At first, I was amused by how often and gladly the AMD representatives made the possible air cooling of both core monsters a topic and distributed thermal side blows against Intel like other people useless pizza flyers. But, lords, basically that would be something, wouldn't it? Experiment makes smart and so there is no stopping for such experiments.
And while Intel's large Intel Core i9-7890XE and i9-7960X (for now) still have to sweat under the greasy layer of thermal paste, the Threadripper family is all in the balance and far from over. By the way, I have beheaded the two Intel CPUs according to the application benchmarks, because the pamphlet is really not suitable for measuring the maximum power consumption and cooling with such bolts.
Air cooling and its limits
Yes, it really works, you can cool both new Ryzen Threadripper with air. I'm going to take that forward as a realization. If you also do without manually activated Precision Boost Overdrive (PBO), then both CPUs still work quite smart and at least with the Ryzen Threadripper 2950X even without much fuss. With this CPU even PBO works with air, only at there is implemented approx. 250 watts then also stop with funny (and quiet).
But: even then it goes quite well and even tactically somewhat lossless, if you do not see the noise level of the fan rotating at 2500 rpm. Because you can hear this clearly in a closed housing. With 45.6 dB(A) at a distance of 50 cm one is already very high, especially since the lower frequency bearing noise of the fan due to the fixed screwing with the motherboard as body sound should also be transferred to the board and later even to the housing.
As good as the series cooler contributed by Cooler Master (which should come from AVC as usual) may be, who really relies on air AND quiet, will land on the Ryzen Threadripper 2950X and probably rely on a potent Noctua cooler. What bothers me personally about AMD's cooler, however, is the nonsensical execution of the cables on the side facing the PCI Express slots.
If you put a very potent graphics card in the first slot, the USB port on the cooler is completely hidden and no longer usable. In addition, a graphics card with a backplate already touches the cooler directly, which is thermally complete nonsense. That's why I sorted out the air cooler (together with the rather weak all-in-one compact water cooling from Enermax) for a part of the next measurements, because I am now about exploring the real limits of these two CPUs. But that is definitely not possible with these two cooling solutions.
What do XFR2 and Precision Boost Overdrive (PBO) really do?
For all the theoretical backgrounds, I refer explicitly to my launch review of the Ryzen 2700X and the very detailed techniques. Despite all the doubts of some media about my measurements at that time, which AMD certified an exact point landing of the power consumption (Package Power) to the specified limits, I have remained to this day with the statements made. For good reason, by the way.
With today's tests, I can even support this, because as long as the motherboard manufacturers do not cheat and adhere exactly to AMD's specifications (e.g. the PPT), as long as the measured values remain exactly within the defined frame! That is what I want to do here, so that the limits of the following measurement results can be better classified. From now on, I measure everything with water and from reaching the thermal limits I also rely on the chiller in the extended circuit of our new water cooling measuring station.
With XFR2 and PBO, of course, the right cooling is important as always. I first pushed the Ryzen Threadripper 2990WX to its limits, but without using the last force. Nevertheless, more than 42 amperes are already a house number when using the chiller on the EPS connection of the motherboard, even if there has nailed thermally nothing yet. With LN2 you can get to over 50 amperes and you need a good power supply to not trigger the OCP. Especially since you also have to manually remove the 500-watt limit on the motherboard.
AMD has come up with a new (old) temperature determination and evaluation for both CPUs this time! All measurements and data are based on Tdie, as the temperature of the package! What is output as a 27°C higher Tctl value for compatibility reasons is only used for fan control and other functions to remain backwards compatible! AMD sets the upper limit for Tdie to 68°C for both new thread ripper CPUs (which in turn results in 95°C Tctl as the control value). Tctl is therefore no longer a measured value!
The curves for the Ryzen Threadripper 2990WX will look like the graphic below. In addition, I have emphasized the range of application of the different cooling solutions in color and would like to show within which performance ranges which solution makes sense or. necessary. With the air cooler and the all-in-one compact water cooling, this CPU monster can no longer get very far under full load of all cores, even if in both cases I had to operate the fans at maximum speeds. After all, the overclocking margin that AMD actually allows ex works with the PBO can no longer be used sensibly.
If you are wondering about the jumps of the individual temperature curves: they are based on the temperature values of the cooling used in each case, which I had to significantly intensify upwards (and thus change several times). Thus, there are then four curves, because LN2 runs out of competition and also well below zero. However, since the sensors only 20°C upwards to show any meaningful values, so I'll leave that better.
Different with the Ryzen Threadripper 2950X, which I also like much more in terms of coolability! Because one thing is also a fact: the effort you have to put into cooling is significantly lower! A normal Custom Loop water cooling system or an all-in-one compact water cooler with 360 radiator and good fans are enough. Even with air cooler there is something else, hat off! However, the blue curve for Tctl is now based only on the values of the Custom Loop water cooling:
Intermediate conclusion
While the Ryzen Threadripper 2950X is also well cooled up to just over 250 watts with air and also boosts properly, you should put on a solution with water again for the last megahertz. Without PBO, however, the air cooler is also comfortable up to exactly 180 watts and remains relatively quiet. The Ryzen Threadripper 2990WX is also available without PBO up to approx. 250 watts still well to cool with air. This is the good news of the day and record-breaking.
After all, it also creates a tick more than the nominally attached 3 GHz, but beyond that it gets loud. The large CPU can be mastered without PBO without enough, but if you have the 4 GHz mark in mind, you have to drive up a larger calibre. Because then the big one sucks like a hole and any efficiency goes into the orcus. Of course, this has to be made up with yourself, especially since PBO is not a guarantee component.
Kommentieren