Intel Demonstrates 65W Broadwell-K Socketed Processors at GDC 2015

[Dufus]

Very likely 14nm high performance is suffering from heat density problems (which is not a process problem) and intel needs some more refinement. That doesn't mean that its impossible or a failure.

That's a possibility. What can be done about it, lower the TDP and lose the gains in frequency improvement so it ends up being about the same frequency as previous lithography but at lower power. Or keep around the same package power by using more integration into the CPU and / or more cores.

Tjc thermal resistance seems to be getting higher as we get smaller dies. Perhaps a rough idea can be made from the temperature differential between working and idle cores or all cores and an idle integrated graphics temperature.

Since Intel have already posted in the ark a couple of BDW Xeon's I would think they would be out this month. Interesting both are spec'd 45W TDP while one is 4C/8T and the other 8C/16T with a Tcase spec of 80C. Both turbo to 2.6GHz with 2.2GHz and 2GHz HFM respectively. So something like 20% less core power for 8 vs 4 at the HFM which seems a bit odd.

 

[DrMrLordX]

Since Intel have already posted in the ark a couple of BDW Xeon's I would think they would be out this month. Interesting both are spec'd 45W TDP while one is 4C/8T and the other 8C/16T with a Tcase spec of 80C. Both turbo to 2.6GHz with 2.2GHz and 2GHz HFM respectively. So something like 20% less core power for 8 vs 4 at the HFM which seems a bit odd.

Those are Xeon-D chips, and they've been paper-launched at the very least.

 

[Dufus]

Those are Xeon-D chips, and they've been paper-launched at the very least.

Didn't I already say that? D-1500 series to be precise.

Since Intel have already posted in the ark a couple of BDW Xeon's I would think they would be out this month.

For anyone else interested.

Socket FCBGA 1667, DDR4
http://www.supermicro.com/products/motherboard/Xeon/D/X10SDV-F.cfm
Coming soon.

 

[Idontcare]

Are they going to sell any/many more desktops with Broadwell than they will with Haswell? Think about what Idontcare was saying, especially in the Q12015 financial thread: Haswell Refresh is only just now catching up with them.

Yep, there is only so may people (a limited quantity of TAM) who are interested in upgrading to a 2-yr old product that is Haswell on the eve of Intel supposedly launching both Broadwell and Skylake.

I'm running a Sandy Bridge laptop, numerous Ivy Bridge desktops, a Sandy Bridge desktop, and an FX-8350 desktop (don't laugh, but yes it is true).

Guess why I don't have a single Haswell inhouse? Rhetorical question of course, because who wants to invest in 2 year old technology today?

I'm waiting for Skylake to throw out the entire existing workhorse lineup (some 7 rigs in total) from laptop to desktop to dronetop*. And you can bet an appreciable portion of Intel's TAM is locked up waiting for the same.

* dronetop is my phrase for yesteryear's headless node, aka the standalone box that is monitor-less, keyboard-less, mouse-less, etc, you simply boot it and log into it, but you make it work for a living like it owes you rent

 

[witeken]

throw out the entire existing workhorse lineup (some 7 rigs in total) from laptop to desktop to dronetop*.

Why do you need 7 rigs?

 

[Idontcare]

Why do you need 7 rigs?

Forex backtesting optimization. I write code, computers process data through the code, magic happens, the planet gets hacked (but only just a little bit, honest) and bills get paid, yaddi yaddi.

 

[witeken]

Forex backtesting optimization. I write code, computers process data through the code, magic happens, the planet gets hacked (but only just a little bit, honest) and bills get paid, yaddi yaddi.

Vague. Don't you work for TI or so as semiconductor engineer somewhere in Asia?

 

[ClockHound]

Thought it was very explicit. IDC said yaddi yaddi, not yadda yadda...if I have to spell that out for you, it will be much too explicit.

Most everyone knows that computational yaddi requires at least 6 discrete Intel systems and 1 AMD system to proportionately simulate the market. *Sheesh*

;-)

 

[Idontcare]

Vague. Don't you work for TI or so as semiconductor engineer somewhere in Asia?

You asked what I do that needs 7 rigs, which is a different question from asking me what else I do that doesn't need 7 rigs

Start with MT4, use the client terminal to enable automated trading which entails writing code in what is generically called an Expert Advisor, which you (ideally) end up backtesting over years and years of historical market data for the purposes of both optimizing and debugging.

It is during that last phase, the backtesting, which you really need to through CPU cycles and cores at the problem.

A bit dated now, but you can read more about MT4, forex, and some of my personal history in this post from when I did my last round of hardware upgrades.

 

[witeken]

You asked what I do that needs 7 rigs, which is a different question from asking me what else I do that doesn't need 7 rigs

Does it make money?

 

[Dave2150]

Does it make money?

Would assume it doesn't make much, or he'd have dedicated server blades instead of old dated hardware in his basement.

 

[crashtech]

Hard for a computer to predict human irrationality.

 

[Idontcare]

Does it make money?

Of course, otherwise I wouldn't be doing it.

Hard for a computer to predict human irrationality.

The computer doesn't predict anything. The code doesn't write itself Someday maybe it will, but for now there is no skynet.

Anyways it is very much like the industry of weather modeling and forecasting.

They can't predict or forecast when or where the next hurricane/typhoon/cyclone will appear in a given ocean body, but once one begins to form they can do a reasonably good job of predicting/forecasting wind strengths, eye trajectory and path, and timeline.

Market analysis really isn't any harder, but to be sure you can be foolish about your approach to it and burn out from either ignorance or arrogance. We can't predict black swan events (when the next hurricane will develop) or market reversals but we can work within trends and channels pretty easily.

As for choice of hardware, amortization and depreciation are common business financials to manage, who goes out and buys hardware just because they have money in the bank? That isn't efficient use of capital.

 

[imported_ats]

Those are Xeon-D chips, and they've been paper-launched at the very least.

Xeon-D is in volume shipment, nothing paper about it. Partners have already announced some products and will be delivering to customers beginning of next month in the retail market. They are apparently already delivering on actual contracts.

 

[DrMrLordX]

Good. I don't know how long ago those parts were put on Intel's internal roadmaps (I don't remember seeing them anywhere on public roadmaps), but if this is an example of Intel going "agile" by reacting quickly to changing market conditions, then it's a positive sign.

 

[mrmt]

Good. I don't know how long ago those parts were put on Intel's internal roadmaps (I don't remember seeing them anywhere on public roadmaps), but if this is an example of Intel going "agile" by reacting quickly to changing market conditions, then it's a positive sign.

That's a 2013 presentation:

http://www.v3.co.uk/v3-uk/news/2284...lus-14nm-broadwell-and-denverton-server-chips

 

[beginner99]

Forex backtesting optimization. I write code, computers process data through the code, magic happens, the planet gets hacked (but only just a little bit, honest) and bills get paid, yaddi yaddi.

Forex.Interesting. Where you affected by the Swiss National bank decision? Positive or negative?

 

[Idontcare]

Forex.Interesting. Where you affected by the Swiss National bank decision? Positive or negative?

Neutral, but not because of any great insight on my part. With the CHF being pegged to the Euro there was no reason to trade CHF versus trading the EUR pair directly. (i.e. why trade CHFJPY when you can trade EURJPY and not deal with depegging risks?)

This is how your trading day went on Jan 15 if you were trading the EURJPY, avoiding the needlessness of trading a pegged currency:

^the red vertical line is Jan 15, the day the SNB made their announcement.

And this is what you got on Jan 15 if you were trading the pegged and correlated CHFJPY instead:

Lots of people in forex got hosed, but really only for silly reasons. Never trade a pegged currency, that's basic forex 101 stuff.

But legitimate businesses were truly harmed by the SNB's decision, but only if they weren't hedged. Most big businesses were hedged and fine, but small business owners don't have time for that, so they got hurt bad.

 

[DrMrLordX]

That's a 2013 presentation:

Okay, so it shows up as "Broadwell SoC". Interesting, I missed that one.

 

[Dave2150]

So, still no news on clock speeds, or anything Broadwell-k related, for that matter.

I guess we'll simply have to wait until June, and hope they release it then?

 

[III-V]

So, still no news on clock speeds, or anything Broadwell-k related, for that matter.

I guess we'll simply have to wait until June, and hope they release it then?

There's also Spring IDF coming up.

I don't think it's scheduled to release until summer anyway. At best, we'd get a bit of teasing a head of time.

 

[ehume]

So, still no news on clock speeds, or anything Broadwell-k related, for that matter.

I guess we'll simply have to wait until June, and hope they release it then?

The amazing thing is there are no leaks from their customers, no cracks in their NDA's.

 

[Idontcare]

The amazing thing is there are no leaks from their customers, no cracks in their NDA's.

been quite a dry spell for coolaler, that's for sure

 

[Dave2150]

The amazing thing is there are no leaks from their customers, no cracks in their NDA's.

Yes, indeed. I guess something changed at Intel HQ to discourage these leaks?

I'm assuming that all the motherboard, system integrator vendors etc have to had broadwell-K samples in order to update the Bios's etc.

We've had so many leaks in the past, what on earth has changed now? Maybe Intel punished the previous leaks/informers heavily? hmm.

 

[Idontcare]

Yes, indeed. I guess something changed at Intel HQ to discourage these leaks?

I'm assuming that all the motherboard, system integrator vendors etc have to had broadwell-K samples in order to update the Bios's etc.

We've had so many leaks in the past, what on earth has changed now? Maybe Intel punished the previous leaks/informers heavily? hmm.

What's there to leak that everyone doesn't already know?

14nm is super healthy, BW binning is spitting out 5GHz+ monsters, Intel doesn't want to do a paper-launch because they know the world+dog is going to want to upgrade immediately so they have spent the past 9 months stockpiling desktop skus.

Oh, and while we are living in fantasy land here for the moment, the S810 has no overheating issues, Qualcomm is just doing 20nm like a boss, and TSMC will never make another Apple chip again.

Its all true, I read it in my post on the internet after I pressed the Submit Reply button once.