Intel Haswell refresh and Devils Canyon reviews/info

[urbanfox]

In other news, #1 anniversary Pentium, week 32, does 4,72 GHz at 1,34 V aircooled. Post @ 4,8 with 1,42V, no post higher no matter what volts.

Cooler/temps?

[tool_462]

That's why I bring the wife and kids and we pay with cash. The first time we used a credit card and it was under our last name they refused to sell us.

But with cash, a phony name and phony address. Ka-Ching. No issues

MC in St. Louis Park, MN won't take cash for CPU's. At least the 2 times I tried they wouldn't. Wouldn't take VISA Cash Card either (that I bought with BTC) since it wasn't linked to a name and address.

[DedEmbryonicCe1]

MC in St. Louis Park, MN won't take cash for CPU's. At least the 2 times I tried they wouldn't. Wouldn't take VISA Cash Card either (that I bought with BTC) since it wasn't linked to a name and address.

What was their explanation for that policy? The one in Fairfax, VA certainly did a couple years ago when I bought a launch 1090T for ~$300 with cash. It might have something to do with returns but most places won't give you cash anyway, just store credit.

[G.Foyle]

Cooler/temps?

Cooler is Grand Kama Cross, temps are far from throttling (exact numbers, being a figment of DTS-es imagination, are disregarded).

[Charles Wirth]

My review sample goes a little something like this.
http://valid.canardpc.com/w90q4y

4Core / 8 Thread
5Ghz ambient F1 with fan 1.25vcore
5.5Ghz -20c 1.5vcore 1.8v input
5.9Ghz -100c 1.6vcore 2.2v input
6.0Ghz -130c 1.8vcore 2.3v input
6.1ghz -135c 1.85vcore 2.4v input
6.2ghz -140c 1.9vcore 2.5v input
6.3ghz -142c 1.91vcore 2.5v input (instant crash)

1.95vcore = instant crash @ -140c
2.6v input = instant crash

I need to test the memory controller more
Memory controller maxed out a little over 1.4v -130c 5.5Ghz, 1.4v @ 5.2Ghz seems to be a close max stable

Cold bug factory defaults in bios at -120c
Cold bug 5.5Ghz boot -140c
Cold bug 6Ghz 4c8t idle -141c ~ loaded -150c

You gotta get chip warmer to get it colder
It clocks better and goes colder with 4c8t, removal of HT and cores makes cold bug worse.

This chip is binned for review sample, it should be close to what you get from retail.
With the new chip you get more Mhz on air and water but you can see the wall trying to reach the last few multilples on the top end.
This chip is better than any retail 4770K that I have tested in the past year.

Two of the major changes help with getting all retail chips to 5ghz on air, The TIM and the added resistor packs on the back of the CPU allow more voltage and that pushes the chip a little harder to push past 6Ghz.

[DeathReborn]

No, non-K versions are plain old Haswell. K versions of those numbers are Devil's Canyon and require a BIOS update, which should be coming soon for Gigabyte Z87 boards.


In other news, #1 anniversary Pentium, week 32, does 4,72 GHz at 1,34 V aircooled. Post @ 4,8 with 1,42V, no post higher no matter what volts. Substrate is not new like Devil's Canyon, it's plain old 2C+GT2 substrate. TIM might be new, but I can't check on this CPU

Out of interest, what can it do @ 1.3v max? I ask because I am considering using one for a little gaming machine for my nephews (Lego games etc) and don't really want to go over 1.3v for them.

ALso thanks, I'll keep my 4770K 4.8Ghz 24/7 @ 1.25v for a while longer then, hoping for a straight cpu swap & 5-5.2GHz under water 24/7 from 4790K.

[FlanK3r]

My review sample goes a little something like this.
http://valid.canardpc.com/w90q4y

4Core / 8 Thread
5Ghz ambient F1 with fan 1.25vcore
5.5Ghz -20c 1.5vcore 1.8v input
5.9Ghz -100c 1.6vcore 2.2v input
6.0Ghz -130c 1.8vcore 2.3v input
6.1ghz -135c 1.85vcore 2.4v input
6.2ghz -140c 1.9vcore 2.5v input
6.3ghz -142c 1.91vcore 2.5v input (instant crash)

1.95vcore = instant crash @ -140c
2.6v input = instant crash

I need to test the memory controller more
Memory controller maxed out a little over 1.4v -130c 5.5Ghz, 1.4v @ 5.2Ghz seems to be a close max stable

Cold bug factory defaults in bios at -120c
Cold bug 5.5Ghz boot -140c
Cold bug 6Ghz 4c8t idle -141c ~ loaded -150c

You gotta get chip warmer to get it colder
It clocks better and goes colder with 4c8t, removal of HT and cores makes cold bug worse.

This chip is binned for review sample, it should be close to what you get from retail.
With the new chip you get more Mhz on air and water but you can see the wall trying to reach the last few multilples on the top end.
This chip is better than any retail 4770K that I have tested in the past year.

Two of the major changes help with getting all retail chips to 5ghz on air, The TIM and the added resistor packs on the back of the CPU allow more voltage and that pushes the chip a little harder to push past 6Ghz.

Seems very similar as my 4770K...I got 6.01 GHz, but I was limited in BIOS with input voltage :-/. Scaling seems similar (CPU voltage/VRIN/ frequency)

[zoob]

Review up at pcper
http://www.pcper.com/reviews/Process...d-Overclocking

[zalbard]

Review up at pcper
http://www.pcper.com/reviews/Process...d-Overclocking

Disappointing overclocking results.

However, temperatures seem to have considerably improved - so no need to delid at least. Although I find it strange that core 1 is so much hotter than others...

[dartaz]

According to intel website, i5-4690K and i7-4790K support VT-d
http://ark.intel.com/products/80811/...up-to-3_90-GHz
http://ark.intel.com/products/80807/...up-to-4_40-GHz

[vario]

According to intel website, i5-4690K and i7-4790K support VT-d
http://ark.intel.com/products/80811/...up-to-3_90-GHz
http://ark.intel.com/products/80807/...up-to-4_40-GHz

Wow, did not expect that, intel actually did a decent thing by NOT DISABLING it, before it was just a douche move because "well F U dear customer"

[zalbard]

According to intel website, i5-4690K and i7-4790K support VT-d
http://ark.intel.com/products/80811/...up-to-3_90-GHz
http://ark.intel.com/products/80807/...up-to-4_40-GHz

I guess this is just a spreadsheet error.

[FlanK3r]

reviews:

http://www.jagatreview.com/2014/06/h...devils-canyon/
http://hexus.net/tech/reviews/cpu/70...-22nm-haswell/
http://www.pcper.com/reviews/Process...d-Overclocking
http://pclab.pl/art58123.html

[LightSpeed]

So hexus decides to review the chip and not overclock it nor mention it in their conclusions. Hmm

[FlanK3r]

http://www.tomshardware.com/reviews/...ance,3845.html
http://www.hardocp.com/article/2014/...y#.U5eAhvl_s0I
http://www.vortez.net/articles_pages..._review,1.html

Next 3 reviews

[PatRaceTin]

Finally new CPU do 5GHz easily like my i5/i7-2X00K series.

[stasio]

Seems everybody for Reviews use same CPU batch.......L336D081...QG81 ES.
Only Hexus use different....L410C691...QG81 ES.

Still waiting for Retail results.

[FlanK3r]

And again next reviews

http://www.purepc.pl/procesory/intel...canyon_w_akcji
http://www.legitreviews.com/intel-co...-review_143880
http://www.guru3d.com/articles_pages...review,21.html

[pH(x)]

Disappointing overclocking results.

However, temperatures seem to have considerably improved - so no need to delid at least. Although I find it strange that core 1 is so much hotter than others...

I just started reading this paper on dim and dark silicon that may shed some light, no pun intended. It's at least very interesting.

http://cseweb.ucsd.edu/~mbtaylor/pap...n_dac_2012.pdf

[zalbard]

I just started reading this paper on dim and dark silicon that may shed some light, no pun intended. It's at least very interesting.

http://cseweb.ucsd.edu/~mbtaylor/pap...n_dac_2012.pdf

Nice! Looks interesting, thanks for sharing!

[paulbagz]

Anyone de-lidded these yet to see what's underneath?

-PB

[LightSpeed]

Finally new CPU do 5GHz easily like my i5/i7-2X00K series.

Most chips seem to max out at 4.8 ghz or so. Now if the retail samples are better ill have to replace the 4670k..

[massman]

Anyone de-lidded these yet to see what's underneath?

-PB

[FlanK3r]

nice, yours chips? Show the research

[PatRaceTin]

Look like thermal paste