Why I won't recommend a Haswell system until August+

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Why I won't recommend a Haswell system until August+

Post by Vicotnik » Fri Jun 07, 2013 12:39 am

yakuman wrote:Really disappointing. Haswell was my planned build because of Ivy Bridge's heat problem. It's ironic that die shrinks are supposed to reduce heat, but each preceding generation has been increasing heat :x. On the other hand I'm kind of tired of waiting for the next generation.
Heat and temperature is not the same thing. A die shrink often reduces heat but increases temperature; less heat but smaller area.

That Intel uses crappy TIM once again, making overclocking more difficult is another story. And really disappointing as you say.

yakuman
Posts: 140
Joined: Thu Jun 07, 2012 3:05 am
Location: Canada

Re: Why I won't recommend a Haswell system until August+

Post by yakuman » Fri Jun 07, 2013 2:46 am

Vicotnik wrote:Heat and temperature is not the same thing. A die shrink often reduces heat but increases temperature; less heat but smaller area.

That Intel uses crappy TIM once again, making overclocking more difficult is another story. And really disappointing as you say.
If each die shrink raises temperature, then Haswell will be cooler than Broadwell, unless they fix the TIM / IHS. I'm never interested in overclocking anyway. What's unsettling is the reports of higher temperatures even at stock idle.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Re: Why I won't recommend a Haswell system until August+

Post by NeilBlanchard » Fri Jun 07, 2013 5:18 am

Higher temperatures do not necessarily mean more heat. The power consumption roughly equates to heat, but with a die shrink, the smaller surface area means that heat has to dissipate over a smaller area - which means that you will see similar temperatures with less heat and less energy used.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Fri Jun 07, 2013 9:47 am

yakuman wrote:
Vicotnik wrote:If each die shrink raises temperature, then Haswell will be cooler than Broadwell, unless they fix the TIM / IHS.
There's more than one variable here:
- smaller process features typically leads to less current used per transistor* (- delta T)
- smaller process may lead to lower core operating voltages (- delta T)
- tweaks to architecture may lower power use (- delta T)
- decreased die area leads to higher power density (watts/sq mm) (+ delta T)
- add features -> more transistors -> more power ( + delta T)
- add features -> more transistors -> increases die area -> lower power density (- delta T)

So, it's not a given that the next shrink will increase die temps.

* up to a point where leakage current takes over - and that's why Intel moved to FinFet as well as some other stuff.

ceh4702
Posts: 6
Joined: Sat Mar 23, 2013 7:25 am

Re: Why I won't recommend a Haswell system until August+

Post by ceh4702 » Fri Jun 07, 2013 2:52 pm

A die shrink does not necessarily equate to lower temp. Even if it is more efficient, often they add more transistors and the surface of the cpu decreases. They also increased the capabilities of the GPU, which could create more heat. It seems like every time there is a die decrease that the processor is also using more watts, which creates heat. It is like Intel uses a formula that calculates how many extra things they can do and still keep the processor within their targeted thermal limits.

I was hoping some motherboards would try to use DDR4 sooner than expected. I saw an article that looked like Micron was trying to cash in on eary deployment of DDR4. Still no agreed upon DDR4 standard. I could envision everyone buying new motherboard and on the first revision around Christmas or the first quarter and all the new motherboards would be DDR4 and everyone would then want a new motherboard.

I took a double look at this XEON Co-Processor thingy. Intel seems to be doing a lot more interesting things with the XEONS. This probably comes at a premium price also, but the low end E3 is very similar to an i5/i7.

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: Why I won't recommend a Haswell system until August+

Post by QUIET! » Fri Jun 07, 2013 9:20 pm

Memory interfaces are more than just a connector and circuit board traces.

A memory interface change would be pretty much an entirely new processor (like the Haswell extreme which I think may have DDR4 but I'm not paying $1k for).

As for heat and die shrinks, transistors kept getting more and more efficient until leakage became as issue, then the fin fet was invented which essentially eliminates leakage but since the active area is an incredibly thin fin instead of the surface of a comparatively enormous wafer, all the heat generated in the fin is trapped and you'll cook the gate oxide if you run it too hot.

Its an incredibly small problem but you'll have about 4 billion of them...

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Re: Why I won't recommend a Haswell system until August+

Post by nutball » Sun Jun 09, 2013 4:59 am

yakuman wrote:If each die shrink raises temperature, then Haswell will be cooler than Broadwell, unless they fix the TIM / IHS. I'm never interested in overclocking anyway. What's unsettling is the reports of higher temperatures even at stock idle.
Why are you worried about idle temperatures? Why would you be concerned about higher temperatures if Intel aren't? Do you think that Intel don't know what they're doing?

Pappnaas
Posts: 726
Joined: Mon May 16, 2011 11:23 am
Location: Germany

Re: Why I won't recommend a Haswell system until August+

Post by Pappnaas » Sun Jun 09, 2013 6:52 am

The die shrink doesn't raise or lower the cpu temps, the complete package does. So if intel does a better job at cooling the next die step, then it will run cooler. See all those before/after delidding reports.

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: Why I won't recommend a Haswell system until August+

Post by QUIET! » Sun Jun 09, 2013 9:25 am

nutball wrote:
yakuman wrote:If each die shrink raises temperature, then Haswell will be cooler than Broadwell, unless they fix the TIM / IHS. I'm never interested in overclocking anyway. What's unsettling is the reports of higher temperatures even at stock idle.
Why are you worried about idle temperatures? Why would you be concerned about higher temperatures if Intel aren't? Do you think that Intel don't know what they're doing?
Intel knows what they are doing but their motives are questionable.

They probably use crummy TIM to save 50¢ each just because they can. Its not a big deal unless you bought a k processor for over clocking...

There is a pretty large gap between what Intel could do and what they actually sell (no vt-d on k processors, wtf?).

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Re: Why I won't recommend a Haswell system until August+

Post by nutball » Sun Jun 09, 2013 12:21 pm

QUIET! wrote:Intel knows what they are doing but their motives are questionable.
Their motives aren't questionable at all. They may not be very agreeable, but they are very clear - make as much money as possible for as little outlay - which is what any business should do, including their competitor if they want to survive.
They probably use crummy TIM to save 50¢ each just because they can. Its not a big deal unless you bought a k processor for over clocking...
*shrugs* To me overclockers seem to overestimate their share of the market. It's a marvel to me that Intel sell unlocked processors at all.
There is a pretty large gap between what Intel could do and what they actually sell (no vt-d on k processors, wtf?)
Yes there are some bizarre combinations of features in their line-up. Most mainstream uses of virtualization are in environments where over-clocking is inappropriate though - servers and the workplace in general.

Hobbyist virtualisers are a small fraction of the market I would guess, hobbyist over-clockers are a small fraction of the market, so hobbyist virtualising over-clockers are likely a small fraction of a small fraction of the market. So if Intel, say, have to decide whether to spend the money to validate Vt-d on over-clocked processors, they may decide it's not worthwhile.

Who knows. It's not like disabling Vt-d on the 4770K is going to push people to an even more expensive part in the Haswell range. Maybe they're trying to shift piles of unsold SNB-Es.

piglover
Posts: 134
Joined: Mon Jun 08, 2009 9:32 am
Location: California

Re: Why I won't recommend a Haswell system until August+

Post by piglover » Sun Jun 09, 2013 2:28 pm

QUIET! wrote: Intel knows what they are doing but their motives are questionable.

They probably use crummy TIM to save 50¢ each just because they can. Its not a big deal unless you bought a k processor for over clocking...

There is a pretty large gap between what Intel could do and what they actually sell (no vt-d on k processors, wtf?).
I don't think Intel's "motives" are questionable or even disagreeable. It is clear - they sell a product with a published spec and they deliver it in a way that meets that spec. Period.

"Overclocking", by definition, means using the product beyond what it was designed for or advertised as delivering. I do think Intel has some passing interest in feeding the enthusiast community. But I don't share your view that they owe any kind of "obligation" to spend extra money making their product more than it is sold as. Keep in mind that the 50 cents a unit cost you think they should spend adds up to $many $millions spent so that a few hundred people worldwide can amuse themselves.

Overclocking is supposed to require some knowledge, skill and work - its not supposed to be easy. And if it requires the knowledge, skill and work to pull the heat spreader and replace the TIM with a better product then so be it. Rather than griping that you have to you should be taking pride in the fact that you can (and that you have the stones to try).

As for the "no VT-d on K" i think the real WTF is why do they deliver VT-d on any non-xeon processor at all. It is a non-mainstream feature and doesn't really belong on their laptop/desktop products.

QUIET!
Posts: 238
Joined: Sat May 11, 2013 8:33 am

Re: Why I won't recommend a Haswell system until August+

Post by QUIET! » Mon Jun 10, 2013 12:21 am

A lot of people would happily pay $10 more to get Sandy Bridge style IHS attach on k processors or non-k for that matter.

Heat=resistance=more power required=more heat so if I can put a huge tower cooler on an i3 and idle near ambient, that's what I want to do even if I can't over clock. Better IHS attach would help.

Plenty of products are sold based on potential rather than advertised usage, like the Hitachi Magic Wand, Phillie Blunts, Playgirl magazine and you know people stick q-tips in their ears even though the package has a warning.

I bought a k processor because I wanted top of the line, not over clocking and what I found was disabled features. I wound up replacing it with a Xeon but I shouldn't have had to.

Mostly its bad marketing combined with silly me assuming that a top of the line CPU would have all features activated.

Bottom line, there are several products that Intel could easily make and find a market for but they aren't available at any price unless you can get your hands on an unlocked engineering sample chip.

Their not alone, I wonder why Ford won't make a US 3 door Fiesta. I don't want 5 doors but I'll probably get one anyway and I can only be jealous of the countries that have that model.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Thu Jul 18, 2013 9:13 am

C2 Stepping available July 29

Probably won't see C2 stepping mobo's until Sept/Oct

:lol:

Sort of the difference between when Intel provides the first qual'd product in volume to mobo mfgrs and when the supply chain runs out of the C1 stepping version. :)

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

Re: Why I won't recommend a Haswell system until August+

Post by cordis » Thu Jul 18, 2013 10:50 pm

Eh, haswell isn't so bad. I've got an i7 4770k overclocked up to 4.2 Ghz now, the temps get up into the mid 70s, but so far it seems pretty stable. I'm also using the built in video, so that might be adding some extra heat. But so far, it seems fine.

T K
Posts: 48
Joined: Tue Jun 18, 2013 2:32 pm

Re: Why I won't recommend a Haswell system until August+

Post by T K » Fri Aug 02, 2013 3:39 am

CA_Steve wrote:Supposedly, it just affects some USB 3 devices like thumb drives and not keyboard/mouse. That said, I use sleep mode all the time during the day. Before I walk away from my PC, I put it to sleep which dismounts my Truecrypt drive along the way.
How about exTernal
- HDDs
- Optical DVD Drives
?

andymcca
Posts: 404
Joined: Mon Jan 11, 2010 8:19 am
Location: Boston, MA, USA

Re: Why I won't recommend a Haswell system until August+

Post by andymcca » Fri Aug 02, 2013 5:58 am

Do Win* and Linux guarantee a write cache flush before entering sleep?

If not, this could mean data loss on thumb drives.

Edit: Also, I'm too lazy to look. Is this an S3 only problem, or does it affect S5 too (probably not)?

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Why I won't recommend a Haswell system until August+

Post by Abula » Sat Aug 03, 2013 7:58 am

Just FYI, seems the C2 stepping will come when ever the manufacturer decides, here is an example from asus and RAJA an asus representative for forums,

Maximus VI Formula Discussion Thread

This and the impact are new boards presented on Computex but were delayed from the initial haswell launch, many believed with this delay they were going to be C2 stepping, and its partially true, the impact will be C2, and the Formula will be C1,
1. Getting a pic/info for you: I'ts an alu heatsink that has an anodized coating in the water channel from what I just got told.

2. Should be C2 as far as I am aware (will get confirmation). Impact is C2, so I would have thought this will be as well. EDIT: Formula will be C1 for a while. Impact will be all C2.

-Raja

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Mon Aug 05, 2013 9:10 pm

Asus C2 stepping mobo part numbers - note the "5" on the end of the part number. If it's a "0" it's C1 stepping.
Full list at PC Games Hardware. This matches with what's listed at Geizhals.at (thanks, Boost).

laststop
Posts: 123
Joined: Sun Apr 15, 2012 12:37 am
Location: Youngstown, OH

Re: Why I won't recommend a Haswell system until August+

Post by laststop » Sun Aug 25, 2013 11:44 am

Im interested in the Asus Z87 Expert with 1x TB port for my next build. Is it safe to assume if I buy one from newegg I will get the newest revision? Is there anything special I need to do to ensure that?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Sun Aug 25, 2013 12:37 pm

I think it's the opposite of safe. :)

- Newegg doesn't list Asus part numbers...just the model number and Newegg's inventory SKU.
- You won't be able to get anyone to tell you what they carry ( I tried, and I think I got a gerbil with opposable thumbs to reply).
- which means you won't know until you receive the board and look at the part number.
- and if it's the old version, it's a restocking fee for you.

laststop
Posts: 123
Joined: Sun Apr 15, 2012 12:37 am
Location: Youngstown, OH

Re: Why I won't recommend a Haswell system until August+

Post by laststop » Sun Aug 25, 2013 12:45 pm

That is really frustrating. What do you recommend to be sure I get the latest revision? Is there another store that will accommodate this? I'll lose out on the combo deal but I don't really care I'd rather pay the extra 55 and get a guaranteed c2 stepping or even just go with the i5-4670k.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Sun Aug 25, 2013 1:53 pm

Walk into a brick and mortar store and look at the part number on the box. Or, find an e-tailer that shows the Asus part number.

laststop
Posts: 123
Joined: Sun Apr 15, 2012 12:37 am
Location: Youngstown, OH

Re: Why I won't recommend a Haswell system until August+

Post by laststop » Sun Aug 25, 2013 4:07 pm

I'm gonna try my luck with newegg I will be sure to report what I get over here. So I can help the next person in my situation.

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Why I won't recommend a Haswell system until August+

Post by Abula » Sun Aug 25, 2013 5:25 pm

Newegg has some asus motherboards that are C2 stepping, but not all, so atm is still a gamble.

laststop
Posts: 123
Joined: Sun Apr 15, 2012 12:37 am
Location: Youngstown, OH

Re: Why I won't recommend a Haswell system until August+

Post by laststop » Sun Aug 25, 2013 6:58 pm

Even if it's not c2 it won't be the end of the world. Infact I dont even use sleep or hibernate or any of that. I just let my pc idle on the desktop fully awake when i'm not on it. I restart it when windows forces me to restart for updates that's it.

whispercat
Posts: 376
Joined: Tue Jul 22, 2008 6:05 pm
Location: Calgary, AB, Canada

Re: Why I won't recommend a Haswell system until August+

Post by whispercat » Mon Sep 02, 2013 5:11 am

So is it safe to buy a Haswell system yet? :mrgreen:

I'm looking for a low power draw at idle, energy efficient ATX desktop mobo with a good IGP. Does Intel make Haswell motherboards?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Mon Sep 02, 2013 6:34 am

whispercat wrote:So is it safe to buy a Haswell system yet?
Sure, the trick is to find a mobo with the C2 stepping. :)

There are 4 Intel Haswell boards. This is a nice article comparing 33 mobo's and power use.

laststop
Posts: 123
Joined: Sun Apr 15, 2012 12:37 am
Location: Youngstown, OH

Re: Why I won't recommend a Haswell system until August+

Post by laststop » Mon Sep 02, 2013 9:50 pm

I can tell you right now the asus z87 expert with 1 thunderbolt 1.0 port from newegg is not shipping the c2 stepping yet. I am calling them on tuesday and fighting against the restock fee if I can take store credit towards the asus deluxe quad with 2x thunderbolt 2.0 (20gbps). I am going to tell them they sold me a defective board with a known defective issue and I should not have to pay the restocking fee. I'll pay the shipping back but you better believe I will get this restocking fee removed. I have spent over 10,000 dollars with this company they better do me this favor.

Live steve said he was right you will need to wait another month for this inventory to be flushed. If you plan on reselling your PC later like I am you really want a c2 stepping. Asus Z87 deluxe quad should be released soon and it will be c2 stepping from the start AND it will be the first motherboard with 20gbps thunderbolt 2.0 and 2x TB 2.0 ports at that. They will also support up to 4096x2160 @ 60 hz.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Why I won't recommend a Haswell system until August+

Post by CA_Steve » Sat Sep 21, 2013 10:44 am

Asrock putting "C2 stepping" stickers on mobo boxes. Don't think they updated the part number.
Image

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Why I won't recommend a Haswell system until August+

Post by Abula » Sat Sep 21, 2013 6:04 pm

wrong thread... opss.

Post Reply