Linus Has Some Things to Say about Core i9 and X299

[trparky]

this lane castration is a bit odd

Nothing odd about it, it's a pure cash grab.

Want more PCIe lanes? Be prepared to open that wallet wide open!

AMD's dividers

Along with updated AGESA.

 

[Frag_Maniac]

Nothing odd about it, it's a pure cash grab.

Want more PCIe lanes? Be prepared to open that wallet wide open!

No, it IS odd. It's clear just as Linus said that they are panicked by what AMD is doing, yet taking away Pci lanes at a time when they are becoming very useful, and while AMD is literally offering competing high end chips for half the price, will in no way get anyone with half a brain to shell out more for the Intel chips. That's very nonsensical marketing, eg, "odd".

Their only hope to offer any kind of competition now is Coffee Lake. It's clear Intel has always been a cash grab company. It's just that now they have no clue how to get away with it. All they need do is follow what Linus said, make the best chip you can for the best price.

 

[trparky]

Maybe it's my more cynical nature that thinks Intel is doing this just to be a dick.

 

[Frag_Maniac]

Maybe it's my more cynical nature that thinks Intel is doing this just to be a dick.

They've always been part dick. They used to be top innovators. Zero competition took them out of that mindset. Now they've been caught off guard and are just panicking. Hopefully when they see what consumers think of this scheme, they'll have an epiphany, and actually try to compete reasonably again.

 

[trparky]

Considering the kinds of reviews that can be had on YouTube regarding this platform Intel better wake up and have that epiphany pretty damn quickly. Because (at least to me) unless you have a bottom-less wallet and you just like spending money for the sake of spending money, the x299 platform is a royal fuck up.

 

[EarthDog]

My largest complaint so far is the lack of 10 Gb Ethernet in X299. This was back in the day planned for X99, but dropped due to cost. Anyone doing productive work knows 1 Gb Ethernet is too slow.

Cost and literally 10 people used it when x99 was in development.





Being more serious, if you need it, get a 10gb NIC (will cost as much as the board...). GBe is just now starting to hit very select areas, and is quite costly. This yoy may find interesting...: http://www.speedtest.net/reports/united-states/

Again.. intel/board makers being smart and playing the numbers. There are so few that would he useful for on a consumer platform, there isnt a point...well, i guess unpess you are one of those baby's hand full of people that manage to have over 1gb net at home... and manage to actually need it.

If its for the office, you are likely on fiber and using 10gb sfp. If not...yep...pcie card.




"Panic"...intel.... haha, oh man.. sorry, need to change my depends...made a booboo laughing.

 

[FR@NK]

Name me one well known YouTube personality who isn't a tool.

...

Ok, you got me.

At this point everyone is going with the 'trend'. Can't blame them they need to make a living with those videos.

I guess you guys dont watch big clive?

Granted, $600 for their 8 core, IF it really does bench 25-35% faster (by syn scores) than an 1800X might seem reasonable

It wont bench much faster maybe 10% or less. But it will be alot faster for lightly threaded workloads since the max turbo will be 4.5GHz for 2 of the cores. Quad channel will be supported but will only make a difference under heavy multi threaded workloads that have large worksets in the ram. The default speed will be 2666.

 

[TheoneandonlyMrK]

They've always been part dick. They used to be top innovators. Zero competition took them out of that mindset. Now they've been caught off guard and are just panicking. Hopefully when they see what consumers think of this scheme, they'll have an epiphany, and actually try to compete reasonably again.

On the innovation side it's interesting that ibm now has ties to Samsung and Gf ,back when intel were innovating im sure they had strong ties with IBmzzz

 

[Frag_Maniac]

It wont bench much faster maybe 10% or less.

Reason I said 25-35% is there's already been LN2 benches shown with those results over an LN2 OCed 1800x, but of course no source to say if it's credible.

 

[Vya Domus]

Benchmarks on LN2 ... why they would bear any meaning is again , beyond me.

 

[Countryside]

This image explains the x299 situation the best way.

 

[Mr.Scott]

On the innovation side it's interesting that ibm now has ties to Samsung and Gf ,back when intel were innovating im sure they had strong ties with IBmzzz

Interesting that you said that. I was just looking at this.
http://redirect.viglink.com/?key=bb...ware/ibm_has_revealed_their_first_5nm_chips/1

 

[TheoneandonlyMrK]

Interesting that you said that. I was just looking at this.
http://redirect.viglink.com/?key=bbb516d91daee20498798694a42dd559&u=https://www.overclock3d.net/news/misc_hardware/ibm_has_revealed_their_first_5nm_chips/1

I saw something similar but the slide I mentioned was for Amds future schedule, after 7nm Zen++ 3 weva they listed gate around transistors to follow sometime 2019-20 i think but for the life of me i cant think where i think it was a minor detail in their Threadripper presentation but can't be sure , could even have been wccftech which would make it dube.

 

[trparky]

If they can get Ryzen down to 7nm maybe they can increase the clock speeds of these chips. That's the number one thing that's holding Ryzen back, the lack of high clock speeds. High clock speeds equals high IPC.

 

[EarthDog]

High clock speeds equals high IPC.

No.

Instructions Per Clock is what it stands for. By definition it means how many instruction per clock cycle the cpu can process. That does NOT change with clockspeed.

In other words, if a cpu can handle 1M instructions per hz/clock cycle at 4.2ghz, its still doing 1M instructions per hz/clock cycle at 5ghz. The difference is you are pushing more clock cycles, not what it can do per clock cycle.

 

[TheoneandonlyMrK]

If they can get Ryzen down to 7nm maybe they can increase the clock speeds of these chips. That's the number one thing that's holding Ryzen back, the lack of high clock speeds. High clock speeds equals high IPC.

That's not really true of multi core despite it appearing so , if clocks were all that we would not have HCC cpus and since they have gone down the wider is better route , similar to gpus ,core clocks are pretty much stalled until graphene or molybdenum chips are sorted out , the 7700 k could well have been a last harrar of 5ghz capable cpus of which mine happily partook five years earlier, that times past, as has the time of pure ipc improvement , there's only so much resource can be put in a core and the prefetching and handling of data must be approaching a plateau.
Neither intel or Amd are going to up integer ,scaler or vector functionality within the next 3 years and by then we Will be talking 124 core chips so , like i said the Hz dream is done.

 

[trparky]

Then tell me why IPC numbers increase as one overclocks a chip? I overclocked my Core i5 3570k from the stock speed of 3.4 GHz to 4.3 GHz and it definitely helped, not in all situations but it did help. Bumping the clock speed up improved performance. If clock speed isn't connected to IPC then what happens when you overclock a chip? Why does a 7700k at 4.5 GHz beat a Ryzen at 3.5 GHz? There's got be some kind of correlation between clock speed and IPC.

 

[EarthDog]

Then tell me why IPC numbers increase as one overclocks a chip? I overclocked my Core i5 3570k from the stock speed of 3.4 GHz to 4.3 GHz and it definitely helped, not in all situations but it did help. Bumping the clock speed up improved performance. If clock speed isn't connected to IPC then what happens when you overclock a chip? Why does a 7700k at 4.5 GHz beat a Ryzen at 3.5 GHz? There's got be some kind of correlation between clock speed and IPC.

Read my passage again bud...

IPC doesnt increase, the number of times the processor can process those instructions for each clock do. To use my example above, you would be able to process 42B instructions in a second at 4.2ghz, and 50B instructions at 5ghz...but the IPC (again, the number of instructions per clock cycle) does not change. But since you are cramming more clock cycles in that same second, performance goes up... IPC does NOT.

 

[okidna]

Then tell me why IPC numbers increase as one overclocks a chip? I overclocked my Core i5 3570k from the stock speed of 3.4 GHz to 4.3 GHz and it definitely helped, not in all situations but it did help. Bumping the clock speed up improved performance. If clock speed isn't connected to IPC then what happens when you overclock a chip? Why does a 7700k at 4.5 GHz beat a Ryzen at 3.5 GHz? There's got be some kind of correlation between clock speed and IPC.

I think you misinterpreted what IPC is. What you described is NOT IPC (IPC is Instructions Per Cycle, the average number of instructions executed per clock cycle at any given clock speed) what you described above is instructions per SECOND.

Already answered by @EarthDog :

In other words, if a cpu can handle 1M instructions per hz/clock cycle at 4.2ghz, its still doing 1M instructions per hz/clock cycle at 5ghz. The difference is you are pushing more clock cycles, not what it can do per clock cycle.

 

[TheoneandonlyMrK]

Then tell me why IPC numbers increase as one overclocks a chip? I overclocked my Core i5 3570k from the stock speed of 3.4 GHz to 4.3 GHz and it definitely helped, not in all situations but it did help. Bumping the clock speed up improved performance. If clock speed isn't connected to IPC then what happens when you overclock a chip? Why does a 7700k at 4.5 GHz beat a Ryzen at 3.5 GHz? There's got be some kind of correlation between clock speed and IPC.

There is but your not getting me , look the feck around.
Show me a 16 core plus cpu doing 5ghz then think 32/ 64/ 128 cores , clocks wont go up because they All (every company making processors/feckin any asic)chose wider and x86 is already at the pinicle of what is possible (obv not technically) for consumers to own and use.
3 years minimum before that could change and im wagering not then.

My future viewing powers arent bad, radeon on intel for instance ,i said it's likely not just myth and i posited it could be for Apple based on rumours and the way shit rolls ,you may be hearing about kabylake G at some point ,not that i knew per say but shit does Always roll downhill.

 

[trparky]

you would be able to process 42B instructions in a second at 4.2ghz, and 50B instructions at 5ghz...

Ah, now it makes sense!!!

 

[Frag_Maniac]

Benchmarks on LN2 ... why they would bear any meaning is again , beyond me.

Even worse, they tossed in a stock clock bench of the 1800x, but no such thing for the 7820x.

And it's not real world similarity I'm implying with LN2, it's apples to apples comparison against 1800x, because the only way they compared them in that regard in that bench was both OCed on LN2.

None the less, take it with a grain of salt like I said. I'm thinking it might be a while before we even get any credible reviews or consumer feedback, because with this scheme not many are going to want to shell out for them, and I wouldn't be surprised if review samples will be cherry picked by Intel to skew results.

 

[FR@NK]

I wouldn't be surprised if review samples will be cherry picked by Intel to skew results.

How would cherry picked review samples effect reviews results? Would effect overclocks for sure but it wouldn't effect the performance of the chips running at the stock clock speeds.

We dont really need reviews anyways. These chips are similar IPC to current broadwell-e chips. Since we already have a good idea how broadwell-e performs, we can conclude that because skylake-x chips are running faster they will be faster. They also will cost less per core. Skylake-x will be the best performance per dollar jump compared to the previous generation we have ever seen from Intel. I'm sure anyone looking to spend $500 on a processor will gladly shell out the extra Benjamin for the 7820x.

 

[JunkBear]

Oh Man Im so happy to not have that Power craving. Im still rocking my Dell Inspiron 640m laptop and the rig under my avatar.

 

[Solaris17]

JaysTwoCents does make a valid point in his video. He talks about the fact that in what? Two or three months Coffee Lake is going to be coming out and if the rumors turn out to be true it will bring six core, twelve thread CPUs to a mainstream platform. We now have Kaby Lake-X with its four cores (eight threads) existing on the X299 platform that's going to be slower in terms of IPCs with all of its threads than something that's going to possibly exist in another two or three months.

The X299 platform with the lower core based CPUs (i5-7640X, i7-7740X, and i7-7800X) really doesn't make much sense when you sit down and do some thinking about it.

The issue with this mindset, is that these people compare consumer and HEDT platforms. talk about lol-fest. x299 life span is going to span multiple desktop generations. x299 isnt going to get coffee lake because its probably going to get its bigger brother. You dont even need to be an Intel fanboy to see thats how its always been.