AMD Mobile 5000 series

I figured now that "CES" has the info out there on the 5000 series mobile CPUs from AMD we could use a place to chat about it.

Looks like a good line up:


But then AMD does this? Why would they not use all the same architecture when the WHOLE point of skipping 4k series naming on the desktop was to limit confusion on what architecture the processors were on:



If you don't realize that thanks to their usage of the Zen 2 architecture in the 5300u, 5500u, and 5700u the single core performance gain doesn't look impressive over the line up:



At least for the 5600u and 5800u the single core gain is ~17%.

Over all I'm excited for the addition of some serious competition in the mobile market. I'm kind of miffed at AMD's decision to keep some Zen 2 CPUs around to take advantage of the unknowning customer...

 

I wonder what the real world difference is going to be like between the 5980hx and 5900hx

 

Probably nothing, lol. I think AMD was jealous of all of the CPU names Intel had so they just started creating more SKUs.

 

Probably, one would hope for better binning on the premium SKU but that would be wishful thinking until we see them in the wild.

 

Still no Thunderbolt presumably?

 

No, maybe with Thunderbolt 4 in the future, but I don't think AMD is chomping at the bit to pay intel money for their spec.

 

No USB 4/Thunderbolt 4 mentioned (that would be a headline).

Because USB 4/Thunderbolt is now 'free', why would they when they're the cough, cough, 'better option'.

 

Not everyone uses thunderbolt, but great try!

 

Duh, no. But everyone would use USB 4.

 

Right, and when laptops come with it they will have thunderbolt 4 as well.

 

Cool, potential customers can wait another gen I guess.

 

Yeah unless the system integrator decides to do their own thing. AMD has never really shown interest in Thunderbolt. If you really need/want thunderbolt then Intel is the way to go.

 

Probably very little to none realistically speaking. Closer to none given how many thermal constraints laptops 'enjoy' due to OEM's lack of interest in designing proper cooling for these parts in the first place (and updating the cooling solutions in laptops in general).

In all honesty, I think the 5800H would be perfectly fine and a preferable chip for most consumers.
400 MhZ difference in single core boost is just under 10% increase in clock speed... or about 5% increase in actual performance (don't think this would be worth the overall price premium the OEM's would inevitably slap onto those laptops).

 

The only way I see the Premium SKU making a difference is...

1. Binning
2. Actually comes with proper cooling

Both would annoy me personally though

 

The CPU has nothing to do with the cooling. That's is entirely up to the designer. So the Asus G series that has it will probably have decent cooling while the Dells that have it will not, lol.

 

Why would you be annoyed by laptop models with proper cooling? You wouldn't have to buy them if the extra power for extra weight is not the right tradeoff for you.

 

Asus in my experience doesn't have 'decent cooling'.
They utterly botched the GL702ZC when it came to cooling and noise... the thing ended up frying itself after a month of use... and it was a 17" laptop to boot. Good specs, but AWFUL cooling design execution and ridiculously loud under load. I was fortunate when I found the PH517-61. This thing is ridiculously cool and quiet under maxed out loads.

It also doesn't help matters to know that even with Zen 2 models, Asus (and other OEM's) allow the CPU to reach over 90 degrees Celsius when fully stressed (meanwhile, the GPU - whichever they pick - tends to sit at a comfortable 75 to 80 degrees C under full load).
Zen 3 models will probably end up the same, or worse.

This is one of the reasons I'm not enthusiastic about these laptops (not to mention the fact Asus has gone bonkers on the price). The only one that caught my interest in regards to cooling and noise that had Zen 2 was Eluktronics RP-15 or RP-17 (and it was reasonably priced too - but the GPU wasn't that interesting considering that my Vega 56 is easily on par with RTX 2060).

Dell on the other hand managed to properly mess up the G5 15 SE in regards to thermals at launch (despite the efficiency of both Zen 2 and RX 5600m), but I think a subsequent BIOS update managed to bring those issues under control for the most part) , and I'm not encouraged by the premise that they will seemingly be allowing future laptop CPU's to reach 100 [or more] degrees Celsius while perceiving those as 'standard operational temperatures').

 

By what means did that Dell BIOS update bring thermals under control, one might ask? Oh yes, it was probably the good old Dell Furious Throttling (tm) technology.

 

Lol.
Here are the test results with latest BIOS and driver updates:


On average, GPU performance dropped around 3-5%, but it mostly had higher minimums. Dell apparently prevented the GPU from clocking as high as it did before, but to be fair that just shows their cooling design was not adequate for this hw to begin with.

 

Which is why it would annoy me if that hypothetical became practice. I didn't claim cpu had a direct correlation to cooling in laptops and will not defend a point I did not make.

Because if you can get proper cooling of a lower sku then I'll get that if there is no tangible benefits between the two? Not really sure about the rest of your post since I wasn't making an argument, merely stating that specific hypothetical would be annoying. That is all.

 

Have you considered that maybe Zen2 is hard to cool, especially in non chunky laptops, due to the thermal density. I mean, the R5 3600 gets to 90C in desktops with the stock Wraith Stealth cooler. The GPU with its bigger die is easier to cool.

 

The die size on the 4800H and 5600M are definitely small - which did not help things with the G5 SE as any issue with heatsink / paste / pad quality control results in very high temperatures. It has some very hard to cool chips paired with a cheap heatsink solution. Frank Azor even bragged in the PCWorld interview about the G5 SE that they were able to lower the size/cost of the heatsink because of Smartshift. Not a good recipe.

 

The marketing departments at AMD and Nvidia were able to able to coin catchy names like SmartShift and Dynamic Throttle Boost 2.0 to cover up the fact that ODMs are reducing R&D and manufacturing cost on cooling. Genius.

 

Any laptop will reach 90C+ temps when stress testing, The Throttle zone is normally around the mid 90C mark anyways. The problem with the Dell G5 SE is that it was reaching temps over 90C+ when gaming. On my Legion 5 the 4800H barely reaches 80C when gaming and doing video editing, I would say the same for the HP Omen 15, and other laptops i cant think of

Yeah If Dell was smart they could easily make a good cooling system for the Dell G5 SE. It has the best price/performance ratio when it was first released. What a missed opportunity there especially if you wanna go full AMD .

 

My PH517-61 doesn't exceed 75 degrees Celsius on the CPU (Ryzen 2700 - 65W TD) and 65 degrees Celsius on the GPU (Vega 56 limited to 120W) when both are fully stressed at the same time while remaining fairly quiet... you can easily game without headphones with maxed out CPU and GPU without nasty noise discomfort.
Then again, I DO have a 17" thick chassis whose cooling was DESIGNED specifically for AMD hw ... however, it does show that laptop manufacturers for the most part are playing fast and loose with laptop cooling in general (and I suspect they don't really spend the time to TAILOR the cooling to the hw that goes inside).

A 15" unit with lower wattage (say 45W TDP on CPU and same 120W on GPU) should be designed with at least those temperatures under maxed out load for both CPU and GPU (maybe also allow the GPU to reach 75 degrees Celsisus) at the same time.

The noise on newest laptops is really not exactly something my ears could manage from what I saw in the reviews. I had a GL702ZC from Asus before, and that things thermals were fine, however, the noise under maxed out load was ridiculous - like a jet engine (resulting in consistent ringing in my ears), plus that thing had a penchant for burning itself out after about a month of use... but that only went to show how BAD Asus designed cooling was and I told myself I won't be going back to them for newest laptops until they demonstrate a consistent and acceptable upgrade in cooling quality).

This is one of the reasons I'm skeptical of 90+ degrees C on laptops during gaming (yes, under gaming), never-mind heavy loads in general - because the operational lifespan of the hw will be reduced under those temperatures (most probably) and will inevitably result in high system noise (some of us cannot stand that - and if we need to work on those machines in content creation [which can stress the CPU and GPU simultaneously to the max] in a classroom, you can forget about using it then and there as it would be a source of constant distraction - fortunately, I don't have that problem with my PH517-61 as the cooling is exceptional, but I'd but rather not risk it with new laptops.

Dell did execute the cooling badly on that design and only now managed to release a BIOS update which brought the thermals under control (they had to slightly nerf the GPU clocks to do that though).

I realize that 7nm process CPU designs will be more tightly packed and produce more heat, but to be realistic, 65W (12nm) vs 45W or 35W on 7nm shouldn't be THAT much of a problem.
And if it is, then manufacturers NEED to step up their game in cooling (but I'm sure they will come out with yet another stupid excuse for this).

 

I can see ODM's reducting R&D and manufacturing cost on cooling (as this has been fairly consistent behavior over a long period of time), but do we have evidence to support the idea both AMD and NV coined SmartShift and Dynamic Boost 2 to 'cover up' anything for ODM's?

 

The G5 SE would be at 85C running the in-store demo program.

 

And my 15" DTR can cool 300W of combined load heat on max fans, or 200W with very quiet fan speed. Companies make what sells, and consumers want to have their cake and eat it too--a svelte form factor and "good enough" performance.

 

Not excusing Dell for physical cooling implementation, but given the current situation on the laptop since it’s already out in the wild, even slight tweaks to clocks would make sense as a solution. Reducing max clocks faster or boost time even a little can at times make a huge difference with, at times an actual gain in performance due to lower chance of throttling harder and by preventing massive intermittent deltas in clocks.

I had an Asus X550 where the GPU would power limit so fast clocks would drop like a rock like every 30 seconds to a minute making gaming stuttery. I had to mod the laptop bios by disabling boost clocks (I think only a difference of 100 MHz ) on the 840M and low and behold no throttling and I had consistently higher frame rates than before. Before that I had used the likes of nvidia inspector but it wasn’t a perfect solution and recovering from sleep could reset clocks. Ended up posting my permanent solution on nvidia forums and informed Asus about either reducing boost clocks or increasing power limits as the laptop ran relatively cool overall, Asus ignored the issue of course.


Basically what I’m alluding to is that even if vendors put highest end CPU/GPU beyond chassis cooling/power capability just so they don’t feel left out by other vendors, they can at least tweak the damn clocks or boost times or have earlier throttling to slightly lower levels so performance is more consistent and you won’t have massive intermittent or constant performance drops under long load times. Some vendors keep default profiles and CPU/GPU sustain high clocks, but they also let surface temps run out of control like 60-70C at the surfaces which is also kinda extreme. Another aspect in which vendors really need to improve.

 

The cheaper cooling system is just part of the lower price. It is the same cooling unit used on the G3 and G5 Intel+Nvidia models. Maybe if Dell used the G7 design which does seem to have better cooling, but we definitely do not want them to use the Alienware m15 design because that means higher price and likely soldered ram.

The bad part is for an all AMD laptop you cannot easily pick something with a better cooling system and build quality because there are just so few options to choose from. The only better option is the MSI Alpha refresh, which does have a better cooling system at a higher cost. Technically the display has worse color gamut too, but replacing a display is much easier than trying to fix a poor cooling system.

 

How much extra does a vapor chamber cost in bulk? $50? And it doesn't have to entail soldered RAM. Even Razer provide SODIMMs. It's just greedy Dell trying to make a few more bucks on locking people into their value RAM.

 

The vapor chamber would not help in the current G5 chassis, the fans / vents cannot keep up with the amount of heat generated stock.

Who knows if Dell will even make a Ryzen 5000 refresh of the G5 SE, but if it ends up being an "Alienware m15 SE" I am guessing it will have soldered ram just like the Intel models.

 

Hopefully customers will relentlessly punish any attempts to solder RAM. Dell must have lost some m15 sales on this already. Also, they don't solder RAM in other models such as the XPS or G-series. It would appear to be an Alienware m15/m17-specific stupidity.

 

AMD always had a problem with laptop manufacturers cutting corners on not just cooling, but also other internal hw (the fiasco with low quality screens - which is actually happening still more or less), usage of 1 RAM stick (no dual-channel usage - low quality and low speed ram, soldered RAM stick - still happening in some cases), no use of an SSD, weaker Wifi, etc.

Things ARE improving, but more and more consumers are becoming acutely aware that manufacturers (who are being increasingly called out on these things by tech review websites) tend to skimp on quality when AMD hw is used just to save a few pennies (even though moving to higher quality/better/faster hw would literally not increase the overall price by much... or even at all in most cases).

 

Dell isn't the only offender.
I've seen Zen 2 offerings from Asus (and other manufacturers) last year that had come with pre-soldered sticks of RAM (8GB for example - in 2020 no less - you'd think 16GB or even 32GB would have become a standard by now like it should have).

 

Looks like there are some benchmarks finally:


Shows that the 5800H is quite the improvement over the already speedy 4800H. In the Puget Systems Adobe photoshop test scores went from 690 to 974, handily beating the mini tigerlake so let's see what Intel can do with their H series. Oh and in Excel it's 46% faster. SO much for Intel's RUGS that they touted so much about even though they don't believe in benchmarks.

A quick summary screenshot from the application tests on Jarrod's Tech:


Look forward to some more competition in the mobile market, Intel needs to get off their butt.

 

Ruh-roh.

 

Yeah, they do take turns getting off their butts. AMD was on extended leave for a decade. And after 4 'amazing' years, is finally making a real entrance back to the mobile space. A solid win for AMD, indeed. For now. I'll wait for Intel's answer too.

But instead of navel-gazing (AMD to AMD), we can look at what is available from Intel already, and it is nowhere near that big of a jump (and in some cases, Intel is still leading in some respects too).

Intel's RUGS are what we must thank for these kinds of improvements. The benchmarks leave one ever so unsatisfied. You know, that 'margin of error stuff, realistically' and all. Especially as it has no bearing on real-world workloads/workflows.

 

Surprised AMD haven't come up with their own eGPU solution, given the lack of TB. Even some OEMs managed to come up with eGPU connectors (and better than TB3) on their own.

 

Where is Intel currently leading? And no it's not Intel relabeling benchmarks to RUGS that is responsible for AMD's gains, but good try.

*on edit*
And just to be clear, RUGS is Intel hand picking benchmarks and saying "oh these aren't benchmarks, they are so much more than that."

Yeah I guess they just don't see the market. Having said that, Asus now has their own PCIE connector on some of their laptops. That's >TB IMHO...

 

You can see those goal posts moving from space.

 

Ha, the ROG Flow X13. Well, the connector is great, 8 lanes, but the eGPU itself has an integrated mobile GPU. Hope they build a larger enclosure and add support for this to their other notebooks.

 

A "proper" enclosure entirely defeats the purpose of the device. Which would you rather transport - the flow with its eGPU or a Razer Blade Stealth with a Core X?

 

@custom90gt, I won't do that homework for you. Even if I held that information under your nose and you could smell the ink, you won't admit Intel leads in some ways, still.

Stop trying to redefine what has been defined already. Intel RUGS are "Representative Usage Guides". The benchmark that matters is time. Not some made-up 'score' that shows real-world squat.

But all right. Here is a bonus for you. A benchmark no less.

So much for AMD dominance. Even at lower power too. Wow, all the stereotypes are going out the window... lol...



@saturnotaku, yeah, I like moving goalposts. Sitting still gets boring fast.

An AMD powered notebook is still the budget option today (overall in feel/usability, not price). Even if the engine under the hood is premium for some workloads.

See:
Asus ROG Flow X13 Review: XG Mobile Meets Ryzen 9 5980HS | Tom's Hardware


I'm so glad about the competition. AMD took its time to wake up. Let's see how long it will stay awake this time.

 

I would not transport either eGPU almost ever (that's kind of against the idea of an eGPU), and both laptops looks equally transportable to me.

 

"Intel RUGs or "Representative Usage Guides" are pre-scripted and packaged benchmark test suites that Intel prepares for the press and analyst community"

Surely this is the reason AMD made Zen 3, it's not to steal Intel's business.

Also thanks for picking one of the ~5 tests that Intel was ahead in and posting it here. Should we post the other ~35 where AMD is ahead?

 

See? I told you he would do it.

 

Show what the definition of RUGS is as per Intel? Does that bother you?

 

No, it doesn't. But that's not complete (as you know).

See:
CES 2021 - 1 - ID:615781 | Performance Index (intel.com)

 

Whats the point of RUGS when they also state alongside it verbatim "Industry benchmarks" in the above link?

Genuinely asking