(Guru3D) kaby Lake i7 is another boring quad

[Sheep221]

ZEN could give us 6 or 8 cores at the price of those Quad Core KabyLakes but no OEM will ever use ZEN (no iGPU) so Intel will not lower prices.

Its the ZEN APUs Intel will need to worry more but those will face 10nm Cannonlake in the mobile segment in 2H 2017 and beyond.

Also, i wouldn't expect any CannonLake 10nm Desktop SKUs in 2017.

Why not? They release something new each year, in fact they released 2 generations in 2015 alone.

 

[inf64]

Yea, back to the great performance, low price hope from Zen. I seriously doubt we are going to get both. They may undercut intel by 50 or a hundred bucks, but I seriously doubt they are going to sell 8 core haswell performance for four hundred bucks. At least for sure not initially.

My most likely guess is that performance of 8 core zen will be competitive to hex core intel, maybe better in highly multithreaded, less in lightly threaded. If that is the case, they might sell it around 300 to 350.

A big advantage ultimately though will be that it is on the mainstream platform. Unfortunately, initially, the platform is new too so there wont be many people upgrading from a lower core count chip to top of the line Zen.

I think you nailed it with this comment.

 

[ehume]

That 875 shows why you should get something with a bit more balls than a basic dual. Its over 5 years since its come out and its still acceptable for an allrounder (minus gaming).

Since she really does use her box for gaming, and no longer for Photoshop; and she uses dGPU's to keep up with games, that 875 is in fact still acceptable for the gaming she does -- and I first put that chip in my testing box six years ago. We're waiting for the Next Big Thing (tm), I guess.

 

[CHADBOGA]

Since she really does use her box for gaming, and no longer for Photoshop; and she uses dGPU's to keep up with games, that 875 is in fact still acceptable for the gaming she does -- and I first put that chip in my testing box six years ago. We're waiting for the Next Big Thing (tm), I guess.

ehume,
In the linked thread in your signature, all the pics in the Opening Post are inoperative.

 

[chucky2]

Just Kaby & Skylake

Has this been actually confirmed though? Or is it just supposition?

 

[CHADBOGA]

Has this been actually confirmed though? Or is it just supposition?

They have confirmed that Skylake will run on the 200 Chipset and they haven't said diddly squat about Cannonlake, but it is hard to imagine Cannonlake won't come with a different pin array on the socket, looking at Intel's past practice here.

 

[AtenRa]

From a reliable source, first Kaby samples can go to 5GHz on air.

Lets see if production SKUs can do the same or better.

 

[chucky2]

They have confirmed that Skylake will run on the 200 Chipset and they haven't said diddly squat about Cannonlake, but it is hard to imagine Cannonlake won't come with a different pin array on the socket, looking at Intel's past practice here.

Ok, supposition then. On wiki (hardly a hard fact site I know) they list the 200 series chipset for both Kaby and Cannon. I'm holding out hope that Cannon will work on Kaby 200 series chipset boards but only like with maybe DDR4 memory (no DDR3 memory controller). I'd be fine with that...

 

[moonbogg]

From a reliable source, first Kaby samples can go to 5GHz on air.

Lets see if production SKUs can do the same or better.

WowWee! If that's on air, then under an ocean of water it may go further. That means the 8 core version would be awesome, but the quad would still be boring.

 

[jpiniero]

5 Ghz on air would be terrific IMO.

Ok, supposition then. On wiki (hardly a hard fact site I know) they list the 200 series chipset for both Kaby and Cannon. I'm holding out hope that Cannon will work on Kaby 200 series chipset boards but only like with maybe DDR4 memory (no DDR3 memory controller). I'd be fine with that...

It just seems so unlikely given Intel's history once they announced Kabylake. Plus bringing back the FIVR is more important.. that would be tough to have compatibility.

 

[frozentundra123456]

Well, I would not get too excited. Seems like early leaks are always higher than production samples get for intel cpus. Maybe overclocking will be decent though. Seems Skylake is a bit better than Haswell, and Kaby will be on a more mature process.

 

[Dresdenboy]

You mean the same AMD that showed this slide to investors/analysts in 2012?

"Focus on power-performance optimized cores"

Thanks, I nearly forgot about that slide. This single point basically says the same, what the October znver1 patch revealed (some not publicly analyzed yet like the iterative multiplier): Leave out, what worsens power/performance (for the targeted software mix), put in what helps. Go from theoretical 20W cores with high throughput 256b FP units + 256b cache data paths to 10W cores with 128b units + data paths, somewhat lower theoretical throughput, but capable to use twice as many cores and maybe 30% higher turbo clocks.

This will look bad in CB ST and x265 single threaded compared to KBL, but will basically double caches, cache ports, AGUs, FP units, reg ports, etc. for MT code vs. the theoretical quad core Zen variant.

 

[Arachnotronic]

From a reliable source, first Kaby samples can go to 5GHz on air.

Lets see if production SKUs can do the same or better.

That would be nice, if true. I would be compelled to upgrade my 6700K if I could get something that ran at 5GHz on high end air; have always wanted a CPU that runs at the "magic" 5GHz mark

 

[Silverforce11]

Are you expecting Zen to deliver better performance than Kaby Lake?

No, but I sure as heck expect it to deliver more cores/$, good performance, and overall excellent perf/$.

For example, a 5960X @ $999 doesn't interest me. But a Zen equivalent at $399 will sure as heck shake things up.

ps. Intel margins & profits are nuts. Skylake is the size of Polaris 11. Think about that for a bit.

 

[Arachnotronic]

For example, a 5960X @ $999 doesn't interest me. But a Zen equivalent at $399 will sure as heck shake things up.

Why would AMD sell something equivalent to a 5960X for $399? AMD tried to sell an FX-9590 for ~$880 for heavens' sake!

ps. Intel margins & profits are nuts. Skylake is the size of Polaris 11. Think about that for a bit.

Size isn't everything. You should see how much $/mm^2 NAND flash costs, makes every logic vendor look like it's ripping you off big-time. Intel's gross profit margins are good, but not out of this world.

 

[RussianSensation]

No, but I sure as heck expect it to deliver more cores/$, good performance, and overall excellent perf/$.

For example, a 5960X @ $999 doesn't interest me. But a Zen equivalent at $399 will sure as heck shake things up.

ps. Intel margins & profits are nuts. Skylake is the size of Polaris 11. Think about that for a bit.

That's not realistic. If 8-core Zen ~ 5960X and Intel succeeds 5960X with 6900K for $999, then why would AMD price "Zen equivalent" at $399? They'd still be competitive even selling it at $799 then.

Another inconsistency is that how can we have an 8-core Zen ~ 5960X when 40% IPC increase over their fastest architecture is not even close to Haswell's IPC? That implies Zen would need to have far higher boost clocks than 5960X and/or overclock better than 5960X which tends to hit 4.4-4.5Ghz.

I also do not see the point of comparing 2017 Zen to the August 2014 5960X. Isn't it unfair to compare a Q1 2017 part to a Q3 2014 part but then ignore that SKL-E is likely a Q2-3 2017 part? I mean if you are going to compare Zen to Haswell-E, considering their time difference, it's only fair then that we compare an Intel equivalent new architecture in 2017 -- that's Skylake-E not Haswell-E/Broadwell-E.

For gaming, i7 5820K/5960X still cannot beat i7 6700K.
https://www.youtube.com/watch?v=ocwwaVGUFtk

That means if Kaby Lake can consistently hit 4.9-5Ghz on air, it'll still be better to purchase a 4-core KL over an 8-core Zen. AMD's main hope will be that most of AAA games in 2017 and beyond are DX12 where the IPC deficiency of Zen is made up by better multi-threaded games. Once again though, in 2017 Intel will have Cannonlake. Zen's biggest issue is timing, just like the Nintendo NX and Vega 10. If you don't show up, you forfeit sales. Every new Intel CPU launch = consumers building new systems buy Intel, and are not waiting for Zen. That means there are less and less customers who will be upgrading to Zen. With the arrival of Kaby Lake and Broadwell-E, many hold-outs on Nehalem, Sandy and Ivy Bridge users will finally upgrade, especially i3/i5 users.

If Zen was so good, why is AMD delaying the launch of dual and quad-core Zen chips? Do you honestly believe their strategy to regain market share will be to launch an 8-core Zen at $400+?
http://forums.anandtech.com/showthread.php?t=2472053

Sounds like it's a repeat of Phenom I/II where Zen will trail Intel in single-threaded performance, but use way less power than Bulldozer.

 

[Arachnotronic]

That means if Kaby Lake can consistently hit 4.9-5Ghz on air, it'll still be better to purchase a 4-core KL over an 8-core Zen. AMD's main hope will be that most of AAA games in 2017 and beyond are DX12 where the IPC deficiency of Zen is made up by better multi-threaded games. Once again though, in 2017 Intel will have Cannonlake. Zen's biggest issue is timing, just like the Nintendo NX and Vega 10. If you don't show up, you forfeit sales. Every new Intel CPU launch = consumers building new systems buy Intel, and are not waiting for Zen. That means there are less and less customers who will be upgrading to Zen. With the arrival of Kaby Lake and Broadwell-E, many hold-outs on Nehalem, Sandy and Ivy Bridge users will finally upgrade, especially i3/i5 users.

I like the bit in bold, because for gamers this is absolutely true. I had a 5960X and actually "downgraded" to a 6700K. I am hitting higher clocks than I was able to with the 5960X and I have more performance per clock, so for what I do most (gaming), my 4.6-4.7GHz 6700Ks are simply better.

A 5GHz Kaby Lake, if it's real, would be even better. Given how well Intel CPUs hold their value, I would be able to pull off that upgrade for probably less than $100 out of pocket.

 

[Silverforce11]

Man, you are not going to perceive the difference between a 5960X vs Skylake or KL in games. Stop with the charade because it's very unrealistic!

Even if you had SLI/CF, game on 1440p crushes GPUs and so they become the bottleneck, always.

There's the old saying, if your GPUs aren't max loaded, your graphics settings aren't high enough.

The difference boils down to "excellent for gaming, excellent for everything else" vs "excellent for gaming, good for everything else". That's the difference between 8 core Intel vs 4 core Skylake.

 

[IntelUser2000]

Silverforce11:

People ALWAYS want better. That's why we went from 1600x1200 screens to quadruple that. The thing is, if you compare them side by side, the differences are not much. And that's the limitation of your eye quite likely. But your brain, when you get used to certain settings/resolutions, make that the new "normal". Meaning you need even higher settings and resolutions.

Same with games. My friend had a 980X + Geforce 780 to play Quake Live! And most games he ran at lowest settings possible, not just to reduce distractions but for absolutely the smoothest experience at ANY scenario.

Also, VR will make CPUs more important, especially if they go from current 90 fps to higher ones. Nvidia was demonstrating 1000+ fps one weren't they? Imagine at that point. You'd really need a Geforce 780 to VR a 1000+ fps Quake Live!

 

[RussianSensation]

Man, you are not going to perceive the difference between a 5960X vs Skylake or KL in games. Stop with the charade because it's very unrealistic!

Even if you had SLI/CF, game on 1440p crushes GPUs and so they become the bottleneck, always.

There's the old saying, if your GPUs aren't max loaded, your graphics settings aren't high enough.

The difference boils down to "excellent for gaming, excellent for everything else" vs "excellent for gaming, good for everything else". That's the difference between 8 core Intel vs 4 core Skylake.

Using your logic, Max IQ 1440p/DSR/VSR/4K, we could even bring Big Pascal SLI to its knees. In that case, what's even the point of upgrading for gaming from an i7 2600K/3770K @ 4.5-4.8Ghz?

If you make the argument that gaming is 99-100% GPU limited, then Zen is even more pointless. That means might as well get 6-core Broadwell-E 6800K, overclock it to 4.5ghz and enjoy it starting this summer.

That's why all the arguments for "waiting for Zen" make no sense. If someone needs a new rig in 2017, sure they will have the choice of Zen vs. Intel. If you say you NEED 8 fast cores ~ 5960X OC for XYZ tasks, then how did you manage to survive since 2014 on a quad core i5/7 when even a 6-core 5820K @ 4.4-4.5Ghz would have made your life easier in encoding, rendering, encryption, etc.?

If you really needed a fast 6-8 core CPU, you would have already purchased one. In fact, for multi-threaded tasks even a 3930/4930 OC would beat i7 2600/3770/4770/6700K, etc. I just don't get why someone didn't need 6-8 cores in 2012, 2013, 2014, 2015, 2016, but then boom the minute Zen launches, it's 8-cores or bust?

Please tell me how that's a consistent argument the people hyping up Zen are making?

I like the bit in bold, because for gamers this is absolutely true. I had a 5960X and actually "downgraded" to a 6700K. I am hitting higher clocks than I was able to with the 5960X and I have more performance per clock, so for what I do most (gaming), my 4.6-4.7GHz 6700Ks are simply better.

A 5GHz Kaby Lake, if it's real, would be even better. Given how well Intel CPUs hold their value, I would be able to pull off that upgrade for probably less than $100 out of pocket.

Early leaked Skylake OCs were showing 5.2Ghz. It turned out to be BS though. I doubt you'd lose $100 selling the i7 6700K. I bet you can do better by timing it 2 weeks before i7-7700K drops. A lot of gamers don't follow CPU releases religiously. A less than a year old 6700K can probably be sold for $60-70 less since it still has at least 2 years of warranty left. That's why I see almost no risk at all buying BW-E, Skylake or Kaby Lake before Zen even shows up. The risk of Zen flopping is huge but the probably that Zen beats a 4.7Ghz+ Skylake in games is very slim. Let's not discount that Zen also has to have similar perf/watt in overclocked states vs. Skylake/BW-E. Nehalem i7s, Bulldozer/Vishera overclocked well but their power usage went through the roof doing so. What makes Intel's 14nm chips so good is that even when overclocked to the max, they sip power. I also love the option of $25/30 Intel Performance Tuning Plan that allows you to murder the chip with 1.45-1.5V if you so desire. There are a lot of things that make Intel K series so good, and it's not just IPC.

 

[Silverforce11]

That's the point. Upgrading your CPU for gaming is meaningless if you're on an i5 of recent generation.

You upgrade the CPU for other reasons besides gaming. What am i going to benefit going from Ivy i5 to Skylake for gaming? Nada. I am 100% GPU limited. If I am not, it's cos I haven't cranked up the visuals. :/

But next year, I will build a new rig, and this one goes down the ladder for next of kin... when the choice comes for a CPU in that new rig, if it's between 8 Core Zen at the same price as 4 Core Kaby Lake, if Zen is Broadwell IPC, I will pick 8 Core, 16 Thread Zen over KB, EASY choice. Because I do other things besides gaming, and in those things, the difference between 8/16 vs 4/8 is huge and very noticable.

People on Haswell+ have zero reason to upgrade to Skylake, no reason to upgrade to Kaby Lake for gaming. But those on Nehalem, SB or even older CPUs may. That's AMD's potential customer.

 

[Silverforce11]

Oh, I still have a SB rig that next-of-kin is using, 4.5ghz. If I plug in a GPU of Titan X caliber, do you think I will need a new CPU?

No. 1440p, max settings, modern games struggle to maintain 60 fps even on the best GPU. CPU bottlenecks don't apply until its well beyond those FPS.

Sure, I could go get a 6 Core Intel for $399 now. But I'm not building a PC now. In early 2017 when I will, the choices available, if Zen is priced aggressively, will mean Zen is a viable option. That's all AMD can hope for since they are coming back from being so far behind.

 

[HiroThreading]

But those on Nehalem, SB or even older CPUs may. That's AMD's potential customer.

C'mon, that's a disastrous strategy, and one that I highly doubt the AMD execs are aiming for.

I have to concur with what RS and Arachnotronic have said in this thread regarding Zen's timing to market and competition.

In my opinion, Zen is not going to be a chip marketed to gamers or overclockers. AMD will leave those markets to mainstream SKL and CL. Zen will however look to take back some of what AMD lost in the server and workstation market, particularly in 2P and MP systems.

 

[Silverforce11]

@HiroThreading

Is Skylake aimed at Ivy, Haswell or Broadwell gamers?

In the GPU side of things, NV's own numbers claims that ~70% of their users are still on Kepler or older. I wonder what it is like for Intel CPUs.

 

[HiroThreading]

@HiroThreading

Is Skylake aimed at Ivy, Haswell or Broadwell gamers?

In the GPU side of things, NV's own numbers claims that ~70% of their users are still on Kepler or older. I wonder what it is like for Intel CPUs.

SKL was aimed at ultrabooks, tablets and servers.

However, Intel -- unlike AMD -- has the resources to scale the core design from dual core to multi-ringbus modules.

30% user penetration for your latest architecture is a pretty good statistic. Not Apple iOS good, but still pretty good for a hardware manufacturer.