Intel CPU Rant

[Chess Gator]

Gaming/Simulations have very little benefit for Intel's quest to just keeping adding more cores and lowering the frequency. Desktop consumers want to see more than 7% gain after 3+ years of CPU "progress" ... adding cores solves NOTHING for games/simulations.

Best stable OC for the 5960X is around 4.3Ghz, the 3960X is 4.8Ghz (under common cooling solution, not extreme ones) ... performance difference is only 7% ... so 3+ years on CPU progress is only producing a 7% gain? Most of that gain is probably related to the chipset X99 and not the CPU.

Consumers want higher frequency less cores, that's what works best for desktop computing, games, simulations. Because this doesn't fit your marketing strategy isn't a justification to NOT provide what consumers really want ... more die space, higher frequency.

I would much rather pay $1000 for a 4 core CPU operating at 6Ghz, than 8 CPUs operating at 4.3 Ghz.

What do you guys think?

CG

 

[myocardia]

Gaming/Simulations have very little benefit for Intel's quest to just keeping adding more cores and lowering the frequency. Desktop consumers want to see more than 7% gain after 3+ years of CPU "progress" ... adding cores solves NOTHING for games/simulations.

Oops, I guess not.

edit: BTW, Tom's Hardware tested five retail 5960Xs, and all five of them did 4.5 Ghz or higher.

 

[NTMBK]

Cool, I'm sure you and the other 10 consumers like you will enjoy your $1000, 300W CPU. But you don't represent a big enough market to make it worth Intel's time.

Intel is focused on mass market laptops, and servers. Both of those markets love power efficient cores. Desktop is just overclocked laptop and server CPUs, has been for years.

 

[Sheep221]

Gaming/Simulations have very little benefit for Intel's quest to just keeping adding more cores and lowering the frequency. Desktop consumers want to see more than 7% gain after 3+ years of CPU "progress" ... adding cores solves NOTHING for games/simulations.

Best stable OC for the 5960X is around 4.3Ghz, the 3960X is 4.8Ghz (under common cooling solution, not extreme ones) ... performance difference is only 7% ... so 3+ years on CPU progress is only producing a 7% gain? Most of that gain is probably related to the chipset X99 and not the CPU.

Consumers want higher frequency less cores, that's what works best for desktop computing, games, simulations. Because this doesn't fit your marketing strategy isn't a justification to NOT provide what consumers really want ... more die space, higher frequency.

I would much rather pay $1000 for a 4 core CPU operating at 6Ghz, than 8 CPUs operating at 4.3 Ghz.

What do you guys think?

CG

In terms of simulations, more cores(or more threads for that matter) are crucial to get more results complete in same time, regardless of the frequency you are running more work units in same time, with higher frequency and less cores you get done some instructions slightly faster but they all must wait in queue until one is done so another can continue. Parallel computing has been key to increasing computing performance since introduction of multicore CPUs and HT which was around 2001 and 2002 respectively(consumer dual cores since 2005). I'm 110% sure that we have moved to faster CPUs since then.
And believe me you wouldn't really want to pay for toasty 6GHz CPU...

 

[tential]

How many of Intel's consumers even understand what you wrote?

In short your post ignores 90% of cpu users.
Congrats for living in a bubble!

Sent from my C6833 using Tapatalk

 

[Dresdenboy]

Gaming/Simulations have very little benefit for Intel's quest to just keeping adding more cores and lowering the frequency. Desktop consumers want to see more than 7% gain after 3+ years of CPU "progress" ... adding cores solves NOTHING for games/simulations.

Do you have examples?

And what kind of simulations do you refer to?

 

[chrisjames61]

How many of Intel's consumers even understand what you wrote?

In short your post ignores 90% of cpu users.
Congrats for living in a bubble!

Sent from my C6833 using Tapatalk

Not many casual users would understand it but that doesn't change the fact he is correct.

 

[Yuriman]

Unfortunately, this is what power consumption does as clockspeed increases:

I would be fascinated to see how high it would actually be at 6GHz.

 

[LPCTech]

Intel is a company whos goal is making money. The money is currently in making small low power processors that can be passively cooled to go in laptops and tablets but still have the performance of an average desktop processor. Very few people want a 6Ghz monster chip. Its cool and everything. But intel knows thats not going to make them much money and also very few people want to pay over 300$ even for a desktop cpu.

So what you really wanna say is that intel's business model is annoying you personally as an overclocking enthusiast because they are not making chips that you can get high overclocks on.

Also, I doubt the "average" consumer wants "fewer, higher clocked cores" the average consumer has no idea how a CPU works and just wants it to be cheap and run all their "apps" without being slow or overheating.

 

[BigDaveX]

What makes this complaint even sillier is that Intel's clockspeeds actually have started to increase over the last couple of years. Between Conroe and the initial versions of Haswell, base clock speeds were firmly planted in the low-to-mid 3GHz area. In the last 18 months, Devil's Canyon and Skylake have pushed that up to 4GHz base.

I suspect Intel could push it even higher if they really needed to, but they're likely waiting to Zen to show up before making any firm decisions.

 

[frozentundra123456]

Not many casual users would understand it but that doesn't change the fact he is correct.

Actually, he isnt. I dont think intel has a strategy of adding cores and lowering frequency at all. In fact, quite the opposite, except in servers. Many are critical of intel for *not* adding more cores to the mainstream products.

 

[PhIlLy ChEeSe]

Gaming/Simulations have very little benefit for Intel's quest to just keeping adding more cores and lowering the frequency. Desktop consumers want to see more than 7% gain after 3+ years of CPU "progress" ... adding cores solves NOTHING for games/simulations.

Best stable OC for the 5960X is around 4.3Ghz, the 3960X is 4.8Ghz (under common cooling solution, not extreme ones) ... performance difference is only 7% ... so 3+ years on CPU progress is only producing a 7% gain? Most of that gain is probably related to the chipset X99 and not the CPU.

Consumers want higher frequency less cores, that's what works best for desktop computing, games, simulations. Because this doesn't fit your marketing strategy isn't a justification to NOT provide what consumers really want ... more die space, higher frequency.

I would much rather pay $1000 for a 4 core CPU operating at 6Ghz, than 8 CPUs operating at 4.3 Ghz.

What do you guys think?

CG

There are many taking a 4 core well past 6Mhz, you just need to know where to look is all.
http://hwbot.org/

I've seen many 5960K'S well past 4.3Mhz. Though Intel will be stopping to TICK TOCK as a pattern to release new CPU'S.

 

[Dresdenboy]

There are many taking a 4 core well past 6Mhz, you just need to know where to look is all.
http://hwbot.org/

I've seen many 5960K'S well past 4.3Mhz. Though Intel will be stopping to TICK TOCK as a pattern to release new CPU'S.

No problem with that. I recently coded some assembler 3D gfx stuff for a 68k CPU, which even runs at 12MHz!

 

[RussianSensation]

Intel is a company whos goal is making money. The money is currently in making small low power processors that can be passively cooled to go in laptops and tablets but still have the performance of an average desktop processor. Very few people want a 6Ghz monster chip. Its cool and everything. But intel knows thats not going to make them much money and also very few people want to pay over 300$ even for a desktop cpu.

So what you really wanna say is that intel's business model is annoying you personally as an overclocking enthusiast because they are not making chips that you can get high overclocks on.

Also, I doubt the "average" consumer wants "fewer, higher clocked cores" the average consumer has no idea how a CPU works and just wants it to be cheap and run all their "apps" without being slow or overheating.

In 2015, Intel set a record for i7s sold and it's only picking up in 2016.

http://www.kitguru.net/components/cpu/anton-shilov/shipments-of-intel-core-i7-processors-set-records-in-q2-company/

Keeps going:
http://www.guru3d.com/news-story/intel-has-record-sales-of-core-i7-and-k-model-processors-in-declining-cpu-market.html

What's happening in reality is the opposite of what's being portrayed on hardware forums: people who want a new desktop, want it to be powerful. Otherwise, there is no point getting a crappy computer when a tablet, smartphone will do. That coupled with the fact that i3/i5 seriously cripples productivity and gaming performance over an i7, and more educated consumers are gravitating towards i7s. Finally, all those users with Core 2 Quad/Nehalem/Sandybehave realized that a solid fast Intel CPU lasts 5-7 years. When it comes time to replace that system, they view $100-120 extra over an i5 as peanuts over 5-7 years to get top-of-the-line performance. But hey, don't believe me, the financials are right there in every quarterly Intel earnings report.

The reason Intel cannot produce a 6Ghz CPU isn't because there is no market demand, but because it's simply not possible to make such a product in the current 140-160W TDP space. Even 14nm mature i7 6700K can hardly overclock much beyond 4.8Ghz, with very few samples going to 4.9-5Ghz. That means, it's simply impossible for Intel to reliably deliver a 5.5-6Ghz quad core processor with Skylake's IPC using today's transistors.

If anything, if Intel actually did, that would literally get 95% of Nehalem > Haswell i7 4790K users on this forum to upgrade. We probably won't even get to the performance of a 6Ghz Skylake core equivalent speed until 2021+. If Intel released this level of performance right now for $350, I would upgrade every single family PC in the household tomorrow.

 

[myocardia]

In 2015, Intel set a record for i7s sold and it's only picking up in 2016.

http://www.kitguru.net/components/c...core-i7-processors-set-records-in-q2-company/

Keeps going:
http://www.guru3d.com/news-story/in...model-processors-in-declining-cpu-market.html

The reason Intel cannot produce a 6Ghz CPU isn't because there is no market demand, but because it's simply not possible to make such a product in the current 140-160W TDP space. Even 14nm mature i7 6700K can hardly overclock much beyond 4.8Ghz, with very few samples going to 4.9-5Ghz. That means, it's simply impossible for Intel to reliably deliver a 5.5-6Ghz quad core processor with Skylake's IPC using today's transistors.

If anything, if Intel actually did, that would literally get 95% of Nehalem > Haswell i7 4790K users on this forum to upgrade. We probably won't even get to the performance of a 6Ghz Skylake core equivalent speed until 2021+. If Intel released this level of performance right now for $350, I would upgrade every single family PC in the household tomorrow.

I concur. I can't imagine getting anywhere near 6 Ghz stock clocks any earlier than 5nm, if even then. Last I heard, Intel was planning on releasing 10nm Cannonlake sometime in 2017, which would give us 7nm in 2020, and 5nm in 2023 at the absolute earliest.

 

[Chess Gator]

Anyone hear of anything exceptional coming out of Intel in future, for example the Teraflops Research Chip? https://en.wikipedia.org/wiki/Teraflops_Research_Chip Or maybe Quantum computing?

 

[myocardia]

Anyone hear of anything exceptional coming out of Intel in future?

It all depends on what you're looking to get. The very next "main" release, ie not server CPUs like the -E series, is supposed to have both hardware encode and decode of HEVC/x.265, the same way that Haswell and Devil's Canyon have hardware encode and decode of H.264.

 

[RussianSensation]

Anyone hear of anything exceptional coming out of Intel in future, for example the Teraflops Research Chip? https://en.wikipedia.org/wiki/Teraflops_Research_Chip Or maybe Quantum computing?

For the next 3 years, there will be nothing revolutionary coming out from Intel CPUs to meet your desired throughout increase.

Broadwell-E in 2016
Skylake-E in 2017
Then we have new architecture in Icelake in 2H 2018, which means the earliest we would see Icelake-E is Q1 2019.

For simplicity's sake, let's just assign Skylake a 25-30% increase in IPC over your chip and assume you overclock it to the max at reasonable voltage of 1.4V to 4.8Ghz. Now think about how long from today it will take to get another 25-30% increase in IPC over this Skylake chip.

Chances are Icelake 2018 will bring the traditional 10-15%. That means a subsequent 10-15% won't come until next true new Intel architecture in 2021 because Intel has shifted to Process-Architecture-Improvement 3-year-cadence. So at this point, you could either just get 6-8 core SKL-E in 2017 and not upgrade for 5 years, or get Icelake and not upgrade for 5 years.

As far as mathematical simulations, financials, research, deep learning, AI simulation/prediction, all of that is moving towards GPGPU. I mean NV bet its entire Pascal architecture and the firm's future with Volta on this. That means if you are truly looking for a revolutionary performance increase, find software that benefits from CUDA or OpenCL and get a flagship compute videocard -- you aren't going to find this revolution on CPUs which are mostly serial in nature. Revolution happens in parallel compute -- GPGPU space.

Another issue is also a matter of software scaling linearly with more CPU cores. Once again, cutting edge CUDA/OpenCL GPGPU software doesn't have this problem either.

On the positive side, given how long your CPU has lasted, you either save $ on upgrades, or you have more $ to spend on other PC parts like a 4K monitor, faster GPU, etc.

 

[SPBHM]

6700K OC is your CPU, a lot more affordable than the 5960X and with a higher gain for CPU bound than 7% compared no any Sandy Bridge OC, very few situations in game will run better with a 5960X OC.

https://www.youtube.com/watch?v=ocwwaVGUFtk

 

[geokilla]

Isn't a 4.4GHz Intel Core i5 or i7 probably equivalent to a 10GHz Pentium 4 in single threaded applications? What is clock speed anyways? Comparing clock speed is futile and that is all I know despite being fairly knowledgeable about computers and what parts to go for.

 

[myocardia]

Isn't a 4.4GHz Intel Core i5 or i7 probably equivalent to a 10GHz Pentium 4 in single threaded applications?

No, not really. When the first C2Ds came out, we did a bit of experimenting with them, versus P4s. The Conroe cores were a few percent shy of 200% higher single-threaded performance, at the same clock speeds...at least in gaming. The P4s were great at things like content creation/video encoding, though.

 

[Exophase]

Intel can't just decide to take a Skylake-like design and clock it at 6GHz, no matter how much they're willing to sacrifice power consumption. Not unless it came with an LN2 cooling setup. There are limits to how high things can be clocked, which is why they've been putting so much effort into extracting more performance per clock cycle, even though that too is hitting diminishing returns.

 

[RussianSensation]

Isn't a 4.4GHz Intel Core i5 or i7 probably equivalent to a 10GHz Pentium 4 in single threaded applications? What is clock speed anyways? Comparing clock speed is futile and that is all I know despite being fairly knowledgeable about computers and what parts to go for.

Intel® Pentium® Processor Extreme Edition 840
(2M Cache, 3.20 GHz, 800 MHz FSB)

That thing had Hyper-Threading too.

in April of 2001, the Pentium III 800 rendered this same "chess2" POV-Ray scene in just under 24 minutes.
http://techreport.com/review/28751/intel-core-i7-6700k-skylake-processor-reviewed/13

In Starcraft 2, an extremely CPU limited title, when paired with a GTX460, a $184 i5 2400 3.1Ghz (3.4Ghz turbo) is ~5.5X faster than $1000 2005 Pentium 4 840.

And for people who tend to always downplay how amazing Nehalem was vs. Kentsfield Core 2 Quad back in the days, Starcraft 2 players would remind you. On Ultra settings with a GTX480, Nehalem was a REVOLUTION.

http://www.techspot.com/review/305-starcraft2-performance/page13.html

 

[Dresdenboy]

Intel can't just decide to take a Skylake-like design and clock it at 6GHz, no matter how much they're willing to sacrifice power consumption. Not unless it came with an LN2 cooling setup. There are limits to how high things can be clocked, which is why they've been putting so much effort into extracting more performance per clock cycle, even though that too is hitting diminishing returns.

There were also significant differences on a lower than microarchitectural level, P4 used more power consuming logic (does "domino logic" and "low voltage swing" ring any bells?) vs. Nehalem's cooler but slower static CMOS logic.

 

[coercitiv]

There we go, problem solved.