this generation’s laptop chips will make for some very fun comparisons. Can’t wait to see 4000U and Tiger Lake head to head, both seem to have advantages that will help over the other.
I expect it's going to compete very well with Tiger Lake, assuming Zen 5000 series APUs aren't out beforehand.
I cannot wait to get my hands on a decent laptop with this chip. The question is, how long will I have to wait? I'd like a thin and light laptop, but maybe with some form of discrete graphics, a decent, high resolution screen and a decent keyboard for when I don't bring one.
If Tiger Lake is released in Q4 2020 then it will be released about a quarter before the Zen 5000 APU series, which should be released in Q1 2021 (a year after the 4000 series). I strongly doubt Intel will manage a Tiger Lake release in Q3 of this year. That's too early. It's more likely it will be pushed back to early 2021 if anything.
Tiger Lake's 96 EU iGPU variant (only that, which will be rare) should clearly be faster than the Vega iGPUs of the 4000 series, but I doubt it will be able to compete with the Navi iGPUs of the 5000 series. And, of course, Tiger Lake will also be capped to 4 cores. So at best Tiger Lake will have only an iGPU edge for a quarter or so over the 4000 APU series (just with their 96 EU iGPU variant) while still offering subpar CPU performance. At worst, if Intel either delays its launch or AMD release their 5000 series of APUs a bit earlier it won't even have that.
To be fair, CPU performance from both vendors is strong enough at this point that the average consumer should probably put more weight on the GPU performance. That said, while we may speculate that Intel's top Tiger Lake will have the iGPU edge, it is hard to say where rest of the lineup from both companies will land especially when cost is considered. I'll decline to comment on release dates, but I expect some good competition when they do release.
It depends what you're doing with it. I'm looking forward to the day when I can get a regular, non-DTR laptop with a 16-core CPU in it - and the way that AMD is making progress, that day doesn't seem too far off.
The majority of the laptops are based on Comet Lake, not Ice Lake. And I doubt you will find a $300 laptop this year, not even with Comet Lake. Pulling arbitrarily low dollar values out of your behinds is not a valid argument. Ice Lake based laptops are much fewer (because Intel have only a couple of 10nm fabs and their yields are rather poor), are quite more expensive and are largely intended for Project Athena certified devices.
What Project Athena boils down to is "higher prices". It is effectively a market segmentation strategy to differentiate between "premium" and "non premium" laptops with just a two word marketing term instead of a list of specs. It is also a tidy way for Intel to isolate Ice Lake laptops from the "common" laptops with Comet Lake, since they don't have enough of the former anyway to fulfill the demand of the market. In short, I doubt Project Athena would have been devised if Intel had been able to manufacture Ice Lake at a yield that would satisfy them and at the volume required to meet demand on its own. That is not the case though, and in 3 to 4 quarters we will have a repeat of more or less the same (though not quite) with Tiger Lake + Rocket Lake...
He! AMD has one a masterpiece. We can not compare well a 4 core with a 8 core. They gave up on four cores with an higher core count.....low clocked by default outside a spare single core turbo. Someone can see the real issue with Ryzen 3 4300U.....4 cores, the clock speeds are absolutely bad to compete with Intel. More or less Renoir is not AMD answer to Ice. It is another thing for another class of users. They try a new street to do some dent in the market. There isn't much to compare unfortunately.
Based on these there is a distinct possibility that the all-core turbo/boost of the two chips are going to be similar. AMD's 'boost' isn't a single-core turbo - its the maximum possible frequency of a single core based on electrical and thermal headroom.
Intel's all-core frequency is nowhere near the base frequency. Usually it's much higher. If you bash the Intel chip with AVX-512 and at the same time use the GPU, you will get the base frequency. But that's not a fair comparison since the AMD chip doesn't even have AVX-512.
It's totally fair to compare dissimilar products as long as they're in the same market. Comparing a $25k "crossover" SUV to a $500k Ferrari isn't fair, but if two companies sell ~$25k crossovers, it's totally fair to compare them, even (especially, actually) if one has a ton of features the other lacks.
Having said that, I'm not sure AVX-512 falls into that category. Sure, for some workloads - and therefore, some users - it's a huge deal, but for many more, it's not a significant driver of performance.
I'd bet that for most users, doubling the cores will have a more significant impact on overall performance, but for others, max (sustained) frequency for single-core is probably most relevant.
I don't understand why all you make such a huge deal out AVX-512 from my post. All I'm saying is that all-core frequency is higher than base frequency. You only hit base frequency if AVX-512 and/or GPU are used. Just for the CPU, base frequency isn't required to stay under 15W. Previous posts were comparing Intel base frequency to AMD base frequency, which isn't fair. That's all.
ahh but it is.. if you look at the desktop cpus... most clam intel is the better cpu because it clocks higher.. but like always... now that amd is the higher clocked cpu on mobile.. its unfair ??
it always seems.. if intel does it.. its ok.. but if amd does the same... its not ok.. look at power usage over the years... to now claim... its unfair.. is just an excuse from the intel fans to try to make intel look better...
oh.. and i take it you think you can ?? come on.. sounds to me like you are one of those " its ok if intel does it, but if amd does it.. then its unfair " type of people..
further.. you also didnt fully explain why you think its unfair.... how is comparing base frequency to base frequency unfair ?? like i said.. on desktop.. the comparison has been intels higher clocks to amds lower clocks..is that fair ?? intel cant really get much higher clocks on these chips..
Well, I think the average frequency under load is far more important than either base or max frequency.
Not saying this is the case for any specific CPU, but if one CPU were to say, average halfway between min and max, while the other normally stays above 70% of max boost, there's a real difference there even if on paper they appear to share the same frequency range.
Korgus, I agree with your sentiment but the thing you're doing with the periods (...) is called an ellipses https://www.merriam-webster.com/dictionary/ellipsi... And it's meant to denote intentionally missing text. I think you're doing it do indicate a dramatic pause? You don't have to do that because your point is valid and it has a negative effect on the appearance of your comment to the reader.
Because you'll never see the base frequencies in operation of either? They'll jump around from between their dormant state speeds all the way up to max turbo until you reach thermal saturation, then it'll jump around even more. Sure, I could throw an ice-water liquid cooling loop on either, put them on mains power and tell them to TDP-up well past their 15W thermal spec and throw some massive server or content creation looped workload on them, but that's not what they were meant for. The base clock speed is a lie, for both camps.
From what I've seen, Intel U mobile CPUs seem to have a base clock frequency of 800mhz no matter what the documentation states. It jumps around from there to max turbo speeds. ;p
And this is the top 15W Ice Lake model vs. the bottom end Renoir...and the 8C 16T 4800 U still has a higher base clock (1.8 Ghz) and a higher turbo...
Of course, we'll still have to wait for the first reviews of Renoir in a - hopefully - properly built and configured laptop to see how it does in real life (clocks, power consumption in idle, light load and heavy load).
man gondalf.. that post makes NO SENSE what so ever.. were you drunk?? high on drugs ?? upset because you beloved intel isnt the leader on desktop.. and may not have an easy win on mobile now ?? your post also is nothing more then an intel shills personal opinion.. with no facts or proof... give it up already...
AMD does not even really need to compete toe to toe with Intel, they just needs to get really close, say 95% of the way. AMD's strategy is to get you almost everything Intel has to offer at 25-50% less at retail. When your talking a gain of 5% to maybe 18% tops better performance with an Intel, pending on chip class, you're paying an additional $75-$150 for it. Add to that the fact AMD it's finally producing similar power consumption, well you can see why AMD shares have been soaring over the last 5 years.
However, the edge that Intel once enjoyed for over a decade has quickly shrunk to a single gate. Give AMD'S advancement in the last five years, Intel has got to be more than a little nervous about future market share.
Well, put another way, AMD's finally making the same proposition with mobile that it did with desktop Zen's launch (only now they're catching up with IPC): For $X, we'll come close on single-threaded performance but we'll give you a lot more cores for multi-thread.
It wouldn't surprise me to see say, Ryzen 7's offered for the same price as an i5, and Ryzen 5's for the same price as an i3.
I'm particularly interested in battery life of these chips, since AMD's finally packing LPDDR4X support, negating a platform advantage Intel's had. (Gut feeling is that AMD will be doing better than before, but Intel still pulls ahead with absolute numbers.)
Ian, I feel the need to bring up that the Zen 2 cores technically aren't 'full blooded' as they have less cache. They still should pack quite a punch though, and I look forward to a review.
Agreed. AMD made something of that 'gamecache'. One would assume that dropping it to a quarter of the size would have a significant effect on 'gamability'.
On the other hand, a monolithic die could have lower RAM latency, which might help compensate for this.
I'll be looking forward to some deep investigations into this when the laptops hit the market.
Just L3 though, AFAIK. 4MB per CCX, so 4+4MB in 4800U or 1MB per core.
Saves quite a bit of die (and power) from 16MB per CCX, but savings was, of course, reapportioned to iGPU.
Though this is AMD's first Zen 2 design with 7nm integrated SoC silicon (DDR4/LPDDR4 PHYs and controllers included). I really want to see the die shot.
So Renoir is all on one piece of silicon? Assume 75 mm2 for the Zen cores, that means the remaining ~75 mm2 contains not only the graphic cores but also the I/O that is on a separate piece of silicon in the Ryzen 3000 CPUs. And that I/O presumably gets a shrink so that it is all done with 7nm.
Judging by your own measurements, the Zen2 core on the APU should be about 50mm.
You stated at release that the L3 cache was 50% of die size of the 75mm. The other 50% was the Zen2 cores. So 37.5mm die, 37.5mm L3. That was for 32MB, so at 8MB that's 1/4, or a little less than 10mm. So about 47.5mm.
It should be roughly 50mm for cores, the rest shared in half between CUs and everything else. That is because in 75mm^2 Matisse Zen2 chiplet, ~40%(~30mm^2) is L3, but in Renoir- there is only a quarter of that cache. This could save >20mm^2, and some space on Zen2 chiplet are IF links, not cores.
I'd estimate about 50 mm2 for the Zen cores with the smaller cache. I'd give the Vega cores half of that, but with the codecs and display engine possibly another 40-50 mm2 together. The rest going mainly to internal and external I/O.
Would be interesting to see a die image when the chip arrives.
I was half expecting this generation of mobile chip to start following AMD's chiplet strategy via leveraging the same Zen 2 chiplets but with a IO die that had an integrated GPU. In fact, I would have thought that the IO die was also going to pull double duty as a lowend discrete GPU die when given a higher power budget.
Keeping it monolithic has benefits for the IF and also idle latency, which is key for a mobile CPU. I don't expect AMD to go chiplet with graphics on this sort of product any time soon. Maybe in something higher up the thermal stack, or for specific customers, perhaps.
What's interesting is that AMD also launched their 1st mobile-focused dGPUs in ages at the same time. My bet is AMD will keep the mobile CPU's monolithic but might possibly have release a version to specific customers with either the iGPU part disabled, or cut down to 1 active CU (for battery life purposes when mostly idle) for use with dGPU designs. Depending on the chip layout, the dead silicon from the other CUs might let the chip run with more power longer before throttling.
Honestly i think we well'never see much in the 15W of this SKU, bet 25 W will be the standard (70% more power headroom). 8 cores eat a lot of power and running then slow do not help much the customer and makes your Laptop not much snappy/delusional.
you are the one that is delusional gondalf.. looks like your thinking is.. if intel can do it.. no one else can too.. maybe thats when intel can only do 4 cores with what ever lake they are making for mobile now.. start posting proof of your opinions or go away..
Some of the best, non-combative comment I've seen in a long time here. Everyone is excited about 2020 laptops due with these chips. Very happy that AMD is back in the game - benefits us all.
>I had stated in our article that I estimated 150 mm2 for the die size. Turns out, I wasn’t too far wrong.
Once you have eaten that many wafers, you can literally smell the die size XD
I was wrong in that AMD wanted to have GPGPU uArch in the iGPU, since Dr Su already stated there will RDNA in Ryzen Mobile in the future. But even with the current Ryzen Mobile, it doesn't look like Tiger Lake is attractive. It will be Quad CPU with AVX512, so it may be a better CPU but at half the Core, and Xe GPU I would bet it will be at best on par with AMD.
And then early next year you get AMD may be with Zen 2+ and RDNA2 ( Which is forms the basic of PS5 and Xbox X ) on TSMC N6.
But generally Speaking, despite everything that has happened over at Intel, and AMD being technically better in Desktop, Server and now Laptop, AMD is still not getting the market shares. I am disappointed in their sales, marketing, and forecasting department. Server is still a dismal 4%. Desktop CPU still under 20%. AMD needs to be a lot more aggressive especially in the Server Department which is where the big bucks are.
Isn't AVX512 use relatively limited for consumers? Besides, AMD would easily retain the advantage with its increased core count and if software devs actually optimised their codes to take advantage of Zen uArch as it should.
That said, Zen2+ doesn't exist. Zen 3 exists which will be a completely new architecture and produced on 7nm+ EUV node.
Also, Zen 3 should be released by the end of THIS year (2020) ... not early 2021. 2021 (most likely end of 2021) is slated for Zen 4 on 5nm (which seems to be on track because the yields have gone over 50% already).
Little use to most laptop consumers. Plus AVX512 is such a power hungry feature that it causes severe throttling to keep temperatures in control.
IC asked Mark Papermaster if a Zen 2+ was planned in the future and Mark responded:
"MP: I have nothing to say on any refresh on current designs, but we always look at where it makes sense and where we’re able to take opportunities to provide a bump in performance, power, or die area."
So interpret that however you will. I don't see AMD squeezing in a Zen 2+ between Zen 2 and Zen 3, considering AMD has been pretty consistent with releases non their roadmap. Maybe Z2+ will live on in lower tier CPU's just as Z1/Z1+ has for the last year.
Very curious about the memory subsystem the Threadripper 3990X uses DMA to communicate to the GPU, and the 4000 chip look as if it does too, what's happening within the APU (oh it's not an APU anymore).
None of the Threadrippers have an on-die GPU, the 4000-series APU does. The more important question you should be asking is how does the 2000-series and 3000-series APUs communicate with their on-die GPUs?
with 90% yield, most outcome will be the 16 thread 4800u. this is great news for laptop consumers. i just hope it has enough pci lanes to accommodate x4 thunderbolt port.
Well, 4 cores takes up about 25mm2 so 8c Ice Lake would be ~145mm2 they seem pretty on par actually -- depends on how much space the GPU takes on each but I wouldn't say it's clearly inferior.
"This image is not to scale." Well, color me disappointed. Thought we were going back to the days of REAL COMPUTERS that filled up an entire filing cabinet.
Would be nice to see a desktop APU (R5-4600G?) equivalent with 6c/12t, and enough GPU power to match a GTX1050 or so at gaming.... (IF folks thought the 2200G/2400G sold well, the above should sell by the proverbial freighter-full!
" And Renoir is barely better than current APU in the GPU side " ahhh so you have bench marks that show this ?? post a link so the rest of us can see then too
Namisecond, maroon1 wont post any links, cause there arent any, yet. this is their own personal opinion, with some personal bias thrown in, nothing more....
OP said "Would be nice to see..." - AKA "I'd like it if AMD did this...", not "AMD have done this".
maroon1, on the other hand, made a performance claim that they can't back up. So maroon1 needs to justify that with evidence, whereas neither OP nor korguz have any such burden of proof.
Seems to me that there currently is a point of diminishing returns with the AMD APUs (sorry ... I'm going to keep calling it that, Dr Su !).
Certainly on mobile, 8 CUs (512 shaders) seems to be the sweet spot. 10/11 CUs raises the performance bar, but not 1:1, and especially on the DT. Maybe DDR5 and Navi/RDNA-II can change the formula a bit, but a chiplet w/eDRAM (as a 4th level cache?) would be the logical long-term path ...
You may be mistaking popularity by price for popularity by performance. I'd argue that 2200G/2400G sold well because they sold for a significantly lower price than their larger core count, GPU-less brethren. If the new desktop APUs sell at a similar price point, it should do well. Don't expect significantly better GPU performance from the 4000-series APUs over their older counterparts. Even 20% may be expecting too much.
Huge chip areas at lower TDP = excellent cooling. Fantastic. Some bold manufacturers may even tinker with PASSIVE cooling?
I mean, we are back on the 486 era, where chips only needed a heatsink. That's a full circle. I can see bold designs, with maybe just passive watercooling, or just a heatpipe plugging the thing into the aluminum case of the notebook, - turning it into a decent heatspreader - being feasible, without Apple levels of sacrifice to performance or durability.
IIRC, 486 chips were right around 10W and I definitely recall my 486/33 system came with a fan for the heatsink. Even now, 10W TDP seems to be the tipping point where vendors consider active cooling for their products. My EDC laptop with a 6W TDP CPU makes do with a heat spreader plate, a big one. The 10W TDP desktop version of that same chip came with a fairly elaborate heatsink, sans fan. Anything 15W TDP and up will most likely come with active cooling or a very elaborate custom passive cooling setup.
They just need to start installing CPUs (and its LP memory, and M2 slots) into the screen half of the laptop, with ribbed aluminum as its cover. Much more cooling area where there is no keyboard, and even 2mm ribs with 2mm pitch increase the area by 3x again, easily.
The table is wrong. Ice Lake is 10+. The Canon Lake that was released in 2017 is the original 10nm. No idea if there is an official word, but wiki says Tiger Lake is 10++ and I think that is probably correct. I think it's also safe to assume that the slides you showed 2 years ago from Intel that had 10++ with a performance advantage over 10+ are now incorrect as Intel has probably changed direction in favor of manufacturability and lower power over raw performance
It would be really cool if you could post an article comparing the die sizes of mobile SOCs over the past decade especially the A series vs the snapdragons and then the snapdragons for windows vs x86 processors for the same!
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
103 Comments
Back to Article
skavi - Tuesday, January 14, 2020 - link
this generation’s laptop chips will make for some very fun comparisons. Can’t wait to see 4000U and Tiger Lake head to head, both seem to have advantages that will help over the other.Cliff34 - Tuesday, January 14, 2020 - link
Me too.. This year is an exciting year for laptops.eek2121 - Tuesday, January 14, 2020 - link
I expect it's going to compete very well with Tiger Lake, assuming Zen 5000 series APUs aren't out beforehand.I cannot wait to get my hands on a decent laptop with this chip. The question is, how long will I have to wait? I'd like a thin and light laptop, but maybe with some form of discrete graphics, a decent, high resolution screen and a decent keyboard for when I don't bring one.
Santoval - Wednesday, January 15, 2020 - link
If Tiger Lake is released in Q4 2020 then it will be released about a quarter before the Zen 5000 APU series, which should be released in Q1 2021 (a year after the 4000 series). I strongly doubt Intel will manage a Tiger Lake release in Q3 of this year. That's too early. It's more likely it will be pushed back to early 2021 if anything.Tiger Lake's 96 EU iGPU variant (only that, which will be rare) should clearly be faster than the Vega iGPUs of the 4000 series, but I doubt it will be able to compete with the Navi iGPUs of the 5000 series. And, of course, Tiger Lake will also be capped to 4 cores. So at best Tiger Lake will have only an iGPU edge for a quarter or so over the 4000 APU series (just with their 96 EU iGPU variant) while still offering subpar CPU performance. At worst, if Intel either delays its launch or AMD release their 5000 series of APUs a bit earlier it won't even have that.
BurntMyBacon - Wednesday, January 15, 2020 - link
To be fair, CPU performance from both vendors is strong enough at this point that the average consumer should probably put more weight on the GPU performance. That said, while we may speculate that Intel's top Tiger Lake will have the iGPU edge, it is hard to say where rest of the lineup from both companies will land especially when cost is considered. I'll decline to comment on release dates, but I expect some good competition when they do release.twtech - Sunday, April 12, 2020 - link
It depends what you're doing with it. I'm looking forward to the day when I can get a regular, non-DTR laptop with a 16-core CPU in it - and the way that AMD is making progress, that day doesn't seem too far off.AdhesiveTeflon - Tuesday, January 14, 2020 - link
It's a great thing. Whoever wins, us consumers will benefit.zamroni - Tuesday, January 14, 2020 - link
forget tiger lake for now. we haven't seen ice lake laptop yetextide - Tuesday, January 14, 2020 - link
Uhh, yeah we have...?eva02langley - Tuesday, January 14, 2020 - link
Ah, cmon, Ice Lake is as of now an exotic product. It is far from being common.Farfolomew - Tuesday, January 14, 2020 - link
What are you talking about? There's a ton of laptops out there with 10xxGx series parts. Even cheap $300 budget laptops. How is Ice Lake exotic?shing3232 - Tuesday, January 14, 2020 - link
those are comet lake I suppose.kenansadhu - Tuesday, January 14, 2020 - link
If it's 10xxGx as what Farfolomew said, it should be Ice Lake, I believe. Comet Lake's series name is 10xxxUSantoval - Wednesday, January 15, 2020 - link
Not even Comet Lake. $300 laptops are based on crappy Celeron SoCs.Santoval - Wednesday, January 15, 2020 - link
The majority of the laptops are based on Comet Lake, not Ice Lake. And I doubt you will find a $300 laptop this year, not even with Comet Lake. Pulling arbitrarily low dollar values out of your behinds is not a valid argument. Ice Lake based laptops are much fewer (because Intel have only a couple of 10nm fabs and their yields are rather poor), are quite more expensive and are largely intended for Project Athena certified devices.What Project Athena boils down to is "higher prices". It is effectively a market segmentation strategy to differentiate between "premium" and "non premium" laptops with just a two word marketing term instead of a list of specs. It is also a tidy way for Intel to isolate Ice Lake laptops from the "common" laptops with Comet Lake, since they don't have enough of the former anyway to fulfill the demand of the market. In short, I doubt Project Athena would have been devised if Intel had been able to manufacture Ice Lake at a yield that would satisfy them and at the volume required to meet demand on its own. That is not the case though, and in 3 to 4 quarters we will have a repeat of more or less the same (though not quite) with Tiger Lake + Rocket Lake...
regsEx - Tuesday, January 14, 2020 - link
Looking at two largest retailers here, there are hundreds of notebook models on ICL.Korguz - Friday, January 17, 2020 - link
hundreds??? yea rightlevizx - Tuesday, January 14, 2020 - link
Typing on one of those you haven't seen.tsk2k - Tuesday, January 14, 2020 - link
Please compare Ice-lake and Renoir as soon as possible In, my body is ready.Gondalf - Tuesday, January 14, 2020 - link
He! AMD has one a masterpiece. We can not compare well a 4 core with a 8 core.They gave up on four cores with an higher core count.....low clocked by default outside a spare single core turbo. Someone can see the real issue with Ryzen 3 4300U.....4 cores, the clock speeds are absolutely bad to compete with Intel.
More or less Renoir is not AMD answer to Ice. It is another thing for another class of users. They try a new street to do some dent in the market.
There isn't much to compare unfortunately.
Tamz_msc - Tuesday, January 14, 2020 - link
i7 1056G7 - 1.3GHz base/3.9 GHz TurboR3 4300U - 2.7GHz base/3.7 GHz Boost
Based on these there is a distinct possibility that the all-core turbo/boost of the two chips are going to be similar. AMD's 'boost' isn't a single-core turbo - its the maximum possible frequency of a single core based on electrical and thermal headroom.
Alistair - Tuesday, January 14, 2020 - link
yeah AMD has double the clock rate vs. Intel sustained at 15w and he says "clock speeds are absolutely bad to compete with Intel"... smh...some people are just dumb (writermies is that you?)
milli - Tuesday, January 14, 2020 - link
Intel's all-core frequency is nowhere near the base frequency. Usually it's much higher.If you bash the Intel chip with AVX-512 and at the same time use the GPU, you will get the base frequency. But that's not a fair comparison since the AMD chip doesn't even have AVX-512.
iranterres - Tuesday, January 14, 2020 - link
AVX-512 does not seem to be relevant just yet, and despite AMD not clocking higher, it's better on clock-for-clock then Intel chips at the moment.Alistair - Wednesday, January 15, 2020 - link
all-core frequency isn't 15 watts, it is higher than 15 watts, base frequency is the one guaranteed at 15 wattssing_electric - Wednesday, January 15, 2020 - link
It's totally fair to compare dissimilar products as long as they're in the same market. Comparing a $25k "crossover" SUV to a $500k Ferrari isn't fair, but if two companies sell ~$25k crossovers, it's totally fair to compare them, even (especially, actually) if one has a ton of features the other lacks.Having said that, I'm not sure AVX-512 falls into that category. Sure, for some workloads - and therefore, some users - it's a huge deal, but for many more, it's not a significant driver of performance.
I'd bet that for most users, doubling the cores will have a more significant impact on overall performance, but for others, max (sustained) frequency for single-core is probably most relevant.
milli - Friday, January 17, 2020 - link
I don't understand why all you make such a huge deal out AVX-512 from my post. All I'm saying is that all-core frequency is higher than base frequency. You only hit base frequency if AVX-512 and/or GPU are used. Just for the CPU, base frequency isn't required to stay under 15W.Previous posts were comparing Intel base frequency to AMD base frequency, which isn't fair. That's all.
Korguz - Friday, January 17, 2020 - link
" Previous posts were comparing Intel base frequency to AMD base frequency, which isn't fair. " how is it not fair ??milli - Sunday, January 19, 2020 - link
Because just like Intel's and AMD's TDP are not fully comparable, the base frequencies aren't either.Korguz - Sunday, January 19, 2020 - link
ahh but it is.. if you look at the desktop cpus... most clam intel is the better cpu because it clocks higher.. but like always... now that amd is the higher clocked cpu on mobile.. its unfair ??it always seems.. if intel does it.. its ok.. but if amd does the same... its not ok.. look at power usage over the years... to now claim... its unfair.. is just an excuse from the intel fans to try to make intel look better...
milli - Tuesday, January 21, 2020 - link
You're not capable of reading comprehensively.Korguz - Tuesday, January 21, 2020 - link
oh.. and i take it you think you can ?? come on.. sounds to me like you are one of those " its ok if intel does it, but if amd does it.. then its unfair " type of people..Korguz - Tuesday, January 21, 2020 - link
further.. you also didnt fully explain why you think its unfair.... how is comparing base frequency to base frequency unfair ?? like i said.. on desktop.. the comparison has been intels higher clocks to amds lower clocks..is that fair ?? intel cant really get much higher clocks on these chips..Korguz - Saturday, February 15, 2020 - link
milli, yea thought so.. no other way to prove your point...twtech - Sunday, April 12, 2020 - link
Well, I think the average frequency under load is far more important than either base or max frequency.Not saying this is the case for any specific CPU, but if one CPU were to say, average halfway between min and max, while the other normally stays above 70% of max boost, there's a real difference there even if on paper they appear to share the same frequency range.
Namisecond - Wednesday, January 22, 2020 - link
But these aren't desktop CPUs with just 200mhz turbo boosts. These are mobile chips with huge turbo ranges.Thud2 - Sunday, February 9, 2020 - link
Korgus, I agree with your sentiment but the thing you're doing with the periods (...) is called an ellipses https://www.merriam-webster.com/dictionary/ellipsi... And it's meant to denote intentionally missing text. I think you're doing it do indicate a dramatic pause? You don't have to do that because your point is valid and it has a negative effect on the appearance of your comment to the reader.Namisecond - Wednesday, January 22, 2020 - link
Because you'll never see the base frequencies in operation of either? They'll jump around from between their dormant state speeds all the way up to max turbo until you reach thermal saturation, then it'll jump around even more. Sure, I could throw an ice-water liquid cooling loop on either, put them on mains power and tell them to TDP-up well past their 15W thermal spec and throw some massive server or content creation looped workload on them, but that's not what they were meant for.The base clock speed is a lie, for both camps.
Namisecond - Wednesday, January 22, 2020 - link
From what I've seen, Intel U mobile CPUs seem to have a base clock frequency of 800mhz no matter what the documentation states. It jumps around from there to max turbo speeds. ;pIrata - Thursday, January 16, 2020 - link
And this is the top 15W Ice Lake model vs. the bottom end Renoir...and the 8C 16T 4800 U still has a higher base clock (1.8 Ghz) and a higher turbo...Of course, we'll still have to wait for the first reviews of Renoir in a - hopefully - properly built and configured laptop to see how it does in real life (clocks, power consumption in idle, light load and heavy load).
gagegfg - Tuesday, January 14, 2020 - link
"... the clock speeds are absolutely bad to compete with Intel."Core i7-1068G7 Base Freq. "1.3 Ghz" (quad core)
Core i7-1035G7 Base Freq. "1.2 Ghz" (quad core)
Core i7-1035G4 Base Freq. "1.1 Ghz" (quad core)
Core i7-1035G1 Base Freq. "1.0 Ghz" (quad core)
Core i7-1005G1 Base Freq. "1.2 Ghz" (dual core)
I do not think so..............
Korguz - Tuesday, January 14, 2020 - link
man gondalf.. that post makes NO SENSE what so ever.. were you drunk?? high on drugs ?? upset because you beloved intel isnt the leader on desktop.. and may not have an easy win on mobile now ?? your post also is nothing more then an intel shills personal opinion.. with no facts or proof... give it up already...Iolite - Tuesday, January 14, 2020 - link
AMD does not even really need to compete toe to toe with Intel, they just needs to get really close, say 95% of the way. AMD's strategy is to get you almost everything Intel has to offer at 25-50% less at retail. When your talking a gain of 5% to maybe 18% tops better performance with an Intel, pending on chip class, you're paying an additional $75-$150 for it. Add to that the fact AMD it's finally producing similar power consumption, well you can see why AMD shares have been soaring over the last 5 years.However, the edge that Intel once enjoyed for over a decade has quickly shrunk to a single gate. Give AMD'S advancement in the last five years, Intel has got to be more than a little nervous about future market share.
sing_electric - Wednesday, January 15, 2020 - link
Well, put another way, AMD's finally making the same proposition with mobile that it did with desktop Zen's launch (only now they're catching up with IPC): For $X, we'll come close on single-threaded performance but we'll give you a lot more cores for multi-thread.It wouldn't surprise me to see say, Ryzen 7's offered for the same price as an i5, and Ryzen 5's for the same price as an i3.
I'm particularly interested in battery life of these chips, since AMD's finally packing LPDDR4X support, negating a platform advantage Intel's had. (Gut feeling is that AMD will be doing better than before, but Intel still pulls ahead with absolute numbers.)
eek2121 - Tuesday, January 14, 2020 - link
Ian, I feel the need to bring up that the Zen 2 cores technically aren't 'full blooded' as they have less cache. They still should pack quite a punch though, and I look forward to a review.ET - Tuesday, January 14, 2020 - link
Agreed. AMD made something of that 'gamecache'. One would assume that dropping it to a quarter of the size would have a significant effect on 'gamability'.On the other hand, a monolithic die could have lower RAM latency, which might help compensate for this.
I'll be looking forward to some deep investigations into this when the laptops hit the market.
Kevin G - Tuesday, January 14, 2020 - link
I'll third this idea.A monolithic die is an interesting data point even with the change in cache amount.
Spunjji - Wednesday, January 15, 2020 - link
Agreed - memory latency tests and their effect on IPC vs. the desktop cores would be a really interesting avenue of investigation.Plus we'll have some actual figures to shut Gondalf up with :D
Korguz - Wednesday, January 15, 2020 - link
spunjji... wanna bet ?? he will still find some way to counter that with some stupid reason, with no proof.. just to make intel appear better...Smell This - Monday, July 20, 2020 - link
*** G O N D A L F ***
Smell This - Monday, July 20, 2020 - link
*** G O N D A L F ***
Smell This - Monday, July 20, 2020 - link
G O N D A L F x x
Smell This - Monday, July 20, 2020 - link
G O N D A L F
JasonMZW20 - Tuesday, January 14, 2020 - link
Just L3 though, AFAIK. 4MB per CCX, so 4+4MB in 4800U or 1MB per core.Saves quite a bit of die (and power) from 16MB per CCX, but savings was, of course, reapportioned to iGPU.
Though this is AMD's first Zen 2 design with 7nm integrated SoC silicon (DDR4/LPDDR4 PHYs and controllers included). I really want to see the die shot.
CrystalCowboy - Tuesday, January 14, 2020 - link
So Renoir is all on one piece of silicon? Assume 75 mm2 for the Zen cores, that means the remaining ~75 mm2 contains not only the graphic cores but also the I/O that is on a separate piece of silicon in the Ryzen 3000 CPUs. And that I/O presumably gets a shrink so that it is all done with 7nm.Ian Cutress - Tuesday, January 14, 2020 - link
Yes, it's monolithic. The cores are less than 75, because there's less L3 cache in Renoir than a standard 8-core Zen 2 chiplet.Fataliity - Tuesday, January 14, 2020 - link
Judging by your own measurements, the Zen2 core on the APU should be about 50mm.You stated at release that the L3 cache was 50% of die size of the 75mm. The other 50% was the Zen2 cores. So 37.5mm die, 37.5mm L3. That was for 32MB, so at 8MB that's 1/4, or a little less than 10mm. So about 47.5mm.
neblogai - Tuesday, January 14, 2020 - link
It should be roughly 50mm for cores, the rest shared in half between CUs and everything else. That is because in 75mm^2 Matisse Zen2 chiplet, ~40%(~30mm^2) is L3, but in Renoir- there is only a quarter of that cache. This could save >20mm^2, and some space on Zen2 chiplet are IF links, not cores.ET - Tuesday, January 14, 2020 - link
I'd estimate about 50 mm2 for the Zen cores with the smaller cache. I'd give the Vega cores half of that, but with the codecs and display engine possibly another 40-50 mm2 together. The rest going mainly to internal and external I/O.Would be interesting to see a die image when the chip arrives.
Kevin G - Tuesday, January 14, 2020 - link
I was half expecting this generation of mobile chip to start following AMD's chiplet strategy via leveraging the same Zen 2 chiplets but with a IO die that had an integrated GPU. In fact, I would have thought that the IO die was also going to pull double duty as a lowend discrete GPU die when given a higher power budget.Ian Cutress - Tuesday, January 14, 2020 - link
Keeping it monolithic has benefits for the IF and also idle latency, which is key for a mobile CPU. I don't expect AMD to go chiplet with graphics on this sort of product any time soon. Maybe in something higher up the thermal stack, or for specific customers, perhaps.Fataliity - Tuesday, January 14, 2020 - link
Do you know if the motherboard has a chipset? Or is the chipset in the die IO?sing_electric - Wednesday, January 15, 2020 - link
What's interesting is that AMD also launched their 1st mobile-focused dGPUs in ages at the same time. My bet is AMD will keep the mobile CPU's monolithic but might possibly have release a version to specific customers with either the iGPU part disabled, or cut down to 1 active CU (for battery life purposes when mostly idle) for use with dGPU designs. Depending on the chip layout, the dead silicon from the other CUs might let the chip run with more power longer before throttling.Namisecond - Wednesday, January 22, 2020 - link
I think the target market you are thinking of is called "embedded"ET - Tuesday, January 14, 2020 - link
I had expected that to be the case for 45W mobile and for desktop APUs, with 15W APUs being 4 cores. I'm glad that AMD gave us 8 cores at 15W.I still don't think that a chiplet configuration is out of the question for future high power APUs.
Gondalf - Tuesday, January 14, 2020 - link
Honestly i think we well'never see much in the 15W of this SKU, bet 25 W will be the standard (70% more power headroom). 8 cores eat a lot of power and running then slow do not help much the customer and makes your Laptop not much snappy/delusional.alufan - Tuesday, January 14, 2020 - link
sigh~~~~ people doing work at x speed
~~~~~~~~people doing the same work at the same speed
simple maths really but lets wait for the reviews and I for one look forward to your comments then
Korguz - Tuesday, January 14, 2020 - link
you are the one that is delusional gondalf.. looks like your thinking is.. if intel can do it.. no one else can too.. maybe thats when intel can only do 4 cores with what ever lake they are making for mobile now.. start posting proof of your opinions or go away..iranterres - Tuesday, January 14, 2020 - link
With workloads becoming increasingly more parallel on CPUs, 8 cores is just so sweet.vortmax2 - Tuesday, January 14, 2020 - link
Some of the best, non-combative comment I've seen in a long time here. Everyone is excited about 2020 laptops due with these chips. Very happy that AMD is back in the game - benefits us all.outsideloop - Tuesday, January 14, 2020 - link
Ian, maybe you could comment on what Intel puts in the I/O chiplet of Ice Lake compared to what AMD has put in/left out of the Renoir chip?ksec - Tuesday, January 14, 2020 - link
>I had stated in our article that I estimated 150 mm2 for the die size. Turns out, I wasn’t too far wrong.Once you have eaten that many wafers, you can literally smell the die size XD
I was wrong in that AMD wanted to have GPGPU uArch in the iGPU, since Dr Su already stated there will RDNA in Ryzen Mobile in the future. But even with the current Ryzen Mobile, it doesn't look like Tiger Lake is attractive. It will be Quad CPU with AVX512, so it may be a better CPU but at half the Core, and Xe GPU I would bet it will be at best on par with AMD.
And then early next year you get AMD may be with Zen 2+ and RDNA2 ( Which is forms the basic of PS5 and Xbox X ) on TSMC N6.
But generally Speaking, despite everything that has happened over at Intel, and AMD being technically better in Desktop, Server and now Laptop, AMD is still not getting the market shares. I am disappointed in their sales, marketing, and forecasting department. Server is still a dismal 4%. Desktop CPU still under 20%. AMD needs to be a lot more aggressive especially in the Server Department which is where the big bucks are.
extide - Tuesday, January 14, 2020 - link
There is no Zen2+ (Why do people keep on about this..?)deksman2 - Tuesday, January 14, 2020 - link
Isn't AVX512 use relatively limited for consumers?Besides, AMD would easily retain the advantage with its increased core count and if software devs actually optimised their codes to take advantage of Zen uArch as it should.
That said, Zen2+ doesn't exist.
Zen 3 exists which will be a completely new architecture and produced on 7nm+ EUV node.
Also, Zen 3 should be released by the end of THIS year (2020) ... not early 2021.
2021 (most likely end of 2021) is slated for Zen 4 on 5nm (which seems to be on track because the yields have gone over 50% already).
sarafino - Tuesday, January 14, 2020 - link
Little use to most laptop consumers. Plus AVX512 is such a power hungry feature that it causes severe throttling to keep temperatures in control.IC asked Mark Papermaster if a Zen 2+ was planned in the future and Mark responded:
"MP: I have nothing to say on any refresh on current designs, but we always look at where it makes sense and where we’re able to take opportunities to provide a bump in performance, power, or die area."
So interpret that however you will. I don't see AMD squeezing in a Zen 2+ between Zen 2 and Zen 3, considering AMD has been pretty consistent with releases non their roadmap. Maybe Z2+ will live on in lower tier CPU's just as Z1/Z1+ has for the last year.
ash9 - Tuesday, January 14, 2020 - link
Very curious about the memory subsystem the Threadripper 3990X uses DMA to communicate to the GPU, and the 4000 chip look as if it does too, what's happening within the APU (oh it's not an APU anymore).Namisecond - Wednesday, January 22, 2020 - link
None of the Threadrippers have an on-die GPU, the 4000-series APU does. The more important question you should be asking is how does the 2000-series and 3000-series APUs communicate with their on-die GPUs?zamroni - Tuesday, January 14, 2020 - link
with 90% yield, most outcome will be the 16 thread 4800u. this is great news for laptop consumers.i just hope it has enough pci lanes to accommodate x4 thunderbolt port.
eva02langley - Tuesday, January 14, 2020 - link
Intel Ice Lake - 4 cores - 122.52 mm2AMD Renoir APU - 8 cores - 149.22 mm2
This should clear up the question about the density. Intel 10nm is clearly inferior to TSMC 7nm.
extide - Tuesday, January 14, 2020 - link
Well, 4 cores takes up about 25mm2 so 8c Ice Lake would be ~145mm2 they seem pretty on par actually -- depends on how much space the GPU takes on each but I wouldn't say it's clearly inferior.Fataliity - Tuesday, January 14, 2020 - link
Actually 2 cores of Ice Lake is about 25mm, (I think 22mm), 4 cores is about 50mm. So 8 cores would be 100mm. 166.52extide - Tuesday, January 14, 2020 - link
Actually 4c is 30mm2 https://en.wikichip.org/wiki/intel/microarchitectu...So -- that puts it almost exactly at 150mm2 like Renoir.
cheshirster - Wednesday, January 15, 2020 - link
ICL has AVX-512 and a large part of silicon dedicated to TB3.I'd say it is densier.
But don't forget this is 10+ already, not the original 10nm.
e36Jeff - Tuesday, January 14, 2020 - link
You're slipping Ian, you were off by 0.52%. I expected better of you.AlB80 - Tuesday, January 14, 2020 - link
How many PCIe lanes for gfx does Renoir have? x8 or x16?shing3232 - Tuesday, January 14, 2020 - link
Also, Renoir cut down L3 by Large margin.Lord of the Bored - Tuesday, January 14, 2020 - link
"This image is not to scale."Well, color me disappointed. Thought we were going back to the days of REAL COMPUTERS that filled up an entire filing cabinet.
MDD1963 - Thursday, January 16, 2020 - link
Would be nice to see a desktop APU (R5-4600G?) equivalent with 6c/12t, and enough GPU power to match a GTX1050 or so at gaming.... (IF folks thought the 2200G/2400G sold well, the above should sell by the proverbial freighter-full!maroon1 - Thursday, January 16, 2020 - link
Gtx 1050 ? You must be jokingCurrent desktop APU barely match GT 1030 GDDR5
And Renoir is barely better than current APU in the GPU side
Korguz - Friday, January 17, 2020 - link
" And Renoir is barely better than current APU in the GPU side " ahhh so you have bench marks that show this ?? post a link so the rest of us can see then tooNamisecond - Wednesday, January 22, 2020 - link
Do you have benchmarks showing 1050 levels of performance? Post a link so the rest of us can see then tooKorguz - Wednesday, January 22, 2020 - link
Namisecond, maroon1 wont post any links, cause there arent any, yet. this is their own personal opinion, with some personal bias thrown in, nothing more....Spunjji - Friday, January 24, 2020 - link
OP said "Would be nice to see..." - AKA "I'd like it if AMD did this...", not "AMD have done this".maroon1, on the other hand, made a performance claim that they can't back up. So maroon1 needs to justify that with evidence, whereas neither OP nor korguz have any such burden of proof.
Smell This - Friday, January 17, 2020 - link
Seems to me that there currently is a point of diminishing returns with the AMD APUs (sorry ... I'm going to keep calling it that, Dr Su !).Certainly on mobile, 8 CUs (512 shaders) seems to be the sweet spot. 10/11 CUs raises the performance bar, but not 1:1, and especially on the DT. Maybe DDR5 and Navi/RDNA-II can change the formula a bit, but a chiplet w/eDRAM (as a 4th level cache?) would be the logical long-term path ...
Namisecond - Wednesday, January 22, 2020 - link
You may be mistaking popularity by price for popularity by performance. I'd argue that 2200G/2400G sold well because they sold for a significantly lower price than their larger core count, GPU-less brethren. If the new desktop APUs sell at a similar price point, it should do well. Don't expect significantly better GPU performance from the 4000-series APUs over their older counterparts. Even 20% may be expecting too much.Gonemad - Thursday, January 16, 2020 - link
Huge chip areas at lower TDP = excellent cooling. Fantastic. Some bold manufacturers may even tinker with PASSIVE cooling?I mean, we are back on the 486 era, where chips only needed a heatsink. That's a full circle. I can see bold designs, with maybe just passive watercooling, or just a heatpipe plugging the thing into the aluminum case of the notebook, - turning it into a decent heatspreader - being feasible, without Apple levels of sacrifice to performance or durability.
Namisecond - Wednesday, January 22, 2020 - link
IIRC, 486 chips were right around 10W and I definitely recall my 486/33 system came with a fan for the heatsink. Even now, 10W TDP seems to be the tipping point where vendors consider active cooling for their products. My EDC laptop with a 6W TDP CPU makes do with a heat spreader plate, a big one. The 10W TDP desktop version of that same chip came with a fairly elaborate heatsink, sans fan. Anything 15W TDP and up will most likely come with active cooling or a very elaborate custom passive cooling setup.peevee - Monday, January 27, 2020 - link
They just need to start installing CPUs (and its LP memory, and M2 slots) into the screen half of the laptop, with ribbed aluminum as its cover. Much more cooling area where there is no keyboard, and even 2mm ribs with 2mm pitch increase the area by 3x again, easily.Namisecond - Wednesday, January 29, 2020 - link
Uh, no...unless you enjoy laptops that fall over backward when you open them up...jjjag - Wednesday, January 22, 2020 - link
The table is wrong. Ice Lake is 10+. The Canon Lake that was released in 2017 is the original 10nm. No idea if there is an official word, but wiki says Tiger Lake is 10++ and I think that is probably correct. I think it's also safe to assume that the slides you showed 2 years ago from Intel that had 10++ with a performance advantage over 10+ are now incorrect as Intel has probably changed direction in favor of manufacturability and lower power over raw performancee36Jeff - Saturday, January 25, 2020 - link
do we know if this is still 2 4-core CCXs or if this is a single 8-core CCX design?peevee - Monday, January 27, 2020 - link
Listing "EUs/CUs" in the table is worse than useless, it is misleading. Just list number of 32-bit ALUs.dhruvhalwasiya - Tuesday, February 25, 2020 - link
It would be really cool if you could post an article comparing the die sizes of mobile SOCs over the past decade especially the A series vs the snapdragons and then the snapdragons for windows vs x86 processors for the same!