Hot things get hot! cont. 7980XE, 4.8GHz, and hell's fire.

[Sauron]

14 minutes ago, done12many2 said:

 

To be honest.  Temps from either Intel or AMD on the big core parts do not look great.  1950x CPUs are getting very hot on AIOs at stock speeds. 7980XE CPUs are going to get very hot on AIOs at stock speeds.

 

The big point is that enthusiast platforms usually call for enthusiast level supporting parts.  

 

Try to run ANY 16c or higher part made by ANY CPU manufacturer on an AIO at much higher than 4 GHz and it's going to get hot.  

The last thing I heard on the threadripper front was that current AIOs were struggling because they're far smaller tha the heatspreader, but new models should solve that - or do you have info on the temperatures with properly adapted aios?

 

Anyway obviously it will get hot, that's what we expect - the article was suggesting that it may be much cooler than we'd expect though.

[MageTank]

1 hour ago, Sauron said:

If this is done on an AiO cooler (which I STRONGLY doubt) then it's truly impressive and would show some massive improvement over everything we've seen from intel thus far.

 

However, if it's a custom loop there's no telling how many radiators the guy is using... and at that point a 4.8 GHz overclock may not sound as good when put into perspective with the price of the cooling solution. Still, it's good to know it's possible at least.

Well, if to criticize a product we had to buy it every time it would get a little complicated  if I can criticize, say, a pentium without having any intention of buying it (despite having more than enough money to do so), why shouldn't I criticize a part simply because I can't afford it? Besides, if we're talking about x299 in general, I might have been interested in the lower end options (which I could afford) - if they weren't terrible.

I said if you have valid complaints, that it's okay. My problem is with people that complain for the sake of complaining, without it making sense, or only to use it as a means to bolster another brand. An example of an invalid complaint: "These 10c X299 CPU's are too hot to use an AIO on, Intel sucks!". For stock clocks, an AIO works fine, but these people expected to use a cheap AIO on a $1000+ 10 core CPU and OC it at the same time? Yeah, it's not gonna work out that well. Even on one of the best 280mm AIO's, I can still hit 90C on a delidded 5ghz 7700k with Linpack MKL. To expect a 10 core to holdup at high clocks and much stronger AVX, that would be silly.

 

An example of a valid complaint: "It is a bad move by Intel to forego soldering these chips, especially when previous data shows how well solder helps." I prefer criticism to be constructive, and have no problem when it is. I just see a lot of people complaining about X299 when they have no real intent of using it, with or without custom loops. 

40 minutes ago, Mr.Meerkat said:

Does this include the large number of people crying about the VRMs on B350 motherboards?  

It includes anyone that refuses to do their own research, and just regurgitates anything they hear online from one source that they trust. The X299 situation is tricky. I like Der8auer, and there is no denying the man is extremely smart in his field. I also appreciate knowing the limitations of these VRM's. However, as I said when this situation first started, he did blow a large portion of it out of proportion. When you intentionally remove current limitations and run loads designed to create a ton of heat, it's going to create a ton of heat, lol. For people like myself that run Small FFT Prime95, it's a great thing to know. For the other 99% of the world, it makes no difference to them, even on this enthusiast platform. I think the bigger picture during that debacle was the fact that Intel surprised vendors with SKU's that were not previously mentioned, and that the board manufacturers chose to waste money on aesthetics, painting a bad board to look good, instead of spending that money on better thermal solutions. That being said, I know plenty of people running the 10 core at 4.6/4.7ghz 24/7 on those boards without any fires breaking out, so (for now) everything seems okay.

 

Since I have an extremely bad way with words, allow me to clarify my points in my own words once again. My point is not to stifle constructive criticism about products that may be bad. If you have valid complaints, voice them. My biggest problem with most people that frequent this (and the CPU/GPU subsection) part of the forum, is the mud slinging that goes on when the side they dislike, has a problem. Blatant statements like "X299 runs too hot, Threadripper is better because it's easier to cool" make zero sense. The heat scales relative to your workflow and your cooling solution. As I have said before (and have been scolded on this very forum for making this exact point), heat is not just a consequence of poor cooling. It's also a direct consequence of higher workflow efficiency. If you are using more of your processor, you are consuming more power as a result, creating more heat. Less of your CPU is being wasted, so it's more efficient workflow. X299's is going to be worlds hotter than X399 when it comes to AVX performance, but it's AVX performance is also going to be worlds faster as well. The people that need that performance, understand that they also need a cooling solution that will facilitate that demand, as do the people that choose TR over standard Ryzen.

 

Any outsider can look at this forum, and see it's full of gamers looking at X299 and X399 as if they are gaming CPU's. This was blatantly obvious when people were arguing over stock clocks, boost clocks and overclocked values for these CPU's. News flash! These are not gaming CPU's. Anyone that buys TR or X299's high core count CPU's with the sole intention of gaming on them, are wasting a serious amount of cash and potential horsepower for no additional benefits in that department. In fact, it will probably negatively impact framerates to choose X299 over a 5ghz+ 7700k (or upcoming 8700k). Before anyone bothers to mention "extreme multi-tasking", there is only so much one person can do at one time before you have to seriously question the practicality of it. "What if I am gaming while streaming to twitch while processing the raw footage to Premier while editing it to upload to youtube while streaming my local DVD collection to my family". To people like that, I'll say: The swiss army knife is great, but nothing beats having the right dedicated tool for the right job. 

 

Man, I need a break. These rants are getting longer and longer, and going absolutely nowhere, lol. 

[done12many2]

51 minutes ago, Sauron said:

The last thing I heard on the threadripper front was that current AIOs were struggling because they're far smaller tha the heatspreader, but new models should solve that - or do you have info on the temperatures with properly adapted aios?

 

Anyway obviously it will get hot, that's what we expect - the article was suggesting that it may be much cooler than we'd expect though.

 

Here's a 1920x running on a custom loop with a full sized TR water block from EK.  It's running stock clocks in the area of 3.5 GHz under load with a simple AIDA64 (we can all admit it's a weak load).  Average die temps (Tdie) are 68c with maxes near 80c.  Average Tctl temps were 95c with a max of 103.3.  I'm assuming that the throttling point for the 1920x is somewhat higher than the 68c (Tdie) throttling point for the 1950x.

 

While this isn't the information that you requested on AIO with new base plates, we could only assume that they would be worse than this.  We could definitely assume that a higher core count 1950x combined with the newer AIOs would also be worse than this.

 

I'd like to point out that my point is not to make AMD TR look bad in any way as it's a great CPU.  My point is the highlighting of Intel chips in this area is going beyond silly.  We're doing it now just for the sake of doing it, whether it's actually worse or better than AMD's situation. 

 

Both the 1950x and the 7980XE are going to take big boy solutions to cool if you intend to run something other than Google Chrome or Microsoft Word.  

 

 

 

[Taf the Ghost]

@done12many2 That was a great video from Paul. He was so close to losing it at the end.

 

As to cooling TR parts, the "standard" AIO setups aren't great for OC'ing because of noted contact issues. 

 

On the 7980XE, in the mid-4 Ghz range, I'd expect we're looking at power draw in the 500w range and related heat to go along with it. Since it seems the second run of boards are coming out with a lot better VRM design, it would seem the Motherboard makers finally got the CPUs in house and rushed to get some better power out to the boards. Though I still expect a few cases of melted boards once these finally launch.

[done12many2]

1 minute ago, Taf the Ghost said:

@done12many2 That was a great video from Paul. He was so close to losing it at the end.

 

Yeah, that AIO pump shitting the bed on him followed by that FML look was pretty funny.  

 

1 minute ago, Taf the Ghost said:

As to cooling TR parts, the "standard" AIO setups aren't great for OC'ing because of noted contact issues. 

 

On the 7980XE, in the mid-4 Ghz range, I'd expect we're looking at power draw in the 500w range and related heat to go along with it. Since it seems the second run of boards are coming out with a lot better VRM design, it would seem the Motherboard makers finally got the CPUs in house and rushed to get some better power out to the boards. Though I still expect a few cases of melted boards once these finally launch.

 

I for one am looking forward to taming the beast man.  I already have dedicated mining plans (haters welcomed) for current stuff, so the 7980XE can't get here fast enough.  

 

[Drak3]

5 minutes ago, done12many2 said:

 

Yeah, that AIO pump shitting the bed on him followed by that FML look was pretty funny.  

 

 

I for one am looking forward to taming the beast man.  I already have dedicated mining plans (haters welcomed) for current stuff, so the 7980XE can't get here fast enough.  

 

I'd imagine Epyc would be better for mining, in the long run.

 

But you can ship me to the 7980XE for some testing while you set that up.

[done12many2]

2 minutes ago, Drak3 said:

I'd imagine Epyc would be better for mining, in the long run.

 

But you can ship me to the 7980XE for some testing while you set that up.

 

Oh, I'm not mining with the x299 rig.  If I do, it's only going to be part-time when I'm not using it.

 

I meant that the 7980XE can't get here fast enough so that my current parts can be switched over to full-time work.

[Mr.Meerkat]

1 hour ago, MageTank said:

An example of a valid complaint: "It is a bad move by Intel to forego soldering these chips, especially when previous data shows how well solder helps." I prefer criticism to be constructive, and have no problem when it is

(And then someone rebuts it with the valid argument of how solder over time will crack from the constant thermal cycling which can cause failure eventually, or at the very least make it run at lot hotter. ) 

 

[MageTank]

Just now, Mr.Meerkat said:

(And then someone rebuts it with the valid argument of how solder over time will crack from the constant thermal cycling which can cause failure eventually, or at the very least make it run at lot hotter. ) 

 

On dies that large, it would take several years or more for it to become a real issue. No paste is 100% immune to thermal cycling, not even the Dow Corning blend used by Intel, specifically designed to handle heavy thermal cycling. 

 

In my opinion, the heat imposed by people pushing constant hot loads far outweighs the potential longevity issues of solder down the road. In contrast, Haswell-E's dies are much smaller compared to Skylake-E, and we've not seen widespread reports of solder failing to that extent. Then again,

 

[done12many2]

5 minutes ago, MageTank said:

On dies that large, it would take several years or more for it to become a real issue. No paste is 100% immune to thermal cycling, not even the Dow Corning blend used by Intel, specifically designed to handle heavy thermal cycling. 

 

In my opinion, the heat imposed by people pushing constant hot loads far outweighs the potential longevity issues of solder down the road. In contrast, Haswell-E's dies are much smaller compared to Skylake-E, and we've not seen widespread reports of solder failing to that extent. Then again,

 

 

Nothing like some properly delivered Samuel L. Jackson!

[Drak3]

1 hour ago, done12many2 said:

 

Nothing like some properly delivered Samuel L. Jackson!

Nothing better than Samuel L. Jackson playing a white guy, emulating a stereotypical black guy.

[Coaxialgamer]

I'm actually morbidly curious how much power that 18 core will pull once overclocked.

The enthusiast in me loves that were getting products like this again , but the sane person in me , well ..

 

but it's pretty clear that if a decent AIO has trouble cooling an OC 7900x , then i don't think it'll fare well against an 18 core...

[ARikozuM]

1 hour ago, Drak3 said:

Nothing better than Samuel L. Jackson playing a white guy, emulating a stereotypical black guy.

Nothing like a forum poster using a comma to form a fragmented, sentence. 

[Drak3]

8 minutes ago, ARikozuM said:

Nothing like a forum poster using a comma to form a fragmented, sentence. 

Bite me.

[ARikozuM]

7 minutes ago, Drak3 said:

Bite me.

[Drak3]

21 minutes ago, ARikozuM said:

You kinky fucker.

[Cookybiscuit]

2 hours ago, Coaxialgamer said:

I'm actually morbidly curious how much power that 18 core will pull once overclocked.

The enthusiast in me loves that were getting products like this again , but the sane person in me , well ..

 

but it's pretty clear that if a decent AIO has trouble cooling an OC 7900x , then i don't think it'll fare well against an 18 core...

Surely it has to be over 1000W total system power consumption.

[Taf the Ghost]

7 hours ago, Coaxialgamer said:

I'm actually morbidly curious how much power that 18 core will pull once overclocked.

The enthusiast in me loves that were getting products like this again , but the sane person in me , well ..

 

but it's pretty clear that if a decent AIO has trouble cooling an OC 7900x , then i don't think it'll fare well against an 18 core...

OC'd 7900X is in the 300w range, depending on the workload and measurement type. So somewhere in the 500w to 600w range. Power scaling has been extremely linear with Skylake-X, which isn't surprising.  Though, in this case, the OC actually could bring a fairly significant performance boost, which is why I'm really curious about the testing when this drops. The stock vs OC results should be hilarious.

[ravenshrike]

On 9/2/2017 at 1:04 PM, done12many2 said:

 

Here's a 1920x running on a custom loop with a full sized TR water block from EK.  It's running stock clocks in the area of 3.5 GHz under load with a simple AIDA64 (we can all admit it's a weak load).  Average die temps (Tdie) are 68c with maxes near 80c.  Average Tctl temps were 95c with a max of 103.3.  I'm assuming that the throttling point for the 1920x is somewhat higher than the 68c (Tdie) throttling point for the 1950x.

 

 

He also uses the single large paste blob, which is decidedly not the optimal solution. Instead the optimal solution, at least for asetek coolers and I can't see why it wouldn't apply to EK ones, seems to be smearing paste across the baseplate like its going out of style.