We have been working with Der8auer to find out what was really going on with X299 heat issues
X299 VRM Issues - An Update
- Thread starter tinytomlogan
- Start date
Thelosouvlakia
Classy-Fried-Chicken
I watched Der8auer's video in the morning... 12K on prime95... Brutal...
TheF34RChannel
New member
Superb work and collaboration Tom, thanks!
AlienALX
Well-known member
Right I watched about half and then skimmed through. So basically in a nutshell what he said before was right. That basically certain loads and certain things will overheat the VRMs because they are not enough to keep a huge 10 core CPU under load like that. This is why when we discussed this before I was talking more about the big picture. IE - if these VRMs and boards and CPUs can't handle running at full pelt all of the time what will happen when you introduce an unlocked 18 core?
I have played around a lot with Prime, LinX and OCCT. You are right, the tests do stop and then a new one begins. This is mostly to test the CPU properly by turning all of the flags to 1 and then allowing them to go back to 0 (so think of it like cells, you need to test them in both directions to make sure they are working properly).
You are also right about what you were saying about it being unlikely to see those sorts of loads in "day to day" use. However, from my time with the FX 8320 there are quite a few. For example, certain levels in Crysis 3 (not all of them strangely) would use all 8 cores, and actually cause the same overheat/lock up issues I had with P95. As I said though, not all of the levels did that but the worst of them was the swamp level. That used to make my rig crash more than any benchmark. I guess because not only was the CPU drawing power through the board but the GPU also.
This won't be much of an issue any time soon. However, when everything wants all of your CPU it will be a much bigger problem.
I have played around a lot with Prime, LinX and OCCT. You are right, the tests do stop and then a new one begins. This is mostly to test the CPU properly by turning all of the flags to 1 and then allowing them to go back to 0 (so think of it like cells, you need to test them in both directions to make sure they are working properly).
You are also right about what you were saying about it being unlikely to see those sorts of loads in "day to day" use. However, from my time with the FX 8320 there are quite a few. For example, certain levels in Crysis 3 (not all of them strangely) would use all 8 cores, and actually cause the same overheat/lock up issues I had with P95. As I said though, not all of the levels did that but the worst of them was the swamp level. That used to make my rig crash more than any benchmark. I guess because not only was the CPU drawing power through the board but the GPU also.
This won't be much of an issue any time soon. However, when everything wants all of your CPU it will be a much bigger problem.
NeverBackDown
AMD Enthusiast
You should do more stuff with him. I love all the technical nerdy stuff
agent_x007
New member
I would like to see a Intel Burning Test (Maximum/Very High) comparison to OCCT.
Rampage Extreme (X48), which Tom mentions at one point, has 16 phases (doubled 8 phase) with heatpiped cooling and with optional additional cooling that can be mounted on top of that
I bet it could take that 400W like nothing was happening
Rampage II Extreme has similar VRM setup (16 phase), with added 3 phase for IMC/QPI and another 3 phases for X58. No additional cooling over stock heatpipes tho :/
Rampage Extreme (X48), which Tom mentions at one point, has 16 phases (doubled 8 phase) with heatpiped cooling and with optional additional cooling that can be mounted on top of that
I bet it could take that 400W like nothing was happening
Rampage II Extreme has similar VRM setup (16 phase), with added 3 phase for IMC/QPI and another 3 phases for X58. No additional cooling over stock heatpipes tho :/
Last edited:
Thelosouvlakia
Classy-Fried-Chicken
By the way, I'm not using OCCT for testing. Prime95 it is, because I might pass occt running for an hour or so, think that my cpu is stable at 4.8Ghz and when I spend two hours in a cpu punishing game (ie arma3) I get thrown out of the game due to an instability. While prime95 is pretty much the tool to go for 100% stability. (and yes even on the old 8350 I had to raise the thermal limits and current limits in order to stabilize prime95)
Great to see Der8auer and TTL go deep on this one.
Far two many reviewers don't go deep enough in my opinion , maybe theres not an audience.
Once of the main reasons I joined this forum many moons ago was for content like this.
I think this week we have seen a new dog teach an old dog some new tricks
Far two many reviewers don't go deep enough in my opinion , maybe theres not an audience.
Once of the main reasons I joined this forum many moons ago was for content like this.
I think this week we have seen a new dog teach an old dog some new tricks
Boards and VRM's are fine - its disabling overcurrent protection etc AND running an overclock where the issue is. The X299-A is capable of pushing out about 650w if you cool the VRM's ok.
A4.6GHz 10Core with OCCT is only hitting 80 on the VRM's and normal use is way less than even that.
Its an issue to a very small amount of people, for most users its just an excuse to test, learn and find a way to make things better....
Prime is punishing
OCCT will normally find an instability
Last edited:
AlienALX
Well-known member
Yeah if you disable things and start thrashing you can expect a fire I did that with my Crosshair.. Thankfully it was stupid solid and never failed.
Most of the problem with VRMs is that they are tiny. Like a couple of mm across. No matter how big your block of copper/alu there's only so much they can take.
We're going to need better boards for the big bugger 299 chips. I bet once you get to 12 cores and beyond most boards will struggle. I just hope they make it clear on Intel (that's my only gripe tbh) and tell people.
Otherwise we are going to end up in Apple land. Where a 2ghz I5 actually runs at about 800mhz all of the time because it throttles to s**t ! IE - I hope Intel don't start writing cheques that boards can't cash.
We'll see I suppose. And BTW Tom great video and a fantastic collaboration. I love Der8auer 'cause he's a total nerd. So it's fantastic to see you working with him
I did that with my Crosshair.. Thankfully it was stupid solid and never failed.Most of the problem with VRMs is that they are tiny. Like a couple of mm across. No matter how big your block of copper/alu there's only so much they can take.
We're going to need better boards for the big bugger 299 chips. I bet once you get to 12 cores and beyond most boards will struggle. I just hope they make it clear on Intel (that's my only gripe tbh) and tell people.
Otherwise we are going to end up in Apple land. Where a 2ghz I5 actually runs at about 800mhz all of the time because it throttles to s**t ! IE - I hope Intel don't start writing cheques that boards can't cash.
We'll see I suppose. And BTW Tom great video and a fantastic collaboration. I love Der8auer 'cause he's a total nerd. So it's fantastic to see you working with him
Dark NighT
New member
That was really informative and it is great to see a collaboration happening with this and working it out together, more please!
Overclocking the 16 and 18 core cpu parts might be the biggest minefield, so much more power consumption.
Overclocking the 16 and 18 core cpu parts might be the biggest minefield, so much more power consumption.
Unless I misunderstood, Der8auer mentioned that allowing the digi+ to run above auto can also see higher scores in other benchmarking / torture tests , is this something that you have seen since retesting ?
Excalabur50
Well-known member
It's really good that yourself and Der8auer were able to amicably sort everything out and it was really interesting to hear what it took to sort it and just goes to prove you never stop learning, great video and really well worth watching
Last edited:
Eddie long
Member
very interesting , thanks too both of you for clarifying the issue
AlienALX
Well-known member
Does anybody remeber these, maybe its time that they made a comeback :
God no. They sounded like you had a case full of bumblebees !
They were only for water cooling too, so when you had poor airflow. Cases have changed quite radically since then.
