, but you need to either (1) increase the voltage (bad for low-power parts) or (2) add support circuits (which increases access times). The other option is to go with 8T SRAM cells, which operate at a lower voltage and have no read penalty (because they're stable), but they significantly increase the size of your cache.

Now this is interesting. Considering current CPUs are power limited, nobody wants a heater instead of CPU, logic says when 8 T has lower power consumption per cell for negligible 1/3 size increase, CPUs should use 8T cell for cache. Considering 22 nm had problem with area for power dissipation, 8 T would help a lot.

Core 2 had large L2 cache because it was shared by two CPUs, and because it was originally designed for 65nm and higher voltages it could use more dense 6T SRAM cells and not pay a serious performance/power penalty.

Well, rumors were: "Intel was unable to design a reliable on die memory controller, thus it was forced to use large caches to compensate". When Intel started to use on-die memory controller, these large caches disappeared overnight. My Wolfdale-M3 has the same cache per core as i5-3xxx and the cache has lower latency.

I wonder if Haswell didn't switch to a pure 8T cache. Also I wonder about Ivy Bridge-E, they might add 8T without any penalty.

 

[Araxie]

lol, seriously.. what is wrong with you?

A. you are insane: seek professional help
B. you are trolling, can a moderator please do something.

I think it's B.

Of course my friend in just crazy i need really a professional help..

I love to troll... Just see this article: http://www.xbitlabs.com/news/cpu/di...0K_Haswell_Overclocked_to_7_01GHz_Report.html

 

[Tsumi]

Of course my friend in just crazy i need really a professional help..

I love to troll... Just see this article: http://www.xbitlabs.com/news/cpu/di...0K_Haswell_Overclocked_to_7_01GHz_Report.html

Where in that article does it say overclocking through the straps does not need raising of the vcore?

/picard facepalm

 

[PGHammer]

Well I have 3 cycles L1, and 17 cycles (3MB) L3 equivalent, which is quite a difference when compared to 39 cycles L3 on i7-3770K.


Now this is interesting. Considering current CPUs are power limited, nobody wants a heater instead of CPU, logic says when 8 T has lower power consumption per cell for negligible 1/3 size increase, CPUs should use 8T cell for cache. Considering 22 nm had problem with area for power dissipation, 8 T would help a lot.


Well, rumors were: "Intel was unable to design a reliable on die memory controller, thus it was forced to use large caches to compensate". When Intel started to use on-die memory controller, these large caches disappeared overnight. My Wolfdale-M3 has the same cache per core as i5-3xxx and the cache has lower latency.

I wonder if Haswell didn't switch to a pure 8T cache. Also I wonder about Ivy Bridge-E, they might add 8T without any penalty.

Kentsfield and Lynnfield (both dual-dual) may well be the last outsize-cache Intel CPUs for a while - both have larger on-CPU caches than Sandy Bridge or Ivy Bridge, let alone Haswell - and pretty much the ONLY use for the outsize cache (in either Kentsfield OR Lynnfield) is operating-system and application installations; for some rather weird reason, applications themselves, or even general OS tasks, are not generally CPU cache pigs.

However, what I would REALLY like to see are more 2+-core-aware applications - especially since the quad-core CPU has laid firm claim to the midrange, if not the mainstream, of x86.

 

[TheFlosser]

Been waiting for a Haswell laptop for a while but my interest has actually went over into a Richland platform as of late.

 

[defaultluser]

Well, rumors were: "Intel was unable to design a reliable on die memory controller, thus it was forced to use large caches to compensate".

This is true, the trend started with the ridiculous 2MB cache Prescott (made absolutely no difference), and continues through Core 2 (4 and then 6 per-pair). Most simple gaming loads did not see a benefit because most cache-heavy operations are tight main program loops, and most heavy processing loads (like video) do tons of processing on each piece of data, which means fetch can keep-up and cache misses are rare. This is why the Athlon 64 could get away with a paltry 512K L2 and just rely on the fast memory controller.

I believe the biggest purpose of those huge caches on the Core 2 were to prevent the core from falling like a rock when presented with heavily-threaded programs. When you already have cache snoop overhead losses when you bolt four cores together, the last thing you wanted was multiple threads thrashing your tiny L2.

I wonder if Haswell didn't switch to a pure 8T cache. Also I wonder about Ivy Bridge-E, they might add 8T without any penalty.

I have read no indications that Haswell has changed to 8T L3 cache. Given the fact that tweaks have reduced the latency to ~30 cycles, I'm not sure how much more there is to gain.

I would think that Intel would rather drop 5-10% extra die size for more GPU units than for a negligible performance increase.

 

[brennok]

I am leaning towards it. My main computer is still an i7 860, but my living room is running the older i5-661.

These days if I do game it is in the living room. My bedroom does a lot more processing based tasks such as ripping, transcoding, and file conversions.

Since I don't game that much the i7 will move out to the living room, and the new PC will go in the bedroom.

 

[/usr/sbin]

Well you are wrong, using the 167mhz blck base profile, does not need any voltage bump, does not even touch the pcie lines, or dmi lines, all keep at 100mhz while processor blck remain at 167 (without the 5-7% extra margin as sandy/ivy was possible to jump in the blck) so if you have 167mhz with the stock 35 multiplier 167mhzx35=5.8ghz instant jump.. WITHOUT do anything more than increase the blck base.. (That mean in a lower increase on the heat... Cuz the chip are under stock multiplier and voltages... Most people will like to use the 167mhz blck base with a 40 multiplier for a 6.6ghz or increase the blck to 170mhz and same 40 multiplier to get 6.8ghz. Its not anything insane... That test was just stated by intel.. Nothing crazy... They said that even using 125mhz blck base allow for even higher stability on increased voltage for enthusiast ppl..

Not sure if serious.

 

[Parja]

Well you are wrong, using the 167mhz blck base profile, does not need any voltage bump, does not even touch the pcie lines, or dmi lines, all keep at 100mhz while processor blck remain at 167 (without the 5-7% extra margin as sandy/ivy was possible to jump in the blck) so if you have 167mhz with the stock 35 multiplier 167mhzx35=5.8ghz instant jump.. WITHOUT do anything more than increase the blck base.. (That mean in a lower increase on the heat... Cuz the chip are under stock multiplier and voltages... Most people will like to use the 167mhz blck base with a 40 multiplier for a 6.6ghz or increase the blck to 170mhz and same 40 multiplier to get 6.8ghz. Its not anything insane... That test was just stated by intel.. Nothing crazy... They said that even using 125mhz blck base allow for even higher stability on increased voltage for enthusiast ppl..

Nowhere in the article you linked after this post does it allude to your being able to just magically bump your CPU up to 6.8GHz without a large increase in CPU voltage. How are you making the (absurd) connection that you can just bump the BCLK up to overclock the CPU 67% without needing to increase the CPU voltage?

 

[lajgnd]

I was waiting for Haswell, but now that 6 core processors are starting to show benefits in modern gaming, I might as well jump on board the 3930K since the Ivy or even Haswell 6 core won't be out for months.

 

[mellojoe]

True, but 1155 is a dead socket. I'd rather wait, buy 1150, then upgrade that if need be (and if cheap enough) when people dump their 'tick' processors for skylake.

I'm using your thought, but its been said many times, and I've always wanted to respond to this:

If you are in the 3-year, 4-year, or 6-year upgrade cyle that so many people are, the dream of buying a platform and then upgrading that platform at a later date is just that ... a dream. Most people buy one time, and then by the time they are ready to upgrade, their platform has been obsoleted by 2 or 3 generations. Sure, they can "throw money away" and buy a final run processor from a dead technology, but finding those cheaply is worse than trying to upgrade the whole system.

Look at me, for instance. On a P35 motherboard. Do you know how hard it is to find Socket LGA775 cpu's? So, at this point, bying a "high-end" LGA775 is really just a non-issue. I love the idea of buying a platform and then just upgrading the CPU at some point down the road. But, I have rarely ever seen that work out for the end user. The only time was back in the 500mhz days, when you could go from 450mhz to 950mhz on the same platform, and that made a big enough difference to warrant the upgrade. Now, the architectures are so similar, and the performance gains are so minimal, that it just doesn't seem to pay to buy a 2nd chip from the same platform.

For power users, they aren't on long upgrade cycles. Sometimes they'll be upgrading multiple times, and for them jumping from one platform to another isn't a bad thing. They're opening their boxes multiple times a year anyway, so plopping in a new mobo isn't a worry. But for many, they like to THINK they will adopt a future-proof platform, but it just doesn't exist. In 3 or 4 years, everything will change anyway. So buy today what makes you happy. Me? I'll wait for Haswell, only because I'm in a position where I CAN wait. For others, buy an IvyBridge and call it a day.

 

[Blackbird]

but now that 6 core processors are starting to show benefits in modern gaming

What games?

 

[socK]

Crysis 3 scales very well

 

[Araxie]

Crysis 3 scales very well

Yes but no more than that... The only one and i think thats for many reasons.. One of that the hell badly optimized... Requiring more brute power than the normal situations cuz IA and physics are just horrible...