[Canard PC Hardware] Intel prepares Ryzen's response behind the scenes

[Jan Olšan]

IPC is a measure of the single-threaded performance of a single CPU core.

Just for the sake of correctness... Last time I checked (in 2006?), IPC stands for "instructions per cycle". As such it by definition can't be a measure of single thread single core performance, since that also depends on a second, unrelated factor: the actual clock of the core.
Obviously, CPUs with high IPC can have their single-core speed hampered by low clock.

Also you could argue that when discussing performance of a *single core* per clock, HT/SMT should be counted in, because even if it is additional threads, it is still extra performance generated by the same core from the same clock cycles. But since single thread performance is very important, I can agree with *single-thread* IPC being singled out.

 

[Insert_Nickname]

Don't confuse the enthusiast market with the mainstream market. I do personally believe that currently (and some years to that) 4 cores is more than enough for the mainstream market. The mainstream market just isn't starved for CPU performance. The additional 4 cores aren't going to make Microsoft Office run any better, nor is it going to make Facebook run any smoother. The total CPU performance just isn't the determining factor in the mainstream market right now. Maybe in some years, but that is hard to tell.

Raven Ridge will be the big mainstream competitor to Intels mainstream lineup.

I would say 4 real cores (excluding Atom based ones) are the minimum one should expect on the desktop today. If nothing else, extra cores help tremendously with all the mainstream bloatware people insist on running in the background. This doesn't affect enthusiasts, since we know what is running on our PC (hopefully... ) and know how to keep them clean. Your average consumer, not so much.

Mobile isn't so clear cut, but lower clocked duals can choke on even simple web-browsing. Don't even get me started on the small cores, they're downright atrocious to use.

 

[2Dtails]

I would say 4 real cores (excluding Atom based ones) are the minimum one should expect on the desktop today. If nothing else, extra cores help tremendously with all the mainstream bloatware people insist on running in the background. This doesn't affect enthusiasts, since we know what is running on our PC (hopefully... ) and know how to keep them clean. Your average consumer, not so much.

Mobile isn't so clear cut, but lower clocked duals can choke on even simple web-browsing. Don't even get me started on the small cores, they're downright atrocious to use.

A dual core with hyper-threading is in my opinion the lowest you should expect. It should be able to handle the default bloat that followed with the machine, and still be up for the task.

Atoms, currently, is a dead end. Performance is sub-par, the only reason I see someone go with it, is its efficiency or cost.

Mobiles have low base, and high turbo for that reason. I personally haven't had much issues with a dual core (with hyper-threading) running a bunch of tabs from anything from video running to a lot of webforums and web-games.

 

[LTC8K6]

OK, so now I think we can say that Intel needs to respond to RyZen, and we can also say that the 7740K and 7640K were not responses to RyZen.

So we can move on to discussing what Intel's response might be.

I think it will have to be price cuts initially.

Unless they have been hiding something from us, it doesn't seem like they could have a new chip anytime soon.

 

[Arachnotronic]

OK, so now I think we can say that Intel needs to respond to RyZen, and we can also say that the 7740K and 7640K were not responses to RyZen.

So we can move on to discussing what Intel's response might be.

I think it will have to be price cuts initially.

Unless they have been hiding something from us, it doesn't seem like they could have a new chip anytime soon.

LTC8K6, would you like to create a new thread for that? This thread is awfully cluttered and kind of unfocused.

I think that would be a great discussion.

 

[LTC8K6]

OK.

 

[Lidiya Nikolayev]

Yes, that is the biggest issue with the graph. So misleading, but smoke and mirrors with respect to process tech is Intel's game now.

What are the major hurdles that have kept Intel from considering opening their own data Center and selling space to companies ? Why isn't intel selling cloud space with their own data center ?

 

[itsmydamnation]

What are the major hurdles that have kept Intel from considering opening their own data Center and selling space to companies ? Why isn't intel selling cloud space with their own data center ?

Intel dont sell data centres, they have Fab's and they have tried really hard to be a foundry but failed. There is a simple reason why, 3rd party IP. Almost all 3rd party IP is built for TMSC, only a select few companies can build a chip on intels fabs without the 3rd parties porting to intels fabs(apple, FGPA makers etc) . So far the 3rd parties haven't and there is no sign of that changing.

 

[tamz_msc]

Intel dont sell data centres, they have Fab's and they have tried really hard to be a foundry but failed. There is a simple reason why, 3rd party IP. Almost all 3rd party IP is built for TMSC, only a select few companies can build a chip on intels fabs without the 3rd parties porting to intels fabs(apple, FGPA makers etc) . So far the 3rd parties haven't and there is no sign of that changing.

That would be an outcome of their business model that they have been clinging to for a long time. Even NVIDIA has made a better attempt at diversification than Intel.

 

[Lidiya Nikolayev]

Intel dont sell data centres, they have Fab's and they have tried really hard to be a foundry but failed. There is a simple reason why, 3rd party IP. Almost all 3rd party IP is built for TMSC, only a select few companies can build a chip on intels fabs without the 3rd parties porting to intels fabs(apple, FGPA makers etc) . So far the 3rd parties haven't and there is no sign of that changing.

Thank you for explaining why they aren't taking that path. It makes sense now.

 

[Phynaz]

Intel dont sell data centres, they have Fab's and they have tried really hard to be a foundry but failed.

Just as an FYI, Intel's 3rd party foundry business is not a failure. They have publicly stated that they will not be a general purpose foundry, they play at the high end of the foundry business.

Notice the name of their foundry business is "Intel Custom Foundry". Custom isn't a high volume business.

 

[jpiniero]

Just as an FYI, Intel's 3rd party foundry business is not a failure. They have publicly stated that they will not be a general purpose foundry, they play at the high end of the foundry business.

Notice the name of their foundry business is "Intel Custom Foundry". Custom isn't a high volume business.

That sounds like spin from Intel though that they are unable to get Apple or Qualcomm or someone of that like.

 

[Phynaz]

That sounds like spin from Intel though that they are unable to get Apple or Qualcomm or someone of that like.

If somebody wanted Intel to manufacture ARM they could, Intel is partnered with ARM. It's doubtful that anyone would want to pay Intel's rates though - outside of the US military at least. You can bet the US Govt is using Intel fabs for manufacturing, that's actually how the whole Altera acquisition took place.

 

[Nothingness]

If somebody wanted Intel to manufacture ARM they could, Intel is partnered with ARM. It's doubtful that anyone would want to pay Intel's rates though - outside of the US military at least. You can bet the US Govt is using Intel fabs for manufacturing, that's actually how the whole Altera acquisition took place.

LG seems to be an ICF 10nm customer: https://newsroom.intel.com/editorials/accelerating-foundry-innovation-smart-connected-world/
That surely is with some ARM CPU.

 

[lopri]

LG looked like a 22nm customer.. and a 14nm customer.. Somehow things did not work out. LG is desperate - see the recent news regarding S835 shortage and Samsung exclusivity - and my guess is that it would have already used Intel if ___________________ . (insert a reason you can think of) Something did not work out and it was not because of lack of willingness.

 

[Lidiya Nikolayev]

LG seems to be an ICF 10nm customer: https://newsroom.intel.com/editorials/accelerating-foundry-innovation-smart-connected-world/
That surely is with some ARM CPU.

Thank you for the link, I will read this.

 

[tamz_msc]

LG seems to be an ICF 10nm customer: https://newsroom.intel.com/editorials/accelerating-foundry-innovation-smart-connected-world/
That surely is with some ARM CPU.

About damn time. Should be interesting to see what LG can do. Perhaps we'll see it in the G7? OEMs are not happy with Samsung hogging the supply of the 835. I expect more to follow suit.

 

[Andrei.]

LG looked like a 22nm customer.. and a 14nm customer.. Somehow things did not work out. LG is desperate - see the recent news regarding S835 shortage and Samsung exclusivity - and my guess is that it would have already used Intel if ___________________ . (insert a reason you can think of) Something did not work out and it was not because of lack of willingness.

They were just late and uncompetitive against Qualcomm, simple as that. Maybe third time's the charm.

 

[Arachnotronic]

That sounds like spin from Intel though that they are unable to get Apple or Qualcomm or someone of that like.

It kind of is. Intel isn't likely to win Apple, TSMC is a very strong competitor and it has built up a lot of trust with Apple -- TSMC nailed it with 20nm, 16FF+, 16FFC, and they look on track to be ready for 10nm and 7nm.

That kind of trust is important.

Intel, by contrast, has issues ramping new new process technologies. Until Intel's execution improves and is consistently excellent for several generations, if I were running the foundry selection show at Apple, I'd steer clear.

 

[Arachnotronic]

Just as an FYI, Intel's 3rd party foundry business is not a failure. They have publicly stated that they will not be a general purpose foundry, they play at the high end of the foundry business.

Notice the name of their foundry business is "Intel Custom Foundry". Custom isn't a high volume business.

Intel has waffled on this. PSO really did not like the idea of building chips for competitors. When BK came in, his proclamation was, "open the foundry to all who can use it." However, now Murthy is now saying that Intel has no intention to be a GP foundry and wants to partner strategically to boost its IP/product portfolio.

In other words, right back to Intel's view under PSO.