Huge temp spike with 5930K OC and H110

[taserbro]

Hello PromitR
I'm running the 5960x with a swiftech h220-x on clu and I'm also stuck behind a thermal wall at 4.6ghz 1.33V where I'm also seeing a massive heat runaway a couple minutes into any stress test that I don't feel like playing chicken with.
There's a pretty huge difference in overclocking behavior difference between the different chip batches; I had the chance to play with a 5690x chip that seemed to output less heat at 1.37V than mine at 1.3V but could not get past 4.4ghz at all no matter what I threw at it.

That said, from wasting way too much time reading through pretty much everything on ocn, I find that most people who do achieve frequencies higher than 4.5/4.6 do so with manually binned chips out of half a dozen samples by abusing amazon returns and often using exotic phase change cooling or custom water loops with 3-4 radiators. For the amount of efforts using off the shelf components, I'm more than happy with 4.5ghz.

Have you tried increasing your cache frequency? In my experience, it will give you a measurable performance increase for a very small thermal cost.

 

[BonzaiDuck]

Hello PromitR
. . . . .
That said, from wasting way too much time reading through pretty much everything on ocn, I find that most people who do achieve frequencies higher than 4.5/4.6 do so with manually binned chips out of half a dozen samples by abusing amazon returns and often using exotic phase change cooling or custom water loops with 3-4 radiators. For the amount of efforts using off the shelf components, I'm more than happy with 4.5ghz.

. . .

I was planning to use two radiators. If you have the links handy for those "exotic . . . custom" projects, I'd like to take a look.

4.5Ghz for the X chip is no mean accomplishment. The Maximum PC review had only printed murmurs and hints from mobo or system vendors that "4.5 was possible" or "in the cards."

There was mention with expectations of newer steppings or improved micro-code. Said it before -- we'll just have to wait and see . . .

The magazine mildly hypes the E processors as an "enthusiast's dream." I suppose it all depends on your cooling budget and your OC expectations, but there's no denying that these are powerful chips . . .

 

[taserbro]

I was planning to use two radiators. If you have the links handy for those "exotic . . . custom" projects, I'd like to take a look.

4.5Ghz for the X chip is no mean accomplishment. The Maximum PC review had only printed murmurs and hints from mobo or system vendors that "4.5 was possible" or "in the cards."

There was mention with expectations of newer steppings or improved micro-code. Said it before -- we'll just have to wait and see . . .

The magazine mildly hypes the E processors as an "enthusiast's dream." I suppose it all depends on your cooling budget and your OC expectations, but there's no denying that these are powerful chips . . .

Here is ocn's haswell-e thread. I'm afraid there's no way I can remember which page exactly I saw the rigs being displayed but right off the bat you can check the updated configs like asg's custom loop using 3 480/360 radiators and munney's phase change build. Also, I found it worth it to read through most of it since there's a lot of useful information that if anything has the calming effect of making you realize your chip isn't defective and that there's a ton of people running into the exact same issues, sometimes worse...

That said, these are really just to show what it takes to break the wall of diminishing returns and get those last 300-400mhz provided you even have the golden chip that can go that high. For the price of some of those setups, I could see myself forgoing the 5690x and instead getting multiple 5820k rigs and starting a render farm, which would make much more sense in a utilitarian sense for my use. But that's the cost of having a single rig that can both destroy more expensive xeon workstations on rendering benchmarks and still game as well as 4790k's without sounding like a tugboat engine.

 

[BonzaiDuck]

I've stumbled across some web-pages at ASUS ROG and Tech-Powerup, which mention something ASUS did with the 2011-v3 socket.

The Haswell E processors have some additional pads in the LGA. These were supposedly added for test purposes, likely related to the FIVR integrated voltage regulation of the Haswell cores.

ASUS apparently "innovated" their socket to "2011 OC" and made use of the extra pads. They had already managed to override the FIVR limitations on the earlier Haswell chips, noted in another ROG page:

http://forums.anandtech.com/newreply.php?do=newreply&noquote=1&p=36811716

Since the OP as an ASUS X99 board (whichever one -- maybe the "Deluxe"?) -- I am . . . just wondering as to how this might all be related.

Shot in the dark.

Even so, without reference to their motherboard choices, some folks reported getting a 5930K stable in the 4.6Ghz range with load temperatures in the low 60's C (with water-cooling.)

Other rumors suggested that Intel was "investigating" the ASUS socket.

What else is known about this? How would it impact the various experiences people are having with the E processors?

 

[taserbro]

I've stumbled across some web-pages at ASUS ROG and Tech-Powerup, which mention something ASUS did with the 2011-v3 socket.

The Haswell E processors have some additional pads in the LGA. These were supposedly added for test purposes, likely related to the FIVR integrated voltage regulation of the Haswell cores.

ASUS apparently "innovated" their socket to "2011 OC" and made use of the extra pads. They had already managed to override the FIVR limitations on the earlier Haswell chips, noted in another ROG page:

http://forums.anandtech.com/newreply.php?do=newreply&noquote=1&p=36811716

Since the OP as an ASUS X99 board (whichever one -- maybe the "Deluxe"?) -- I am . . . just wondering as to how this might all be related.

Shot in the dark.

Even so, without reference to their motherboard choices, some folks reported getting a 5930K stable in the 4.6Ghz range with load temperatures in the low 60's C (with water-cooling.)

Other rumors suggested that Intel was "investigating" the ASUS socket.

What else is known about this? How would it impact the various experiences people are having with the E processors?

The extra pins that most x99 asus boards (at least the deluxe, ws and rve) have allow them to access some extra uncore/cache settings afaik. The pins were always there on the cpu, yet not in the final specifications of the socket but asus decided to tell their oem to leave them in and the asus rep on ocn mentioned that it helps in some overclocking scenarios.
It was mentioned in the ocn haswell-e and the asus motherboard support thread over there.

In my limited experience (I didn't get to use my gigabyte x99 board for long before it went belly up on me and I switched to the asus one), if there was a difference I really couldn't really see anything measurable whereas going from a bad chip to a decent one was like day and night. But currently asus makes the better x99 overclocking boards anyway so if that's what you were planning on doing, it should just make your choice a tad easier.

 

[BonzaiDuck]

The extra pins that most x99 asus boards (at least the deluxe, ws and rve) have allow them to access some extra uncore/cache settings afaik. The pins were always there on the cpu, yet not in the final specifications of the socket but asus decided to tell their oem to leave them in and the asus rep on ocn mentioned that it helps in some overclocking scenarios.
It was mentioned in the ocn haswell-e and the asus motherboard support thread over there.

In my limited experience (I didn't get to use my gigabyte x99 board for long before it went belly up on me and I switched to the asus one), if there was a difference I really couldn't really see anything measurable whereas going from a bad chip to a decent one was like day and night. But currently asus makes the better x99 overclocking boards anyway so if that's what you were planning on doing, it should just make your choice a tad easier.

I'm curious . . . How did the Gigabyte board go "belly up?" There had been rumors of "smoke and fire" with some ASUS X99 Deluxe boards . . .

I currently prefer ASUS because I like their UEFI BIOS, and my current rig's board makes the rig a winner.

If the troubles people report with their E chips are due entirely to the chip lottery, it just seems that there's a wider range of behaviors than I'd ever seen with the Sandy cores.

Especially -- two folks here (maybe on different threads and forums, too!) -- report temperature spikes. And I don't know if there was info about voltage as monitored -- or not.

If I build an X99/Haswell-E system in 2015, I want to make the right choices up-front. I'm going to take my time, but you can see -- I'm looking closely at the threads here and elsewhere "early in the game."

It may be there are so many of us still "warmed up" over our SB-K systems that we're all waiting for each other to build their Haswell-E's and post reassuring info about it.

 

[taserbro]

I'm curious . . . How did the Gigabyte board go "belly up?" There had been rumors of "smoke and fire" with some ASUS X99 Deluxe boards . . .

I currently prefer ASUS because I like their UEFI BIOS, and my current rig's board makes the rig a winner.

If the troubles people report with their E chips are due entirely to the chip lottery, it just seems that there's a wider range of behaviors than I'd ever seen with the Sandy cores.

Especially -- two folks here (maybe on different threads and forums, too!) -- report temperature spikes. And I don't know if there was info about voltage as monitored -- or not.

If I build an X99/Haswell-E system in 2015, I want to make the right choices up-front. I'm going to take my time, but you can see -- I'm looking closely at the threads here and elsewhere "early in the game."

It may be there are so many of us still "warmed up" over our SB-K systems that we're all waiting for each other to build their Haswell-E's and post reassuring info about it.

My gigabyte board went down to a faulty bios that bricked it; no burning, no fireworks, nothing spectacular happened. I flashed it to the newest version at the time because the stock one that it shipped with was entirely unusable and would freeze after a few seconds no matter what I did. It could have been any number of problems in the process, bad flash drive, microsurge while flashing, bad luck gamma ray flipping a bit, I don't know but the board itself wasn't bad at all. In fact, it was my first option over the asus models because I was used to and liked the gigabyte uefi from using their z97 boards and I loved the ability to apply changes in real time without reboot.

As for the temperature "spikes", I'm pretty sure I'm just running into the limit of how much heat a single 240mm low fpi radiator with stock fans can deal with. Besides the very different behavior from different chip batches, I think the thermal properties of haswell-e has been the least complicated issue to deal with so far and I consider myself a complete noob when it comes to liquid cooling.

Also, keep in mind that the folks on ocn were almost all day one early adopters and received costa rica made early batches of chips which explains the larger variance between samples. The later L4 batches are much more consistent and tend to overclock more gracefully so it's unlikely you'll run into something like the infamous 3422b90...

 

[BonzaiDuck]

. . .
As for the temperature "spikes", I'm pretty sure I'm just running into the limit of how much heat a single 240mm low fpi radiator with stock fans can deal with. Besides the very different behavior from different chip batches, I think the thermal properties of haswell-e has been the least complicated issue to deal with so far and I consider myself a complete noob when it comes to liquid cooling.

Also, keep in mind that the folks on ocn were almost all day one early adopters and received costa rica made early batches of chips which explains the larger variance between samples. The later L4 batches are much more consistent and tend to overclock more gracefully so it's unlikely you'll run into something like the infamous 3422b90...

The reviews show promise for "E". My own intention is to build one next year without worry of coming "late to the party" or missing out on Broadwell. My plan is a deliberate plan.

Cooling will get special attention in "the plan." If I make an outlay on any "E" processor and better-than-decent motherboard, I don't want to experience frustration over temperatures and clock-speeds. So I've been eavesdropping on the discussions about these chips.

If I commit to custom water-cooling, I won't need any other problems . . . heh-heh . . . .

 

[PhIlLy ChEeSe]

You spent all that money on a top of the line system then use some economy water cooler, sell what you have n buy the pump, hose, a block and a rad. Be done with it, other why's change the temps in the room. Turn off the heat open the window(I do), the change alone could be 20 or 30 degree's depending on outside temps.

 

[BonzaiDuck]

You spent all that money on a top of the line system then use some economy water cooler, sell what you have n buy the pump, hose, a block and a rad. Be done with it, other why's change the temps in the room. Turn off the heat open the window(I do), the change alone could be 20 or 30 degree's depending on outside temps.

I guess the only question which lingers for me is this. What will the AiO makers and even air-cooler designers like Noctua release before middle of next year?

Since I don't know the answer to that, my own plan calls for custom water parts with more radiator cooling capacity than the best AiO's can offer. Costs more, but so will the motherboard, processor and DDR4. So -- Yea-ah!! Absolutely!