Monday, January 6th 2020
AMD Announces Ryzen 4000 Mobile Processors: 4800U and 4800H
AMD today announced its Ryzen 4000-series mobile processors designed to compete with Intel's fastest, across both its 10th gen "Ice Lake" and "Comet Lake" mobile processors lines. At the heart of these processors is the 7 nm "Renoir" silicon, which doubles the CPU core count over the previous generation "Picasso," and improves IPC (single-thread performance) by a double-digit percentage. "Renoir" combines a CPU with 8 cores based on the "Zen 2" microarchitecture, with an iGPU that has the number-crunching machinery of "Vega," but with display- and multimedia-engines of "Navi." It is a monolithic piece of silicon with a dual-channel IMC that supports not just conventional DDR4 memory, but also fast LPDDR4X.
There are two distinct classes of Ryzen 4000 Mobile: U and H. The Ryzen 7 4800U, with its 15 W TDP, targets ultra-portable notebooks, and goes head-on against Intel's Core i7 "Ice Lake-U" processors, winning on the CPU front with its high core-count and IPC. The Ryzen 7 4800H, on the other hand, taps into the 45 W TDP headroom to dial up CPU and iGPU clock-speeds significantly, offering CPU performance that beats the desktop Core i7-9700K. It also introduces SmartShift, an iGPU + dGPU virtualization technology that lets your notebook dynamically switch between the two based on graphics load.
There are two distinct classes of Ryzen 4000 Mobile: U and H. The Ryzen 7 4800U, with its 15 W TDP, targets ultra-portable notebooks, and goes head-on against Intel's Core i7 "Ice Lake-U" processors, winning on the CPU front with its high core-count and IPC. The Ryzen 7 4800H, on the other hand, taps into the 45 W TDP headroom to dial up CPU and iGPU clock-speeds significantly, offering CPU performance that beats the desktop Core i7-9700K. It also introduces SmartShift, an iGPU + dGPU virtualization technology that lets your notebook dynamically switch between the two based on graphics load.
53 Comments on AMD Announces Ryzen 4000 Mobile Processors: 4800U and 4800H
Considerations for 15W APUs are entirely different as the systems they are used in are entirely different. First off, these systems are entirely designed for low power (unlike laptops for the H-series, which try to balance very high performance and middling power), but are also trying to maximise performance within that low power envelope. That means setting high target clocks and then power throttling. Also, 15W chips are far less likely to be paired with a dGPU (though it does of course happen from time to time), meaning that the actual performance of the iGPU becomes more important than on the 45W parts. A 15W APU iGPU might be used for gaming, while a 45W APU iGPU will in all likelihood not be.
As for boost clocks: these are maximum single-core boost clocks. If you don't think the 45W chips will boost higher and for longer than the 15W ones, you don't understand how power and clock balancing in modern PCs work. A 15W part will not match the performance of a 45W chip outside of a few edge cases, despite the spec sheets listing similar performance.I wasn't stating my opinion, I was referencing statements from the OEMs. Don't put words in my mouth.Again: have I said so? But ultimately, which laptops are made and which sell well is a complex balance in which both customer demands and OEM desires play a part.Yes, and? How does this apply to what I was saying?Wow. Nice of you to assume I don't know something as basic as components downclocking when idle. Classy. But beyond that, a large GPU can never be as power efficient as a small integrated GPU - not least of all because having a dGPU active means you have another large piece of silicon with its own RAM and power delivery circuitry active, no matter how idle it is. 15-25W is the entire power budget of a small laptop, or half the TDP of the 45W APUs. You see how power draws on this level might be a problem, right? It will always be more efficient to switch to a smaller and less powerful integrated GPU in non-performance sensitive use cases. An iGPU in desktop use might consume anywhere from a fraction of a watt to a handful, not 15-25. That's an order of magnitude less, and a change that has massive implications for battery life.
iGPU save power but like I already told you, so can discrete card save power so this is irrelevant.I wasn't putting any words in your mouth and you haven't stated that this is a reference to an OEM's statement which I haven't seen in a form of an article. So if you quoting or referring to an article say that you are and maybe share a link to the article(or whatever form it is) so that when somebody reads it knows, that this is a reference not your own conclusion.That's the thing. It doesn't because I wasn't interested in what you said, previously OEM want to have low power laptops and don't want to use discrete cards. I mentioned it because according to You and OEM nobody wants discrete cards in laptops because high power needs.Of course it can't be as energy efficient but you also have more possibilities in terms of gaming making your laptop more flexible.
So you are saying that they have pushed the CPU to 45W with lower frequency of the iGPU and less CU so that you can save power when not using dGPU because your computer is idling?
my vega uses 15-20W when idling. and yet you have 45W CPU bumped from 15W just to save power? and the iGPU is slower in the 45W.I see why but i still think you are scrambling. Because instead of getting a 45W cpu with slower iGPU you can use 15W CPU with better iGPU and slap a dGPU in there instead. You already save power because the CPU you are using, uses three times less power.
And that was my question. Why 45W with slower iGPU when you have 15W with better iGPU. 45W with discrete card for power efficiency is nonsense because you can use discrete card with the 15W CPU no problem and even more efficient than 45W.
At least that is how I see it.
What I have been trying to explain to you this entire time is the rationale behind giving the low-power SKU a more powerful iGPU than the high-power SKU, but you seem entirely resistant to even understanding that this is the question I've been trying to answer for you (which you posed in the first post I responded to). Please try reading my response more carefully so that we might understand each other.Wait, did I say the 45W chips don't have iGPUs? Have you even read my posts? I'm trying to explain to you why it makes sense for the 45W chips to have a less powerful iGPU than the 15W ones. Put into plain language:
15W APUs will in the vast majority of cases not have a dGPU attached.
45W APUs will in the vast majority of cases have a dGPU attached.
15W APUs have a very tight power limit, but people still want passable GPU performance.
45W APUs have much looser power limits, and have attached dGPUs, so iGPU performance doesn't matter as much.
15W APUs will have good battery life, so sacrificing a little for better iGPU performance makes sense.
45W APUs will not have good battery life, so saving power at low loads by scaling down the iGPU by a bit makes sense.
15W APUs have a tight power budget and (normally) no dGPU, so they need to maximize performance for all scenarios as best is possible. For gaming, this means prioritizing GPU power, as few games are CPU limited below high frame rates that an iGPU won't reach.
45W APUs still have a power limited CPU, but they (normally) have a dGPU to take care of 3D workloads, meaning prioritizing power for the CPU part of the APU improves overall performance.
I sincerely hope that makes it clearer for you.Yes you were. Is there anything in my statement indicating that it is my opinion? Such as me saying "I think" or "I believe"? No, but you claimed that this was my opinion. It isn't. So you were putting words in my mouth.Asking for sourcing is fine, and I should probably have provided it right away. Then again, asking about it is probably what you should have done, rather than putting words in my mouth. Here, from AnandTech:What? I never said anything even remotely close to that. Can you please reready my post? I said OEMs specifically don't want to prioritize iGPU performance in 45W APUs because the laptops using these have dGPUs to do that job.Yes, absolutely. But there's nothing at all stopping you from doing that while still switching to a low-power iGPU when the dGPU isn't needed - in which case it makes perfect sense for the iGPU to be as low power as possible, no?No. But you are so close. Here, in the fixed version:After all there are quite a few workloads where CPU power is useful but GPU power isn't, so having a lower power iGPU allows for better performance in these scenarios.Please reread my posts. This is nonsense.But the CPU will be dramatically slower! Do you seriously think a 15W 8-core APU with a 4.2GHz max single-core boost and a 45W 8-core APU with a 4.2GHz max single-core boost will run at the same sustained turbo clocks even with one core active, let alone several? The 45W APUs will be much faster, as that's how multi-core turbo works. Single-core turbo is high on the 15W parts to make the PC responsive in bursty workloads (say, opening an application), but it will slow down A LOT to meet its 15W TDP. The 45W part will run its CPU much faster for the vast majority of time.
You should really read some articles exploring modern boost and turbo systems, as you clearly don't understand how these work, and therefore misread the spec sheets entirely. 4.2GHz max boost does not necessarily mean the CPU can sustain 4.2GHz for more than a short period, before power and thermal limits take over and slow things down. And looser power and thermal limits, as a laptop designed for a 45W APU rather than a 15W one will thus allow the APU to boost higher for longer periods of time. They might max out at the same level, but for anything beyond short bursts the 45W will be dramatically faster. Why? Because the 45W has a higher base clock.
Base clock is the "guaranteed" all-core clock speed that the CPU should deliver. And if you look at the two 8c16t chips there, the 15W is 1.8GHz while the 45W is 2.9GHz - a 61% advantage to the 45W part.Again, you fundamentally misunderstand the logic behind this, not to mention misread what I'm saying.
The 15W has a "better" iGPU mostly because its performance matters more in the systems it will be used in, as it's the only GPU most of these systems will have. Maximizing iGPU performance thus has a point, and sacrificing some power to get this makes sense.
On the other hand, 45W APUs will (mostly) have dGPUs, so iGPU performance doesn't actually matter (as the dGPU will do anything performance-sensitive), and it thus makes more sense to make the iGPU (specifically, and not anything else) smaller and more efficient so that more power can be allocated to the CPU and dGPU, or battery life can be extended when not gaming (as dGPUs don't scale low enough to allow for good battery life). The CPU will run faster than the 15W anyhow, but saving power on the iGPU means the CPU can consume more, and thus be even faster.
You seem to assume that the 45W part is competing against the 15W part, but that's not the case. The 45W part will be competing against 45W Intel parts, in similar PCs (i.e. relatively thick and heavy high-performance laptops) while the 15W parts will be competing against 15W Intel parts, again in similar PCs (mostly thin-and-lights). If reading the spec sheet makes you think the 15W chip will outperform the 45W chip in absolute performance, you are fundamentally misinformed as to how these things work.
Since laptops are bigger than desktops as a market - just as phones and tablets are bigger than laptops - and that was where Intel was still in the lead, this is a big deal.
Of course, now we know Intel's 10nm+ is coming soon. But before things get better for Intel, they are going to get worse.
Why in your opinion 15W will not have dGPU but the 45W will have one? Because the processor uses more power so the dGPU can be used? The dGPU uses it's own power source it has nothing to do with processor because the dGPU is integrated in the motherboard. Unless you are trying to tell me something else but I don't get what it is.You said it here. OEMs' are not interested in making 45W chips because they would rather focus on power saving GPUs. Slimming down to use less power?I've no idea where you have taken that but I disagree. Lower power better performance and it doesn't matter if it is CPU, iGPU or GPU it is not true. Maybe you misspoken because that is just crazy.I don't need to reread anything. I simply disagree with you but you don't want to take it. What am I suppose to do with this.
It isn't nonsense. 15-25W when idling. Using browser, skyping. I'm not talking about my Vega but the conversation you had. 45W with dGPU 15W without because? .... Power saving? Now that is nonsense.Slower? Elaborate on that because how I see it and how the specs say it, these two are exactly the same. The only difference is iGPUis faster and base clock for the 15W is lower. Both have same Turbo. 8c16t 4200Mhz both. Yet 4800U will be slower. How did you get into that conclusion?
Would it be possible that the 4800U is better quality and can sustain 4200Mhz with less voltage? Would that also contribute to less power draw? Like better silicon quality for instance?
If the 15W power draw envelope would not allow 4.2Ghz or the CPU couldn't sustain this speeds with given power envelope, the CPU would not be spec as one. The max boost time does not make the processor use more power.I simply disagree with this.Well, you don't have the new Ryzen but I'm sure, there will be more efficient R9 coming out for desktops but there will definitely not be a R9(R9 meaning more than 8c although I'd welcome 8c16t with better iGPU) with iGPUs for the mobile segment. I assume that is what you are after?
The thing is, on the other hand, 2700 65W and 2700x 105W and 4800 is based on that arch but includes an iGPU. The power consumption is substantially lower so maybe EUV has been introduced here after all?
I said the 4800 APUs are based on the 2700 series architecture not share same node.
I'm trying to deduct this EUV DUV by comparing the TDP of the new CPUs with previous release desktops to get more less an answer. Not sure what you are doing.
non EUV it is for sure now.
So, the basics:
"TDP" for both manufacturers for mobile parts roughly means "maximum sustained long-term power draw", though the number is actually a target for OEMs to design laptop cooling solutions around. There are BIOS power limits and other mechanisms involved in enforcing this, and the limits can be adjusted and tuned by OEMs.
For Intel, this means that a 15W CPU will (roughly) be capped at 15W for sustained power (say, gaming, a long render, etc.), but can use far more than this for short-term bursts (opening an application, the first seconds of said render, etc.). Intel's 15W CPUs routinely use 45-55W in short-term Turbo modes, dependent on how the laptop in question is configured. This is how peak turbo speeds are achieved.
This brings us to turbo/boost specs: these are peak numbers, not sustained numbers. Base clock is the number which tells you the guaranteed* sustained all-core clock speeds during a CPU-only load** for a given part. A 1.8GHz base, 4.2GHz boost part like the 4800U will hit 4.2GHz only for short periods of time and in single-core or lightly threaded loads, with peak turbo decreasing dependent on the number of cores being stressed. For Intel this is based on a fixed turbo table with max limits, while AMD scales this dynamically based on available power and thermal headroom. You are only guaranteed that the CPU won't throttle below 1.8GHz, though in reality (again, dependent on the laptop in question) it is likely that it will run faster than this indefinitely. A good source to look into this is an in-depth laptop review like the ones NotebookCheck does, where they look at both peak and sustained power draw numbers for each laptop.
*again, dependent on OEM choices when it comes to laptops, some ultrathin laptops might have cooling solutions that cause clocks to drop below base clock, though that's rather shady.
**A CPU+GPU load will likely cause CPU clocks to drop even lower - many 15W Intel chips with base clocks of ~1.8-2.5GHz drop to ~1-1.6GHz when the iGPU is stressed.
This is the same for desktops - only base clock is guaranteed within TDP - but most desktop motherboards either allow you to configure them to ignore TDP or come configured like this out of the box, meaning that they sacrifice power to maximize performance.
So, for a part like the 4800U, the 4.2GHz number is only something you will see as long as it's possible to maintain within the power and cooling limits of the laptop in question. When those limits are met or exceeded the CPU will slow down while still trying to maximize its clock speed to what can be achieved within its limits - i.e. no more than 15W sustained power draw and whatever the temperature/fan speed limits of the laptop is. This means that clock speeds will drop unless there is power and thermal headroom left over. Which, for a 15W thin-and-light laptop, there never is.
Again, this is easily demonstrated by running longer benchmarks, such as when NotebookCheck runs subsequent Cinebench renders and logs how the score drops over time. Note how scores for both Intel and AMD chips drop dramatically within the first 1-3 runs - this is exactly due to the chips' peak turbo power windows being exceeded and the chips therefore slowing down to reach their steady-state, sustainable performance levels.
This is (mainly) why any 45W chip will (normally) outperform any 15W chip - because it's allowed to use more power, clock higher, and has a cooling system built to handle a bigger thermal load. Peak turbo numbers might be the same or even higher for the 15W part, but sustained (base clock) numbers will always be higher for the 45W part.
You're of course right that binning does affect things, and U-series chips do tend to be better binned (for low power) than H-series chips, but binning doesn't work miracles, and certainly won't allow a 15W chip to outperform a 45W chip in any kind of sustained workload. If that was the case, the silicon manufacturer would be having some very serious quality or design issues. There are cases where a top-bin 15W-class chip, let's say an i7, has matched or even beaten a low-binned 45W-class chip, let's say an i3 or i5, but in those cases the 45W-class chip is usually actually limited by its spec sheet rather than power or thermals - i.e. it isn't allowed to boost as high as its power budget would allow. It's quite well documented that despite having the same TDP and better binning, higher-tier CPUs on a given wattage level consume (slightly) more power on average than lower-tier options. A 15W i7 will have slightly worse battery life than a 15W i5 in the same chassis, especially under heavy load.
But, in short: you seem to expect U-series chips to maintain their rated boost speed indefinitely (and for all cores?), which is not how things work in the real world. They will reach that speed for short durations of time (might be a few seconds, might be half a minute or even a full minute in extreme cases), but will always throttle afterwards.No. I said "There might be a 45W R9 at some point" You responded with "There's no way there will be a 12- or 16-core mobile part". See the issue there? You responded to something I didn't say by assuming that an R9 part must be >8 cores. I was simply talking about a top-end 45W SKU above the 4800H, which would mean boosted clock speeds and enabling the last CU on the iGPU. I never mentioned higher core counts or anything of the sort, which is why your response misses the mark. There's no rule saying an R9 must be >8 cores, just as there's no rule saying an R7 must be X number of cores. R7 mobile chips in the 3000-series are 4c8t, after all, while the desktop chips are 8c16t. R[number] does nothing more than indicate a broad "performance tier", i.e. an R7 is generally faster than an R5, an R9 faster than an R7, and so on.Jesus. The dGPU power is irrelevant. The relevant point is that laptops using 45W chips will be built for heavier loads, with beefier cooling, so that they can sustain higher boost clocks for longer. The laptop designs will be dramatically different. The point of a power saving iGPU in this case is twofold: allowing the CPU to consume more power (the iGPU isn't being used when the dGPU is in use, so any drop in power from it goes directly into increasing CPU speeds), and increasing battery life when the dGPU isn't in use (as these larger laptops tend to have terrible battery life they generally need all the help they can get in this regard). Having a beefier iGPU in this case might therefore cause the laptop to be slower as the CPU would not be able to consume quite as much power, and decrease battery life for general desktop usage.What? No. Stop misquoting me, please. I said OEMs are not interested in making 45W-class laptops without a dGPU (see the AnandTech article about the 4000-series launch), therefore they don't care about peak iGPU performance in 45W chips. This is nothing like what you are saying above.I can't believe I have to spell this out, but here goes. When the iGPU is not being stressed due to the dGPU being used (say, for a game), any power consumed by the iGPU is wasted. This wouldn't be an issue if the iGPU could be disabled entirely, but it can't as it is responsible for the display controller and other central functions. (Large parts of it are disabled or clock gated, but unless power gated they still consume some power.) This is true of any laptop with switchable graphics. So, in this case, making the iGPU consume less power means the CPU (which shares the same power budget) can consume more power and thus perform better. The dGPU is entirely separate from this, but is what makes iGPU performance irrelevant.Yes, you do need to reread, as you are fundamentally misunderstanding what I am saying. I'm not saying that 45W chips with dGPUs save power compared to 15W chips without them. If you are reading that, you seriously need to work on your reading comprehension skills. I am saying that in a laptop with a dGPU, a 45W APU with a lower power iGPU will have better battery life and better CPU performance than a 45W APU with a higher power iGPU.
I never questioned your power numbers for your desktop graphics card either - they're quite reasonable - I simply stated that this is idle power at 15-25W, which is half the full load power budget of a laptop CPU. If your laptop consumes >20W when idle, you have a serious problem. Which is why laptops need iGPUs to save power, even when they have dGPUs for high performance, as the dGPU can never scale to as low idle power levels as the iGPU.This is a rather long part to "simply disagree" with. Care to clarify or elaborate? Do you disagree with the OEM-stated fact that they are not interested in making dGPU-less 45W laptops? Do you disagree with the fact that 15W thin-and-light laptops do not compete with 45W high performance laptops? Do you disagree with the fact of how opportunistic boost/turbo scaling algorithms work?
And this. in red mine to point out with what i have problem.I don't have problem understanding. I comprehend everything you said correctly but your way of expressing yourself leaves unsatisfying urge for me to tell you be more specific because what you say is not what you mean and vice versa. At least.Yes you didn't say 12c or 16c. I said it assuming you were talking about R9 (analogy to Ryzen 9 which is 12 and above) . next time say you are talking about iGPU and there will be no problem.
BTW, the fully enabled iGPU is the 15W 4800U. Unless there's more than 8CU?Misquoted? it is exactly what you said.
here.Though you mentioned AMD but there is no link so..... Btw the entire sentence is kinda weirdly written so God knows what you have in mind here.I think you mean power balance/draw between the iGPU and the CPU. When the iGPU is not being used the CPU get's all the power?You have misspoken using power instead of performance. It didn't make sense "CPU power is useful and GPU power isn't. and we move on to your last sentence which makes even more confusion but I think I get what you are trying to say. I hope, considering our conversation history, I'm not so sure now.Whatever. No time for essays here.I'd love to but I'm tired of trying to understand what you wrote or what you meant. Honestly, just let it go and move on.
For example: The Core i7 9850H has 6 cores, 12 threads, a 45W TDP and a max boost clock of 4.6GHz. The Core i7-10710U has 6 cores, 12 threads, a 15W TDP and a max boost clock of 4.7 GHz*. Yet the 9850H can sustain Cinebench R15 scores of ~1100 given sufficient cooling (there's a link there) while the 10710U even in it's 25W cTDP-up mode in the Dell XPS 13 fluctuates between ~820 and ~950, never quite flattening out. I couldn't find a comparable review of the 10710U in 15W mode, sadly (the MSI Prestige 14 pushes it to ~30W, and the Prestige 15 goes even further, again in-line links in both of the names). In other words, there's no correlation between boost clock specs and sustained performance, particularly across different TDP classes. And a higher TDP especially improves sustained performance.
*Yes, these chips are of different generations and marginally different silicon, but both are made on the same process node (Intel 14nm++), are based on the same architecture (Skylake) and the biggest differences in silicon are the number of PCIe lanes and memory support - they even have the same amount of cache.I never said 15W APUs can't have dGPUs - I very specifically said that they mostly won't. That's a substantial difference, no? Beyond that, you keep making this 15W-to-45W comparison, which is not what I'm arguing at all. What I'm saying is that a 45W APU with a slightly toned-down iGPU benefits from this by having better battery life in low loads than a similar 45W APU with a fully-enabled and higher clocked iGPU would have. 15W APUs will still have better battery life by virtue of having much stricter power limits. You are misreading me and turning this into a straw man argument. Please stop.
Put very simply:
A 45W APU with 7 CUs would have better battery life in normal desktop usage than a 45W APU with 8 CUs given that everything else (clock speeds etc.) are equal. I never compared this to 15W APUs.
Silicon quality can do a lot, but it can't make a 15W APU match the sustained performance of a 45W one. They are in different performance classes. That does not mean that the 45W will be 3x the performance, obviously, as clock speeds don't scale linearly with power draw, but the 45W chip will be able to boost higher for longer and thus perform better overall.No. You don't. I have read your responses, and they make no sense whatsoever based on what I'm saying, and you are continuing to attribute statements and claims to me which I have not made. So no, it is quite impossible that you are understanding what I'm saying. I have pointed out a number of clear-cut and relatively simple misunderstandings both here and previously, but you have handily skipped over every single one of these clarifiations. I'm not going to speculate as to why, but this is a direct hindrance to actually understanding each other, and this is entirely on you.Wait, I'm not saying what I mean? Are you a mind reader? Beyond that, you're welcome to ask me for any clarification you might want, but you don't tend to do that, but rather seem to prefer paraphrasing what I'm saying in fundamentally mistaken ways and then arguing against it.Why should that be necessary? We're talking about mobile APUs. An APU has an iGPU - that's what makes it an APU and not a CPU. You're the one that made a significant logical leap here, in assuming that an R9 in this context (that context being mobile, APU-centric, reated to the specific piece of silicon that underpins the 4000-series APUs) would somehow be a >8c CPU and not an APU. I never did or said anything to indicate that I meant anything more than a higher-binned and higher-clocked SKU of the current silicon.Yes, it is fully enabled (as far as we know, at least). As I have been saying the whole time: we don't have a fully enabled 45W APU. You are somehow arguing against that by saying "But we do, it's the 15W 4800U!" Which, to be clear, is not a 45W APU. Are we really going to keep going in circles here? A 15W APU is not a 45W APU.Yes, that is what I said. But then you paraphrased this into the following:Which is not at all the same. Seriously, please read the two sentences, one after the other, and tell me if they say the same thing. You're saying I said "OEMs are not interested in making 45W chips..." while what I said was "OEMs don't seem interested in making dGPU-less laptops with the 45W chips". You see how these sentences say fundamentally different things, right?
You then asked for a source, which I provided:The link is in "AnandTech" directly before where I then quoted their article. Sorry if you missed that, but putting source links in attributions before quotes is quite standard practice across the internet.See above. There is most definitely a link. And even a direct quote from the article! It sounds like you didn't actually read my post ...Yes, that balance is the point of this entire argument. But no, the CPU does not get all the power when the iGPU is not in use, as the iGPU is still running the display controller and performing various other (non-3D/compute/rendering) tasks when the dGPU is active. Unless the iGPU's compute units are power gated (which they haven't been in designs up until now, at least), they will still be consuming some power - which the CPU then can't use. Which lowers CPU performance compared to how high it would then be with a smaller iGPU.
Let's illustrate this. While these numbers are made up for this example, they shouldn't be unreasonably far from reality.
The APU has a 45W total power budget for sustained loads. Of this, it's not unreasonable to think that 5-10W goes to SoC/non- CPU core power dependent on the workload even when the iGPU is idle. Why?
-the 3D/compute portion of the iGPU is idle, but likely not power gated, just clock gated. This means it's on, just running at a very slow speed, thus still consuming power.
-the non-compute portion of the iGPU is (partially) active, running the display output(s) and other fundamental functionality.
-other SoC tasks use the APU's power budget, so memory controllers, SATA, PCIe, Infinity Fabric, etc. all consume some power. Communicating with a dGPU over PCIe consumes power. Communicating with an SSD over PCIe or SATA consumes power (I'm not talking about the SSD's power consumption, which is separate and unrelated, but PCIe and SATA link power, which comes from the APU's power budget).
So, that means that the CPU is at all times unable to use a certain portion of the APU's power budget. It thus stands to reason that any reduction to the above power draws will allow the CPU cores to draw more power within the 45W max budget, and thus perform better. The difference might not be a lot, but it's there.
Now, which of the above points can be changed? PCIe link power can't be changed, as you'd either need to disable your dGPU or SSD, which ... well, wouldn't let you do much with your PC. Display controller duties and other fundamental I/O that the iGPU handles can't be disabled either, as ... well, you wouldn't have a working display. So the only thing that can scale down is the idle iGPU. Disabling CUs in this directly correlates to a drop in idle power draw, as disabled CUs don't use power.
So: disabling a CU or two in a 45W APU might therefore allow the CPU cores to perform ever-so-slightly faster than if the same APU had all CUs enabled. As an added benefit, this partially disabled iGPU would consume less power when active compared to a fully enabled version when in use, improving battery life when the dGPU isn't in use.You're right here, I said "power" in some places where I meant "performance". Sorry about that, and it's a useful clarification.That's a massive cop-out if I ever saw one. You're entirely welcome to bow out of the discussion any time you like, but don't project this onto me.
Why 4800H exist? Simple it is a different tier of laptop chip. Since this is a laptop segment the most important thing is power draw. This has 45W and the U version 15W for ultrabooks. I already mentioned it but you must have missed it.
The problem here is, your way of expressing yourself confuses me and that is why it is better to let the conversation go. I understand what you said the way you wrote it not they way you think it should have been understood or what's on your mind and I'm tired of guessing :)
It doesn't matter now. Back on topic please.
If you claim that this is all down to silicon quality, and that all 45W parts are simply less efficient bins that the 15W parts: this is both inaccurate and an oversimplification. "Power efficiency" is not one simple linear variable in silicon, but several combined. Some parts clock high at low voltages. Some parts clock high at higher voltages. Some parts refuse to clock high at all. Some parts may clock very high, but need high voltages even for lower clock speeds. Some parts may run at very low voltages, but fail to reach clock targets for high-bin parts regardless of voltage. Some parts are efficient at low speeds, but inefficient at high speeds. The differences are complex and often very subtle, with many variables in play. This is why there are both multiple tiers and multiple performance levels within each tier. A die used for a 4800U might not be able to sustain the base clock of a 4800H at 45W power consumption, or a die used for a 4800H might not be able to sustain the base clock of a 4800U at 15W. Some chips are able to do both, and end up being made into whatever is needed. You're presenting this as if 15W chips are all massively more efficient, provide the same performance, and 45W chips are simply all lower binned parts. This is not true. If this was as simple as you seem to be presenting it, there would be a single tier, with the most efficient parts sold as the fastest chips (Ryzen 7, Ryzen 9), while less efficient parts were partially disabled and/or clocked lower and sold as slower chips (Ryzen 5, Ryzen 3).
By the way, can you at least acknowledge that there is such a thing as a base clock specification, and that these are quite different between the 4800U and 4800H? 'Cause these specifications are indeed real, and are 1.8GHz for the 15W part and 2.9GHz for the 45W part - a full 61% higher. I would love to see you explain what the base clock specification means according to your belief that the max boost specification indicates a long-term sustainable clock speed.
Also, if the 4800U performs the same as the 4800H just at 1/3rd the power, why would Asus pay for a 6-month exclusive bin of the 4800H at 35W but the same performance, called the 4800HS? Wouldn't they then simply use the 4800U and be happy? Again: no, because that's not how these things work. And lastly, if 15W parts are more efficient and perform the same, why do they cost roughly the same? Wouldn't that be a premium selling point to make these more efficient parts more expensive?I was talking about your logic in that specific statement: that you say you are confused, yet still claim to understand everything, and that it is me who is failing to express my points rather than you who is misunderstanding. Wouldn't the logical part be that your confusion causes you to not understand? That seems logical to me, at least. I'm sorry if this debate has been confusing, but I've been as clear as is possible without spending an exorbitant amount of time on this, and I've provided extensive background for you to look into, written by people with very high levels of technical expertise. That should be enough.
Also, this is on topic. We're discussing the newly launched 4000-series APUs, after all.
Bar the "performance is the same" (it doesn't have to be), is it really surprising that chips from the same process could be split into better/worse power consumption groups? I.e. "harvested"?
Take better ones and sell them at a premium, not so efficient ones, for smaller sum, instead of having people win/lose lotteries.