Where is Anandtech's review of Devil's Canyon??

[ViRGE]

These cpu's live on the edge at stock. Currently the stock turbo boost from 4.0 is 4.4/4.3/4.2/4.2. I thought I would just bump it to 4.4 even (without change of voltage) and that was enough to blue screen under stress.

That's fairly typical, no? If you go to 4.4 flat, then you have all 4 cores going at 4.4 versus just the single core, which would increase vdroop at that frequency. While your voltage is staying high enough under the load of 1 core, it's probably sagging a bit too much for that frequency under 4.

 

[Homeles]

I meant 4 Ghz.

They're guaranteeing that it'll turbo up to 4.4 from 4.0 stock for 3 years, essentially.

 

[mikk]

GPU tests from anandtech are GPU limited as usual, what a waste this is for a CPU test.

 

[BFG10K]

These cpu's live on the edge at stock. Currently the stock turbo boost from 4.0 is 4.4/4.3/4.2/4.2. I thought I would just bump it to 4.4 even (without change of voltage) and that was enough to blue screen under stress.

If you want to do that you need to leave the voltage on auto.

Even at stock settings my motherboard turbos all four cores to 4.4 GHz, and I presume it raises the voltage a bit (I haven't checked). Actually I have it set to a "Green" profile right now which I believe slightly undervolts.

 

[ShintaiDK]

You would have to be a bit more precise. There are a lot of graphs and a lot of individual results on that page.

I see most is fixed now. But there is still results that doesnt fit in.

The 4690 and 4690K and stock should perform exactly the same. However the K model is roughly 10% faster. And the 4.7Ghz 4690K is not much faster than the stock 4690K(~)5%, yet its supposed to have a 800-1000Mhz lead depending on threads used. And the benchmark shows its CPU limited.

These numbers simply dont add up for the 47xx chips either. The 4790K being ~40% faster, yet only ~10% higher clock.

There is a few more, that could be explained by a large benchmark variance if you only ran the benchmark once.

 

[ShintaiDK]

If you want to do that you need to leave the voltage on auto.

Even at stock settings my motherboard turbos all four cores to 4.4 GHz, and I presume it raises the voltage a bit (I haven't checked). Actually I have it set to a "Green" profile right now which I believe slightly undervolts.

Its worrying that some mobo makers still autooverclock peoples CPUs out of the box with the loss of warranty to follow.

 

[StinkyPinky]

I've been playing around with the software that came with the board and it's actually very good.

I set the CPU voltage to auto and changed the cores to 4.5 and stress tested it for a few hours and the voltage never went above 1.215 which I'm happy with. Maybe things have changed but a few years ago leaving a board on auto would result it in feeding way too much voltage to the CPU. The temps were only early 80's so that gives me a bit of leg room. Especially since these stress tests heat up the cpu more than normal heavy use would anyway.

 

[BFG10K]

Its worrying that some mobo makers still autooverclock peoples CPUs out of the box with the loss of warranty to follow.

I updated the BIOS today and checked it again. When you enable XMP it asks if you want to boost the turbo (i.e. use 4.4 GHz on all cores). I obviously answered yes before.

I use XMP to get 1600MHz 1.35V which is much better than the stock JEDEC 1.5V 1333MHz profile.

The auto voltage is 1.189V when running LinX (i.e. worst case), so I'm leaving it like that. Technically I'm overclocking, but technically each core is guaranteed to hit that speed, too. Plus my low voltage RAM should ease the load a little.

I like Asus' UEFI interface. :thumbsup:

 

[WhoBeDaPlaya]

I set the CPU voltage to auto and changed the cores to 4.5 and stress tested it for a few hours and the voltage never went above 1.215 which I'm happy with. Maybe things have changed but a few years ago leaving a board on auto would result it in feeding way too much voltage to the CPU. The temps were only early 80's so that gives me a bit of leg room. Especially since these stress tests heat up the cpu more than normal heavy use would anyway.

1.215V is already plenty to cause near-meltdowns in bad 4770Ks. Also, watch out for AVX workloads that cause Vcore to spike.

 

[blackened23]

I'm not so sure. Voltage threshold has increased on the Z97 chipset, numerous Z97 motherboard reviews have noted that they have been able to go up to 1.4V with both the 4770k and 4790k on the Z97 chipset and the temperatures are different than what Z87 chipset boards output. On z87, yes, 1.25V is about the point to where you want to start being careful. Apparently Z97 does something different on a chipset level that allows voltage threshold to go a bit higher.

Why? I don't know. But many reviews have pointed out temp differences between Z97 and Z87. HardOCP's Asus Z97 pro review is one that I remember talking about this in particular. Their limit on Z87 was around 1.25, Z97 increased the threshold. I have no idea why or how.

 

[cytg111]

I'm not so sure. Voltage threshold has increased on the Z97 chipset, numerous Z97 motherboard reviews have noted that they have been able to go up to 1.4V with both the 4770k and 4790k on the Z97 chipset and the temperatures are different than what Z87 chipset boards output. On z87, yes, 1.25V is about the point to where you want to start being careful. Apparently Z97 does something different on a chipset level that allows voltage threshold to go a bit higher.

Why? I don't know. But many reviews have pointed out temp differences between Z97 and Z87. HardOCP's Asus Z97 pro review is one that I remember talking about this in particular. Their limit on Z87 was around 1.25, Z97 increased the threshold. I have no idea why or how.

That is very interresting! Sort of the thing you'd like your favorite hardware site to dig into ... (hint hint)..

 

[ehume]

Covert Adaptive Voltage? Do the MB's do this on normal+V, or do they also do this with fixed Vcore?

 

[WhoBeDaPlaya]

Why? I don't know. But many reviews have pointed out temp differences between Z97 and Z87. HardOCP's Asus Z97 pro review is one that I remember talking about this in particular. Their limit on Z87 was around 1.25, Z97 increased the threshold. I have no idea why or how.

Interesting. Hadn't bothered reading the Z97 reviews.
Fairly certain that the Z87 isn't doing anything out of the ordinary (logged Vcore using a scope).

When I tell my Z87-GD65s to be fixed, the Vcore remains fixed under any and all workloads (albeit with Vdroop). Then again, I always run my CPUs at fixed voltage, fixed multi, and all C-states disabled.

 

[crashtech]

I'm not so sure. Voltage threshold has increased on the Z97 chipset, numerous Z97 motherboard reviews have noted that they have been able to go up to 1.4V with both the 4770k and 4790k on the Z97 chipset and the temperatures are different than what Z87 chipset boards output. On z87, yes, 1.25V is about the point to where you want to start being careful. Apparently Z97 does something different on a chipset level that allows voltage threshold to go a bit higher.

Why? I don't know. But many reviews have pointed out temp differences between Z97 and Z87. HardOCP's Asus Z97 pro review is one that I remember talking about this in particular. Their limit on Z87 was around 1.25, Z97 increased the threshold. I have no idea why or how.

Found this odd, but it is corroborated:

From the ASUS Z97 Deluxe LGA 1150 Motherboard Review on HardOCP:

In manual mode things were quite different. On a Z87 motherboard using this CPU I need about 1.285v max to get a 4.7GHz overclock out of the system. I typically need to adjust power phase settings or load-line calibration to near maximum but not always. With the Z97 Deluxe I found out very quickly that 1.285v was not going to work. Instead I needed to use 1.365v or higher to achieve any kind of stability past 4.5GHz. Here is the interesting part, the system actually ran the CPU cooler than it would on a Z87 motherboard. On a Z87 motherboard that type of voltage setting would have me hitting the wall for thermal throttling quickly. I’d be looking at 90c+ temperatures with those settings. On Z97 I had sub-80c temperatures.

What’s even more interesting about this is that 4.5GHz using the auto-tuning feature achieved that result at 1.28v which is what would normally get me to 4.7GHz on most systems. On this one 4.7GHz eluded me. I know this CPU can do it but either BIOS maturity or voltage design is holding me back. If it’s the latter it’s because I suspect this is designed more for Haswell’s successor in mind than current Haswell CPUs. For now the best I could achieve is 4.6GHz/DDR3 1866MHz at 1.38v. And again it was fairly boring as all I did was play with one setting and that’s the CPU voltage.

This would indicate that there is something going on that is not being properly understood and/or measured.

 

[blackened23]

There's something different going on with Z97 and it isn't just that asus board - i've seen it mentioned across various reviews of different Z97 boards. The voltage threshold is higher. BUT, the weird thing is, the maximum overclock is generally no different between Z87 and Z97. I really don't know how to explain it and I haven't seen an explanation anywhere. You would normally think a higher voltage threshold means a higher overclock. But for some reason, it doesn't. It does mean you can kick a 4770k up to 1.35V+ without issue (using aftermarket cooling) on a Z97, but that definitely WAS NOT the case on Z87. Z87 had a preferred range of 1.2V-1.25V if using an aftermarket cooler (1.2V) or a closed loop like the H100 (a little higher, 1.25V ).

I think anyone buying a new system would get Z97 anyway, but it would be interesting if anyone could investigate and explain the voltage threshold differences between the two chipsets. I really can't think of an explanation since it seemingly happens on all Z97 boards. I don't know if Z97 measures voltage differently or what....pretty strange.

 

[ehume]

Some of these motherboards have Voltage measurement points. A DMM can read the Voltage. Does the measured Vcore = the indicated Vcore?

 

[bononos]

The review article mentions many enthusiasts testing dozens/100s of cpus to find that golden sample. Do those enthusiasts actually buy that many and sell off the parts that didn't make the grade in 2nd hand ebay?

....with many enthusiasts testing several dozen (or even 100s) in order to find a good CPU for benchmarking competitions....

 

[Homeles]

The review article mentions many enthusiasts testing dozens/100s of cpus to find that golden sample. Do those enthusiasts actually buy that many and sell off the parts that didn't make the grade in 2nd hand ebay?

The guys that are sponsored and such, sure.

 

[DrMrLordX]

I think some of them RMA the bad overclockers, which is often what you get in open box sales from NewEgg and others. They used to do that, anyway.

 

[tw33k]

Some of these motherboards have Voltage measurement points. A DMM can read the Voltage. Does the measured Vcore = the indicated Vcore?

I've tried connecting my DMM to the back of my ASUS z97 Deluxe but all it gives me is the input voltage (1.8v). Every point gives me the same reading.

 

[ehume]

I've tried connecting my DMM to the back of my ASUS z97 Deluxe but all it gives me is the input voltage (1.8v). Every point gives me the same reading.

Bummer. I have a bunch of measuring points on my Gigabyte board. But my multimeter is something like 35-40 years old, and analog.

 

[CrystalBay]

Well , I have my DC 4790k batch# L418C169 running 4.8 and stress testing now, I am using a Noctua U14S at about 85F ambient 110F outside. I used this lil guide fromm JJ at Asus . http://pcdiy.asus.com/2014/06/6-4970k-cpus-overclocked-oc-impressions-of-devils-canyon-on-asus-z97-motherboards/ .

 

[WhoBeDaPlaya]

Well , I have my DC 4790k batch# L418C169 running 4.8 and stress testing now, I am using a Noctua U14S at about 85F ambient 110F outside. I used this lil guide fromm JJ at Asus . http://pcdiy.asus.com/2014/06/6-4970k-cpus-overclocked-oc-impressions-of-devils-canyon-on-asus-z97-motherboards/ .

Hope you used more than AIDA. Prime95 28.5 is just brutal on Haswells.

 

[CrystalBay]

Hope you used more than AIDA. Prime95 28.5 is just brutal on Haswells.

Aida Sure , OCCT works fine , Prime 28.5 fails even bone stock Haswells are going to, I understand ! For my daily needs so far so good ...I'm not new to PC's ,however my stability needs are not the same as others. Part of my success is my motherboard I feel.

Going for 4.9

 

[Galatian]

Aida Sure , OCCT works fine , Prime 28.5 fails even bone stock Haswells are going to, I understand ! For my daily needs so far so good ...I'm not new to PC's ,however my stability needs are not the same as others. Part of my success is my motherboard I feel.

Going for 4.9

Well to each their own...I feel stable when I get pass 24h Prime95. Sticking to 4,6 GHz now as it's the sweet spot I feel.