AMD Ryzen 2000H Series APUs for Mainstream Notebooks Spark TDP Debate Again

[btarunr]

AMD introduced the Ryzen 2000H series APUs for mainstream notebooks. These chips are physically identical to the Ryzen 2000U series designed for ultraportable notebooks and convertibles; but come with higher CPU and iGPU clock speeds, and hence a higher TDP. The lineup includes two models for now, the Ryzen 7 2800H, and the Ryzen 5 2600H, both of which are based on the same 14 nm "Raven Ridge" silicon as the Ryzen 2000U series.

The 2800H features a 4-core/8-thread "Zen" CPU, with 512 KB L2 cache per core, and 4 MB of shared L3 cache; with clock speeds of 3.30 GHz, with 3.80 GHz maximum boost. The iGPU is a Radeon Vega 11, with 704 stream processors, and engine clocks of up to 1.30 GHz. If you'll recall, the Ryzen 7 2700U has very similar specifications, but only differs with a lower CPU nominal clock speed of 2.20 GHz (but same boost clocks), and one of the 11 Vega NGCUs being disabled. The difference in TDP between the two chips is enormous - 45W default TDP with configurable TDP as low as 35W for the 2800H; while the 2700U is just 15W default TDP, with configurable TDP as low as 12W.



The story repeats with the Ryzen 5 2600H. This chip has the same 4-core/8-thread CPU configuration as its Ryzen 7 counterpart, but with lower CPU clocks, and a slower iGPU that has just 8 NGCUs translating to 512 stream processors, clocked at 1.10 GHz engine clocks. The CPU ticks at 3.20 GHz nominal with 3.60 GHz maximum boost. The Ryzen 5 2500U, again, only has lower nominal clocks at 2.00 GHz, and even has the same iGPU core configuration; but the difference in rated TDP is huge: 45W vs. 15W.

So just how is it that the enabling of a tiny few components or increasing nominal clock speeds have such a tremendous impact on TDP? Perhaps there are other under-the-hood settings these chips have that make them more eager than their U-series siblings. Also, nominal clocks are clock speeds that each of the four CPU cores on the chip are guaranteed to run at, beyond which, depending on a number of factors, the PrecisionBoost algorithm awards higher clocks. On the U-series chips, PrecisionBoost is extremely conservative with boost clocks. Even so, could clock speeds really have such a profound impact on TDP? Join the debate in the comments below.

View at TechPowerUp Main Site

 

[SIGSEGV]

The lineup includes two models for now, the Ryzen 7 2800H ....
so, where's Ryzen 7 2800X hiding at?

 

[Steevo]

Cache takes a lot of power to run.

 

[R0H1T]

You know with "regular" (Intel) H processors ~ the CPU can turbo for longer, the IGP also gets more "TDP" headroom which ULV chips don't provide! All this assuming the laptop has "good enough" cooling, no controversy here folks move along

 

[_Flare]

So use their native DDR4 3200 with 2x 8GB and a 1TB SSD and drive the full cTDP 54W, or go home Manufacturers.

UNLEASH THE RAVEN

 

[Darmok N Jalad]

I don’t think it’s that big of a mystery. A 25W APU is not going to run at sustained boost speeds on either the CPU or GPU as much as a 45W APU will. If one can bench both products, you will probably see similar scores on short workloads, but much more noticeable improvements on sustained workloads.

 

[silentbogo]

So just how is it that the enabling of a tiny few components or increasing nominal clock speeds have such a tremendous impact on TDP? Perhaps there are other under-the-hood settings these chips have that make them more eager than their U-series siblings. Also, nominal clocks are clock speeds that each of the four CPU cores on the chip are guaranteed to run at, beyond which, depending on a number of factors, the PrecisionBoost algorithm awards higher clocks. On the U-series chips, PrecisionBoost is extremely conservative with boost clocks. Even so, could clock speeds really have such a profound impact on TDP? Join the debate in the comments below.

It's not as simple as "increasing clocks" on identical silicon. H-series are binned to run at higher clocks, while semi-rejects go into U series. This differentiation between the two is a good step forward for AMD, cause you should remember the confusion over FX-9800p, where AMD did put a lot of effort into marketing this high-performance 35W chip which can play Dota 2 in FHD, and then HP f#$%ing it up by capping it at 15W in almost all of their products... as a result - lots of angry customers who paid upwards of $600 for a netbook-level performance.

 

[TheoneandonlyMrK]

You know with "regular" (Intel) H processors ~ the CPU can turbo for longer, the IGP also gets more "TDP" headroom which ULV chips don't provide! All this assuming the laptop has "good enough" cooling, no controversy here folks move along

Exactly , one's cooled to allow 15W tdp The other has significantly better cooling to hold a 45w tdp.
As you say more turbo , and everyone loves more.
They're are also power save modes via Os that could help with efficiency so all good , no secret sauce here just binning for teired products.
Not all laptops are equal as my merrydale tablet (win10)thing would adequately demonstrate.

 

[FordGT90Concept]

Hugely different clock speeds are mostly the cause of the TDP difference. Really nothing special about that.

I do like how they breakdown the TDP. If the HSF isn't designed to handle 54w all of the time, the CPU may be expected to thermal throttle.

 

[NC37]

It's not as simple as "increasing clocks" on identical silicon. H-series are binned to run at higher clocks, while semi-rejects go into U series. This differentiation between the two is a good step forward for AMD, cause you should remember the confusion over FX-9800p, where AMD did put a lot of effort into marketing this high-performance 35W chip which can play Dota 2 in FHD, and then HP f#$%ing it up by capping it at 15W in almost all of their products... as a result - lots of angry customers who paid upwards of $600 for a netbook-level performance.

Lenovo did the same thing too with the i5-8250 and the MX150s. Naturally Ryzen models were similar. Really gotta read a lot of customer forums to find out how bad some of these laptops are gimped. Dishonest marketing reigns supreme.

 

[king of swag187]

Looks interesting, its competitive to a 8300H/8250U/7700HQ and 1050 laptops and yet in the TDP of just the CPU, interesting. While not offering top performance (ofc its an APU) this could easily power laptops for hours with a full 97WH battery

 

[Valantar]

Is this really worth a news post?

15W chips have to balance their tiny power limit between the CPU and iGPU. Under gaming, that means throttling both of them. I've seen 15W thin-and-light Raven Ridge laptops reported to run the GPU between 350-500MHz while gaming. Increasing the TDP to 25W on the same chip absolutely helps, but still puts a hard limit on performance.

The TDP of these parts should thus allow both the CPU and iGPU to run at or above their rated speeds without issue. For the iGPU, that means no throttling, while for the CPU, it might mean boosting above base clock even while the GPU is under load. Neither of these are even remotely possible for mixed workloads like gaming in the 15-25W space. And you know what? That's fine. It's not a problem, as we're still getting the best performance possible in that thermal/power envelope at this time. Releasing binned-for-clocks-and-power (not just low power like the U-series, or clocks regardless of power like the G-series) SKUs like this to allow the silicon to stretch its legs in a mobile form factor is a very good move. No different than Intel's 28+W Iris SKUs in previous generations, really. Well, except for those still having weaksauce GPUs, of course.

 

[medi01]

"We needed to post something bad about excellent chips, so here you go"