Intel processor architectures tested clock-to-clock in gaming - results all over the place

[vorticalbox]

On ‎06‎/‎01‎/‎2017 at 11:35 PM, Princess Cadence said:

I honestly feel that silicon has reached its maximum, Intel or AMD will not really upgrade much from what we already have, aside from optimizations the performance gap is always going to be smaller and smaller... Intel better be already studying replacements for Silicon otherwise the 8th gen will again be nothing more than a Kaby Lake polishing and if you have a "Lake" CPU you'll stay fine enough.

I think it will go like AMD "the only way is up" and, like memory, they will stack CPUs cores.

[SpaceGhostC2C]

1 hour ago, LAwLz said:

Source?

I don't see them mentioning timings in their page.

 

Source for what? I never said I knew their timings, I said typical timings are lower for DDR3 so you need much higher frequencies to match the speed with DDR4. They may as well have 2400 CL11 DDR4 kits for all I know. The point wasn't that timings were known, but that timings mattered, and that a run of the mill DDR4 2400 kit will be slower than a "run of the mill" (2400 was pretty high for DDR3 after all) DDR3 kit. Whatever they did, it would make sense to do these tests with equally fast RAM (which isn't the same as equal frequency RAM).

 

1 hour ago, LAwLz said:

Intel is holding back though.

They could easily make it like this:

Pentium - dual core with hyperthreading

i3 - quad core

i5 - quad core with hyperthreading

i7 - 6-core with hyperthreading

 

And then on their enthusiast platforms starts at 6-cores with extra PCIe lanes, quad channel memory, and even more cores at the higher end..

Those are marketing strategies and have no influence on IPC. Increased competition could change Intel's marketing; unfortunately, it won't make IPC gains magically appear. Intel may stop skinning us alive for a CPU if more competition arises. The point remains that drastic innovation doesn't happen deterministically every X years, nor can it be granted by competitive pressure and increased research effort, and this holds true also if you look at past developments, while AMD was still competing (like Core 2).

[Ekst4zy]

My 2500k is with me for pretty damn long. I switched my GPU 4 times (7770->7870->770->970), but CPU... I never felt like I have to.

 

Ryzen is kinda tempting me tho. I just need to this reviews and prices  

[TopWargamer]

On 1/6/2017 at 8:35 PM, crisro996 said:

*looks at 3770k*

 

Pretty much how I feel right now, although I already made a deal to sell my Ivy based PC to my brother in law. Oh well. I'm Zen bound though!

[Godlygamer23]

5 hours ago, LAwLz said:

They could easily make it like this:

Pentium - dual core with hyperthreading

i3 - quad core

i5 - quad core with hyperthreading

i7 - 6-core with hyperthreading

What if Intel made the Core i5's quad cores with hyper-threading, and Core i7's eight cores without(?) hyperthreading. Though removing hyper-threading may not be the best idea. That may actually make it worse. 

[Drak3]

16 minutes ago, Godlygamer23 said:

What if Intel made the Core i5's quad cores with hyper-threading, and Core i7's eight cores without(?) hyperthreading. Though removing hyper-threading may not be the best idea. That may actually make it worse. 

I'd rather that it be like this:

i3- Quad core, stripped of Virtualization features, basic iGPUs.

i5- quad core at the low end of the lineup, and hexacores w/o hyperthreading at the high end, with the full gamut of virtualization features, the best iGPUs outside of the Iris Pro/Plus lineup

i7- quad and hexa core with hyperthreading on all cores, full gamut of virtualization features, and the Iris Plus graphics on all Sku's.

i7 Extreme- basically same progression as what we currently have, but the entry level i7e has 150% the PCIe lanes of the i7, top end with 250% the PCIe lanes of the i7, and everything else just doubles the PCIe lanes.

[leadeater]

6 hours ago, LAwLz said:

Intel is holding back though.

They could easily make it like this:

Pentium - dual core with hyperthreading

i3 - quad core

i5 - quad core with hyperthreading

i7 - 6-core with hyperthreading

 

19 minutes ago, Godlygamer23 said:

What if Intel made the Core i5's quad cores with hyper-threading, and Core i7's eight cores without(?) hyperthreading. Though removing hyper-threading may not be the best idea. That may actually make it worse. 

 

Or do what they could have done ages ago:

i3 - 4 cores with HT

i5 - 6 cores with HT

i7 - 8 cores with HT

HEDT i7 - 8 cores with HT 40 PCIe lanes

HEDT i7 - 10 cores with HT 40 PCIe lanes

 

Don't really see a need for so many HEDT i7's and some with reduced PCIe lanes, why on earth do that? Old architecture without the benefit of why that platform exists, just buy standard i7. 

[Godlygamer23]

9 minutes ago, leadeater said:

Or do what they could have done ages ago:

i3 - 4 cores with HT

i5 - 6 cores with HT

i7 - 8 cores with HT

HEDT i7 - 8 cores with HT 40 PCIe lanes

HEDT i7 - 10 cores with HT 40 PCIe lanes

 

Don't really see a need for so many HEDT i7's and some with reduced PCIe lanes, why on earth do that? Old architecture without the benefit of why that platform exists, just buy standard i7. 

My main issue(which you seem to go around) is segmenting the market even more with dual and quad cores within the same bracket. It's the issue that currently plagues the mobile market, and one I would rather not see affect the desktop market too much(even though it already has).

[Drak3]

28 minutes ago, leadeater said:

Don't really see a need for so many HEDT i7's and some with reduced PCIe lanes, why on earth do that?

Because not everyone needs 40 PCIe lanes, but they can need 6 cores. Intel just offers an option to them, where they aren't wasting money on something they won't use. It'd be different if Intel brought mainstream hexacores to market, but they haven't.

[leadeater]

31 minutes ago, Drak3 said:

Because not everyone needs 40 PCIe lanes, but they can need 6 cores. Intel just offers an option to them, where they aren't wasting money on something they won't use. It'd be different if Intel brought mainstream hexacores to market, but they haven't.

So then don't buy a HEDT CPU, the only reason you have to is due to Intel not offering higher core counts on their mainstream CPUs (my actual point in the post) which if they really wanted to could. Buying HEDT CPUs for gaming isn't actually the better option, and Intel actually has to disable PCIe lanes which are there in the chip to give us HEDT CPUs with less. I object to it mainly since it's wasteful and serves no purpose and doesn't increase usable chip yields, really is just an artificial product segmentation but I don't specifically object to those that much when it makes sense or serves a purpose.

[leadeater]

1 hour ago, Godlygamer23 said:

My main issue(which you seem to go around) is segmenting the market even more with dual and quad cores within the same bracket. It's the issue that currently plagues the mobile market, and one I would rather not see affect the desktop market too much(even though it already has).

Sorry I likely missed key parts of the conversation chain since I was replying at work and only had so much time to read. Going back to read thread after this reply.

 

I tend to prefer product lines that serve functional purposes hence why I went with what I suggested. Keep in mind I much more often buy Xeon CPUs and that is very much how those SKUs work, apart from the high clock rate workstation ones but there are only a few of those.

 

Disabling cores increases the usable yield much more than dropping frequency, at lest I'm pretty sure it does. Intel has 22 6th generation i5 SKUs and I think that is way too many, 16 quad core and 6 dual core. 10 are desktop, 9 are mobile and 5 are embedded. Do we really need 10 minor i5 variants?

 

I would make every core equally feature rich and equipped across i3, i5 and i7. Have a couple of models with varying frequency but not to many and 1 or 2 low power ones.

 

i3 (4 cores) = Mass consumer: Internet browsing, email, office applications, light/casual gaming. Likely see i3's in most business desktops.

i5 (6 cores) = General gamer: Single GPU, light content creation, light streaming etc + i3 tasks. Most people on this forum would fit here.

i7 (8 cores) = Enthusiast gamer: Single or dual GPU, moderate/advanced content creation, avid streamer/streaming as job etc + i3 tasks. Significant number on this forum also fits here.

HEDT (8 cores) = Enthusiast gamer or professionals: Single or dual GPU, many local storage devices, PCIe devices e.g. 10Gb NIC etc. Mostly find this in production studios and media design companies.

HEDT (10 core) = Professionals: Same as above but running much larger/extreme workload tasks. Mostly find this in engineering teams running smaller simulations, video production studios, research projects for doing smaller tasks or optimizing before running on larger cluster.

 

 

I'm not sure I fully understand the market segmentation your talking about in the desktop space, would like to here more of your thoughts about that. Did what I suggest make it worse? Again I may have missed some important information further back in the thread.

[leadeater]

17 hours ago, SpaceGhostC2C said:

Now AMD is apparently catching up with this lot and plans to keep the same socket for at least 4 years: how many revolutionary improvements can you achieve without changing the socket? Bulldozer to Excavator was an impressive overall IPC increased, when compared to Intel's past 4 years, but it required FM1->FM2-FM2+ (I guess the reason they never got past Piledriver in AM3+ was the inability to keep the same socket - but just guessing). So, AMD's bet is to remain competitive 4 more years without stellar upgrades in IPC, which requires Intel to spend 4 years more in the same ballpark...

AMD has quite a good history of keeping socket/chipset compatibility for their CPUs. Even when they released a new chipset and CPU sometimes that CPU was still backwards compatible with the old chipset, same socket.

 

It's something I appreciate about AMD, even though as of late I have not used any of their products (CPU) except for the horridly bad E2-1800 in my laptop (get what ya pay for lol).

[Vode]

3 hours ago, leadeater said:

Or do what they could have done ages ago:

i3 - 4 cores with HT

i5 - 6 cores with HT

i7 - 8 cores with HT

HEDT i7 - 8 cores with HT 40 PCIe lanes

HEDT i7 - 10 cores with HT 40 PCIe lanes

This would be a really dumb thing to do for Intel:

 

HEDT needs more to separate itself from standard i7s. More PCIe lanes is really a server feature. Even the more cores-feature is becoming a hard sell and the benefit is limited to specific apps.

 

The software is still catching up.

 

 

[Space Reptile]

On ‎09‎.‎01‎.‎2017 at 8:06 PM, Inkz said:

*looks at 4770k* 

*looks at 2600*

[leadeater]

5 minutes ago, Vode said:

This would be a really dumb thing to do for Intel:

 

HEDT needs more to separate itself from standard i7s. More PCIe lanes is really a server feature. Even the more cores-feature is becoming a hard sell and the benefit is limited to specific apps.

 

The software is still catching up.

In the space where HEDT is used the software is more than there, the problem is improperly using HEDT in use cases where it doesn't make sense. This is happening due to the strangle hold on low core count desktop platform.

 

Almost every researcher at the university I work at would love a 8 core or 10 core, which they can get now, and make use of it instantly but that is still rather costly. Why we have compute clusters.

 

We also have multiple computer labs for video editing with 30 computers in them all with 10Gb networking and lots of fast local storage, non HEDT CPUs don't have the PCIe lanes for the configuration we use.

 

We also have workstations that connect up to measurement devices in the laboratories which ingest huge amounts of data points and process them, these still have to be small devices but have the power to do the job with the storage speed and network speed to keep up. 

 

HEDT literally uses the same CPU die as the high end Xeon CPUs, without multiple socket support and higher clock rates to better support GPU accelerated tasks. These high clock rates mean they have to reduce the core count to stay within acceptable power and cost limits. The high PCIe lanes seems like a server feature because it is but can be shared in the very high end desktop computing space.

 

I think the issue is more what is the definition of High End Desktop Computing (HEDT), and this differs depending on person.

[ZothOmmog]

21 hours ago, Fetzie said:

Wait for Ryzen, and either 5820k or 6800k if Ryzen falls on it's face.

Ugh wait wait wait wait wait. That's it?

 

By the way why is both Kaby and Sky limited to 4 ram slots?

[Drak3]

6 hours ago, leadeater said:

I object to it mainly since it's wasteful and serves no purpose and doesn't increase usable chip yields, really is just an artificial product segmentation but I don't specifically object to those that much when it makes sense or serves a purpose.

Again, it provides a cheaper option to those than need 6 cores, but don't need more than 28 PCIe lanes.

It's like asking Ford or Dodge why they sell base versions of most of their vehicles: some people want something that'll fit their needs, but don't have features useless to them that ramp up the price.

[Morgan MLGman]

I call BS on their (probably flawed) testing methodology:

 

This is real proof.

 

[leadeater]

13 minutes ago, Drak3 said:

Again, it provides a cheaper option to those than need 6 cores, but don't need more than 28 PCIe lanes.

It's like asking Ford or Dodge why they sell base versions of most of their vehicles: some people want something that'll fit their needs, but don't have features useless to them that ramp up the price.

Your still missing the point. A 6 core offering in the standard desktop line of CPUs is a better more consumer focused/friendly option. That platform is on a faster development cycle and already suits the people that don't need 40 PCIe lanes.

 

The 6800K makes zero sense when you look at the 6850K, at all. It makes so little sense they had to cut 200Mhz from the 6800K. The ludicrous price increase for the very same CPU die without the PCIe lanes disabled is nothing more than a cash grab, the 6800K is proof the 6850K is massively over priced.

 

The 6800K is a Ford Mustang with the EcoBoost engine, nobody actually wants that. UK were almost forced to only have that option, luckily they later got the full proper Mustang.

 

Edit:

Just because the 6800K option exists now doesn't make it the best possible option, it's the only option.

[Drak3]

1 minute ago, leadeater said:

Your still missing the point. A 6 core offering in the standard desktop line of CPUs is a better more consumer focused/friendly option. That platform is on a faster development cycle and already suits the people that don't need 40 PCIe lanes.

I'm not missing the point. In a reply to you, I said:

8 hours ago, Drak3 said:

It'd be different if Intel brought mainstream hexacores to market, but they haven't.

 

But hexacores on Z270 and lower still doesn't exist. Those that want hexacores, but don't want to pay for things they don't need. The 5920K and 6800K are the next best things to a mainstream class hexacore.

 

7 minutes ago, leadeater said:

The 6800K is a Ford Mustang with the Eco Boost engine, nobody actually wants that.

The 6800K is more like an older Mazda B4000, exactly what some people want and need.

[leadeater]

17 minutes ago, Drak3 said:

I'm not missing the point. In a reply to you, I said:

 

But hexacores on Z270 and lower still doesn't exist. Those that want hexacores, but don't want to pay for things they don't need. The 5920K and 6800K are the next best things to a mainstream class hexacore.

 

The 6800K is more like an older Mazda B4000, exactly what some people want and need.

Sorry I missed that (part of the) reply. But that is what I'm advocating.

 

I totally disagree with the Mazda comparison though, the Ford Mustang ecoboost more accurately fits the real technical difference between the 6800K and the 6850K.

 

But that is still my point you are using, it is the only 6 core option but it shouldn't be. It also doesn't cost Intel any more or less to sell the 6800K and 6850K, the 6800K shouldn't exist and the 6850K should be cheaper.

 

Edit:

This chain that @LAwLz started wasn't about what is available now, it was about what we think Intel should offer.

[Drak3]

1 minute ago, leadeater said:

But that is still my point you are using, it is the only 6 core option but it shouldn't be. It also doesn't cost Intel any more or less to sell the 6800K and 6850K, the 6800K shouldn't exist and the 6850K should be cheaper.

Every serious Intel offering should be cheaper. But there still is the need to have steps correlating to common use cases amongst the target audiences. The 5820K/6800K serve as an intermediate step between their Z97/Z170 counterparts, and the more expensive, and to a decent amount of people, unnecessary, 5930K/6850K and up.

 

There's also the possibility that many 5820K's and 6800K's start out as their more powerful counterparts, but are defective. Should that be the case, I'd rather Intel salvage them and sell them as a lower performance CPU than just scrap it.

[leadeater]

Just now, Drak3 said:

Every serious Intel offering should be cheaper. But there still is the need to have steps correlating to common use cases amongst the target audiences. The 5820K/6800K serve as an intermediate step between their Z97/Z170 counterparts, and the more expensive, and to a decent amount of people, unnecessary, 5930K/6850K and up.

 

There's also the possibility that many 5820K's and 6800K's start out as their more powerful counterparts, but are defective. Should that be the case, I'd rather Intel salvage them and sell them as a lower performance CPU than just scrap it.

See my edit above btw,

 

Also cutting PCIe lanes likely wouldn't turn a chip that failed a 6850K validation in to a working 6800K chip, if it failed that validation there is likely too much wrong with it to be usable at all. Could be wrong on this though, that area of the die is significant in size but there is more than just PCIe lanes in that area. If I am incorrect and cutting lanes gains significant amounts of usable yield then I'd be happy to say the 6800K makes a lot more sense than I currently think it does.

[Fetzie]

2 hours ago, ZothOmmog said:

Ugh wait wait wait wait wait. That's it?

 

By the way why is both Kaby and Sky limited to 4 ram slots?

Because they only have a dual channel memory controller.

[ZothOmmog]

28 minutes ago, Fetzie said:

Because they only have a dual channel memory controller.

I know, I was kind of more thinking out loud why they actually just went with dual and not quad, I do understand that quad might be more enthusiast grade component requirement but but.