How can this be? Air better than water?

[ShrimpBrime]

Ok back on topic, back when the FX9590 came out I got 1 and went through 3 coolers, 2 air and 1 water (V6GT, V8GTS and Nepton 280L) all 3 failed, I lost on the silicon lottery, the Nepton 280L lasting the longest under load.
I then replaced the FX9590 with FX8350, and the temps on the 8350 under max load maxed at 30c (30c as reported by HWM), then the pump on the AIO failed after 3 months, then swapped with V6GT with the same temps as the Nepton 280L.
So if the cooler (air or water) is not rated at or above the power the CPU is pulling (OC'd or not) then there will be high temps.

Oh man you are comparing AMD's 2nd hottest chip to a mid range considerably lower wattage cpu. Both with very different leakage characteristics.... The 1st hottest chip would be the Agena 9950. I don't think even the PHII 980BE put out nearly as much heat as the 9590 and 9950 processors. AMD even out right said to use water cooling on the FX-9590. They didn't ship air coolers with them chips for exactly that reason. The second you turn on that PC, the heat sink was heat soaked.

A super great FX chip for temps would be like the 8370E or perhaps lower ended like 8300 and 8310 chips. You could almost passively cool them ones in comparison.

79c to 81c is within margin of error. There. Video debunked. Moving on then lol.

- - - Auto-Merged Double Post - - -

I should just delid it

I support this statement.

 

[WhitehawkEQ]

Oh man you are comparing AMD's 2nd hottest chip to a mid range considerably lower wattage cpu. Both with very different leakage characteristics.... The 1st hottest chip would be the Agena 9950. I don't think even the PHII 980BE put out nearly as much heat as the 9590 and 9950 processors. AMD even out right said to use water cooling on the FX-9590. They didn't ship air coolers with them chips for exactly that reason. The second you turn on that PC, the heat sink was heat soaked.

A super great FX chip for temps would be like the 8370E or perhaps lower ended like 8300 and 8310 chips. You could almost passively cool them ones in comparison.

79c to 81c is within margin of error. There. Video debunked. Moving on then lol.

- - - Auto-Merged Double Post - - -



I support this statement.

When I had the FX9590 on the MB, everything was fine at idle, it's when I ramped it up is when the MB would lockup or reboot. What I was comparing was the AIO to the V6GT on the FX8350 in terms of cooling after I switched CPU's. (p.s. there was no FX9950, just 9370 & 9590)

 

[ShrimpBrime]

Ah I see the comparison you where making now. Gotcha.

I wrote Agena 9950 which was a 2.6ghz phenom 1. Never said it was FX. No biggie.

 

[Alaric]

Agena 9950 which was a 2.6ghz phenom 1

Loved my 9850. First CPUI OC'd and first rig I built. I think I still have it somewhere (a pin may be missing). Hit 3.038 GHz and never had a temp problem.

 

[ShrimpBrime]

Loved my 9850. First CPUI OC'd and first rig I built. I think I still have it somewhere (a pin may be missing). Hit 3.038 GHz and never had a temp problem.

The 9850 was the first soldered de-lid I had ever done.

The low TDP threshold made those processors somewhat difficult. If I remember it was 61c or 64c.... something of that nature.

Todays AMD chips at 61c would be considered decent temps while the Ryzen chips seem perfectly fine well into the 70c ranges. Good stuff.

 

[EarthDog]

Ryzen chips seem perfectly fine well into the 70c ranges

They are fine up until 90C. IIRC, their throttling point is 95 or 100C.

70C would be considered "great" temps.

 

[ShrimpBrime]

They are fine up until 90C. IIRC, their throttling point is 95 or 100C.

70C would be considered "great" temps.

For sure, try to keep mine sub 80c when possible.

 

[freeagent]

70 is great! I used to be ok with 80 for years of course less is better but.. now I'm ok with 90.. 95..ish. That is also with a 1.2ghz overclock.. seems like overclocks are fading these days. Intels fast cpus boost to 4.7, and guys are happy with 5.0-5.1ish .. call it a 400-500mhz overclock of course ambient cooled. Seems to be the average oc for years was at least 900 to a ghz. Usually more like 1.3 or so it seemed most folks were getting. Now it seems like we are getting back to the 939 era where you were lucky to get 300-500.. Anyways I could probably drone on and on but its dinner time and I'm hungry.. staring at my bag of tacos but don't want to spoil my dinner. Its a tough decision. I'm going to leave this room now.

 

[ehume]

70 is great! I used to be ok with 80 for years of course less is better but.. now I'm ok with 90.. 95..ish. . . .

It is true the CPUs will happily run that hot, but I found in testing that I got the most GFlops when I held the temps at 80c or below. The faster clocks did not result in better produtivity.

 

[Alaric]

70 is great! I used to be ok with 80 for years of course less is better but.. now I'm ok with 90.. 95..ish. That is also with a 1.2ghz overclock.. seems like overclocks are fading these days. Intels fast cpus boost to 4.7, and guys are happy with 5.0-5.1ish .. call it a 400-500mhz overclock of course ambient cooled. Seems to be the average oc for years was at least 900 to a ghz. Usually more like 1.3 or so it seemed most folks were getting. Now it seems like we are getting back to the 939 era where you were lucky to get 300-500.. Anyways I could probably drone on and on but its dinner time and I'm hungry.. staring at my bag of tacos but don't want to spoil my dinner. Its a tough decision. I'm going to leave this room now.

Yields are hugely improved now. For every 1000 MHz OC there were scads of chips that got 500 MHz or less, mostly buried in the ranks of non OCers. We are a dying breed, but sales keep the factories churning, not geeks. What is good for the industry is, by definition, our death knell.

 

[ShrimpBrime]

Yields are hugely improved now. For every 1000 MHz OC there were scads of chips that got 500 MHz or less, mostly buried in the ranks of non OCers. We are a dying breed, but sales keep the factories churning, not geeks. What is good for the industry is, by definition, our death knell.

There is a lot of truth here. Dying breed.... to an extent...

BUT

AMD has made the processors overclock via artificial intelligence. That means everyone with a Ryzen processor is already overclocking and they don't even know it.
For example, using my 2700X here, the Max P-state according to the Cpu-z txt file is 3.7Ghz, NOT the 4ghz all core boost or max single core XFR of 4350mhz.

The forums are dying though. Mobile communities and social software has taken a big chunk of people from these wonderful forums. Reddit would be an example or perhaps discord.

But yea, more people are overclocking now than ever before. They just don't know or realize it lol.

 

[Alaric]

I love this platform. You get to know people, understand nuances, share stories and know that they get it. It started as a place to learn how to OC the worst OC chip ever and became the corner computer bar where people know me and may actually give a s***.

 

[mackerel]

AMD has made the processors overclock via artificial intelligence. That means everyone with a Ryzen processor is already overclocking and they don't even know it.

I thought the "AI" was in the branch predictor, not related as such to the operating clock. Also, define overclock? The boost behaviour is an intended AMD designed in feature, so can you call it an overclock? The difference today is, they can use sensors and whatever to run CPUs much closer to their limit. What we used to have was much more headroom to explore manually.

 

[ShrimpBrime]

Yes SenseMi features branch prediction and overclocking beyond base clocks via temp, voltage and current.

Auto OC. 2700x max p-state is 3.7ghz.
Fx 9590 max p state is 5ghz.

Neither processor has OC head room.

 

[EarthDog]

Marketing. How is it AI when it doesn't adapt? It just fits within (3) parameters, right?

Overclocking, to me, is going past listed values - this includes listed boost clocks. For example, If the all core boost clock is listed at 4.2 GHz, manually going past that on all cores is overclocking.. same with single core.

 

[mackerel]

"Neural Net Prediction" is listed on AMD's SenseMI page, but it gives at best a marketing description of what it does and it is unclear to me if it is any more than the branch prediction part.

A CPU operating beyond base clock but within the manufacturer's intended boost strategy is not an overclock to me. Out of the box boost/turbo is not overclocking. I'm not sure exactly where PBO sits though, as I've not looked into it enough. On Intel side, MCE is an overclock as you are exceeding the clock table. Relaxing PL2 power limit/time is NOT an overclock.

 

[trents]

Since modern CPUs boost only one or two cores, and those being the strongest cores (or at least that's my understanding), it could be argued that a CPU configured to run all cores at the boost speed is an overclocked CPU since the weaker cores that don't boost are running faster than stock.

 

[Bustos]

Never realized we were so sensitive about what constitutes as ‘overclocking.’

 

[EarthDog]

Since modern CPUs boost only one or two cores, and those being the strongest cores (or at least that's my understanding), it could be argued that a CPU configured to run all cores at the boost speed is an overclocked CPU since the weaker cores that don't boost are running faster than stock.

Modern Intel CPUs have a base clock, an all core boost, and dual thread boost.

For example, If the all core boost clock is listed at 4.2 GHz, manually going past that on all cores is overclocking.. same with single/dual core. If single/dual core boost is listed at 4.4GHz, anything past that on single/dual core is past spec and overclocking.

Never realized we were so sensitive about what constitutes as ‘overclocking.’

 

[UltraTaco]

As per Intel Guy, no one is overclocking if turbo rooster is enabled. It is considered an "in spec" operation. Intel can just do it for certain times period, and amd can run in turbo if cooling and power limit is within limit.