General Discussions

[elstaci]

When I saw the term TGP on the NAVI21 Article that I posted I was wondering what that meant. This Tech site explains the difference between TDP, TBP, & TGP: Differences Between TDP, TBP and TGP on a Graphics Card | ITIGIC 

_______________________________________________________________________________________________

Each generation that passes the graphics cards, like any component, are more and more specific, and this leads to changing or introducing certain parameters to explain different aspects of them. Three of them are called TDP, TBP and TGP , something that will sound like a completely different language to many, but what are they and how do they differ?

There is an error that the average user makes very often and that he is surely not aware of. It is nothing more than naming the term TDP for a CPU and a GPU with the same meaning, because no, they are not the same, although they are related.

TDP vs. TGP vs. TBP, what are the differences?

In a CPU the TDP is translated as Thermal Design Power and refers to the thermal design that said processor has and that it is necessary to cool, but in GPU the term has a double connotation from a technical point of view.

TDP

In a TDP graphics card it does not mean Thermal Design Power, but Thermal Design Parameter , which has to do and is directly related to the term GPU Power, which is currently deprecated, but which comes to refer to the same thing.

The TDP of a graphics card (GPU really as such) is defined by the manufacturer and refers to the thermal parameter that the GPU generates, that is, the heat that it gives off in the form of watts and is therefore defined with a parameter such as watts.

This is the parameter most used by everyone and as such, manufacturers advertise their models with it, but currently there are two more that are even more specific, since knowing the consumption of the GPU is only giving the information in a way incomplete.

TGP

Therefore, the TGP is the acronym for Total Graphic Power and is understood in only one way: the consumption of the GPU and its entire PCB, but without its cooling and lighting system. Formerly the TGP was called Total Board Power, but this parameter changed its definition for what was discussed just below.

In other words, the TBP referred to the total peak consumption of the graphics card itself, with all its dissipation systems and LEDs included. Therefore, TGP has ended up being more specific, being a value that makes more reference to the card itself. To understand it better, it is the typical case of the reference PCB that all manufacturers use, but each one adds its own dissipation and lighting system.

Therefore, the TGP of those models would be the same because they share PCBs, instead, they would have different TBP. TBP

This parameter refers to the acronym for Total Board Power and differs from the TDP or TGP in a very basic aspect: it affects the total and peak consumption of the graphics card. What does this refer to specifically? Well, we can summarize it as the heat output given off by the entire card as a whole, including auxiliary systems such as LEDs, fans and more.

Here they enter from the GDDR6 memories, passing through the HBM and its versions, voltage controllers, VRM, coils, capacitors and even the LEDs, fans or independent consumption systems that the graphics card itself has.

In summary, if we want to know how much a graphics card will consume as a general and functional concept of it, we have to look at its TBP. If we only want to know the consumption of the GPU only, we will look at the TDP and if we want to know GPU + PCB without the auxiliary systems we will look at the TGP.

[leyvin]

TDP stands for Thermal Dissipation Power or Thermal Design Point.

This is essentially the "Optimal" Thermal (Wattage) Dissipation required by the CPU/GPU under-load.

This is linked to the Power Consumption., but is more specifically in relation to the Cooler Required.

Now as a note if you have a 65w TDP Processor (Temperature Max. 82c) and a 65w Rated Cooler, such-as the Wraith Stealth... then it will be capable of keeping said Cooler running Optimally, albeit close to 82c and likely at a very high Fan RPM (so far from "Quietly").

The "Peak" Draw for TDP is ~1.46x... meaning that let's say we have a 65w TDP Processor, it can Draw up to 95w.

It is common to call the "Peak Draw" either TPD / TDW., but for a while AMD and Intel were somewhat using these interchangeably for marketing purposes.

Now when it comes to Graphics Processors., it's a bit pointless listing the TDP as you have no real option when it comes to the Cooling Solution. Still as a keynote... something like the R9 FURY (500w) hits ~270w TDP., which believe it or not is just below what a High-Performance (usually a Server / Workstation CPU Cooler) is capable of as those peak at 280w TDP and that's one of the Highest Draw GPUs to date.

Now TBP (which is a Metric AMD have adopted since switching the 'RX' Schema) means the Total / Typical Board Power., this specifically refers to the Power Draw of the GPU, VRAM, Display Engine, Media Engine and Fans.

So for example the RX 5700 XT has a 220w TBP., but of course it rarely, if ever actually uses that.

The only time you do actually see such is in Synthetic Tests that push the GPU to it's Operational Limits.

Thus the Fans are at their Peak, Memory is at Peak, PCIe is at Peak and GPU is near 100% Utilisation.

In said scenario however, realistically (for Navi 10 XT) you are looking more specifically at the GPU SoC only drawing ~185w... this means it's producing ~100w TDP.

To put that into perspective, that's roughly what most Mainstream CPUs are producing.

Most CPU Coolers will easily dissipate that, and often much more quietly than GPU Cooling Solutions. 

What you'll find is most GPU Cooling Solutions are actually massively overengineered but seriously inefficient., and this isn't strictly because they're inefficient designs by nature but rather because of the Form Factor they have to take.

I'd argue if we're going to stick with the All-in-One Graphics Board Solution., it's about time that AMD / NVIDIA actually invest some of their R&D Budget into a Cost-Efficient All-in-One Liquid Cooler approach; as it works better in the Form Factor that Graphics Cards take.

The reason such isn't used is because their Expensive Solutions,. part of this is because it isn't widely used; so you simply have low unit production driving costs up... but I also think the Liquid Cooler Solutions we have right now are really focused on being Enthusiast and "Showy" thus again adding to the overall cost as opposed to being bespoke, efficient solutions.

[elstaci]

I am familiar with TDP but never heard of TGP until I started reading about it in articles concerning the new GPUs.  But thanks for the detail technical explanation. Very informative.