Zen 5 is not for gamers or Windows machines. It does deliver promised performance on Linux, which shows it's for datacenters and AI/workstation uses, not gamers. This is the trend now in the space created by AI and HPC.
Mainly because their zen5 core die is designed and used to make server product 😂 It just a server driven design surprisingly not bad for gaming task Replace the IO die with consumer grade and vola
Nah man. That’s not it, not quite. Linux’s scheduler is better aware of the latency between the CCDs which doubled this generation. It’s faster than Windows on cross platform benchmarks. Windows mostly needs patches but that could be a while. 24H2 release maybe. Zen 5 is still _incredible_ in terms of AVX-512 which I look forward to using for emulators and video encode. There’s big architectural changes that should pay off down the line, but they regressed or just traded blows with Zen 4 in other microbrenchmarks. Software issues too. I’ll probably wait till Zen 6 for the next I/O controller and memory controller.
Yea, as long as the power envolope is significantly lower than a cpu + gpu combo. I think for mini pcs (like the ones that minisforum makes) this would be really cool aswell
@@SiliconSteak i suppose we could use the apple M chips as reference for that. it's probably going to be ballpark the same stuff as an m3, if a bit higher
Although AMD's recent handling of Zen 5(and a lot of things) as been pathetic, to say the least, I'm still stoked about getting new super powerful APUs. It would be absolutely market shattering if a CPU is able to get similar performance almost that of a Laptop RTX 4070, while consuming less power overall, especially if it's a lower overall cost. I hope Intel also follows suit, cuz having a CPU + GPU + NPU in one package is really useful.
@@SiliconSteakAMD is well positioned but definitely not 100% certain to be the first as it’s not the only one. Intel’s got all the required processor architecture design and advanced packaging technology to bring lower power high performing platform in the form of a SoC with CPU + next-gen improved 1080p capable iGPU + NPU with super small data type support. Also the iGPU has XMX bringing extra TOPs which is ideal for software that aims to utilise multiple AI accelerators as such. Interesting to see both releasing, very exciting for the future industry direction nontheless.
It all depends on price. Since it´s an integrated chip it will depend on what the whole product costs. I think it´s cool that we will get this level of performance in integrated chips. Will probably make the product cheaper as a whole as well. It will probably be enough for most people needs without breaking the bank.
Manufacturing cost may be cheaper but final end product will be more expensive than a regular Strix Point + RTX 4060/4070 laptop. It’s efficiency tax. It also requires new integration from laptop manufacturers which add up.
APUs are not monolithic due to bandwidth and latency issues. Chiplets do not cause bandwidth or latency issues significant enough to be noticed. The reason for APUs being monolithic is power consumption. It takes more power to drive signal lines between chiplets. As such, devices that must be power efficient (mobile/handheld/laptop) are all going to want to use parts based on monolithic designs. But as soon as power is not a concern, chiplet based designs are just as fast and can be manufactured at a lower price.
@@SiliconSteak Sorry, you're the one who is COMPLETELY wrong. Stating "previous APUs are monolithic due to bandwidth and latency constrains caused by chiplet design" shows you have pretty much zero knowledge about chip design and inability to research or even logical thinking...
@@yibozhao1012 Remember, APU include both consoles APU, PlayStation and Xbox, where power isn't much of a concern. So, it's because of both bandwidth/latencies and power consumption. And you can expect AMD to use a next-generation interconnect on Zen 6 to drive down latency and increase the bandwidth.
@@807800 There were interviews with the product manager of Zen 4 APUs and he stated very clear: they chose the monolithic design due to power consumption, nothing eles mentioned. Unless you think he was intentionally misleading the public and you have enough knowledge about chip design on silicon level to debunk his claims, you must take his words i guess. Fact is moveing data between chips (through copper wires as the current AMD methodes) consums about 100 times the energy compared to moving same amount data on the same chip. And with your logic, laptops, handhelds and consoles had a bigger concern about latency and bandwidth than high-end PCs and servers... and lastly: with consoles power consumption is not a big concern, but heat dissipation IS.
@@yibozhao1012 "Zen 4 APUs", Yes, for Zen 4 APU, but, as I said, there's consoles APUs. The fact is, monolithic design offer both bandwidth/latencies and efficiency. In fact, as I mentioned above, you can expect Zen 6 to use next-gen interconnect on Ryzen desktop which would bring the improvement in bandwidth/latency and efficiency. I'm sure you are probably aware that one of the issues with Ryzen dual CCD CPU like 9950X is the latency between the CCD, which is why for a gaming CPU, you would be better served with single CCD like 9700X.
About memory bandwidth: It's basically just bus width * transfer rate. On a 256bit bus it's 256bit=32B => 32B/T*8000MT/s=256000MB/s=256GB/s At 7500MT/s it would be about 240GB/s
@@SiliconSteak doesn’t have any effect on theoretical throughput. Latency is pretty much independent from bandwidth. It should influence real world throughput though.
Yes, this product is welcome, I myself looking for this because of AMD's Linux support, but there is something which many people are not aware, as he mentioned everybody wants a PS5 SoC into a laptop, and here we are with obvious reasons that AMD could to that for a long time, especially when they license to others GPU IP and other things under unknown to the public royalties fees. Lastly don't forget that AMD have had for a long time IP for both CPU/GPU usage especially when compared to intel who has to license from Nvidia to be able to sell you a product with an iGPU on board.
Amd might have or might have had deals with MS, Sony to not bring those powerful console level APUs to desktop, pc. I can't think of any other reason for them not doing it till now.
When this thing comes out I get my first laptop in ~2 decades xD Been waiting a hell of a long time watching. Switching GPUs sucks for Nvidia or AMD, they both have issues.
Just found your channel and subbed Also this new APU is what I have been wanting for a long time I am a x360 user who needs good graphics like not max graphics but like a mobile 4060 and it might or might not reach that level but if it isn't that much slower i would still love it for the weight and battery life improvements
I'd like to point out that Strix Halo is chip for "mobile" laptop, not desktop. laptop gamer (gaming laptop with discrete GPU) have quite lower performance expectation than desktop gamers. They don't expect recent AAA game run 60 FPS at ultra setting+1440p. More modest setting and performance is OK for them. Expensive laptop models with high performance GPU like 4080 doesn't sell well. Gaming performance of gaming laptop is largely limited by its battery or cooling. So most laptop gamers are OK with laptop with 4060 mobile or 3060 mobile GPU. So performance of RX6550 (or 4060 mobile) is not satisfactory for desktop gaming user, but enough for satisfying gaming laptop gamer/buyer.
Has nobody noticed that the uplift of Zen 5 is huge in the mobile space? the AI 9 370H and such? There was never going to be a super uplift in gaming, that's your GPU now anyways. The performance uplift at desktop and high power is more modest, and the gamers have so bellyached about their niche that they haven't paid attention to what people reviewing laptops in the low-power mobile niche think of Zen 5. There it is hugely successful ...
After quick google search it seems like the Ryzen 300 series is a monolithic design. This is likely the reason there is actually an increase in performance.
@@SiliconSteak most of AMD's designs are monolithic. Only dragon range (7045) were chiplet designed. These cpus were probably intended to replace ryzen 8000 series mobile and ryzen 7040 series instead. Which are mono designs. The main performance boost from this generation likely stems from the big little design from the zen c cores instead.
Ooof. This is what I was talking about in the vid. If amd wants to make a product like this, they should use the cost savings to lower the price. They never utilize a opportunity to not use the advantage they have
Why haven't AMD done what we all know they can do? One word...sandbagging ;) Don't put too much hope on Strix Halo. It definitely will NOT reflect what we all hope for, and definitely will NOT reflect what AMD could easily produce. Additionally, expect Strix Halo to be expensive...very expensive, especially if AMD includes decent CPU cores and speed. Unfortunately, it is what it is.
Why do they sandbag though if that is the case? Just a matter of not getting into an actually competitive struggle with nvidia? I guess this kind of "implicit cartel" can happen with a market of only 2 or 3 players.
2:20 While yes, on mainstream desktop Zen 5 only offer little to gains specially in gaming, on server/AI workload it's actually impressive. It offers on average 35% uplift in cryptography (server workload), 30% in machine learning, 22% in HPC (simulation stuff). Basically EPYC stuff. And Ryzen desktop is basically just EPYC die combined with mainstream I/O die. So, his leak is actually accurate. EPYC is where the money goes after all.
If Mlid said Desktop and your speaking of Server, and now the uptic is like 6-10% he was wrong. Avx512 is what made chip better but desktop rarely uses it so no the leak was highly misleading. The Amd fanboys took the leak to heart and talked down the 14900k based on this copium. Because of Mlid wrong information I almost didnt buy the 7800x3d on offer. It turned out this was the best purchase since the new cpus sucks without overclocking. If Mlid really new what he was talking about he would know that its the Soc die using like 17w and core ccx uses 55w. I found this out by looking at Hwinfo64 after buying the 7800x3d. So even with an marginal improvement on a ccx that wouldnt help much because of the improvements would only apply to the 55w, and IMC we knew would stay the same since its on the SoC die. To me it looks like Mlid doesnt use any of the cpus he is talking about at all.
Really hope this comes out as a mini pc variant or socketed. Like others have been saying though - why have they not done this before now? What is the business strategy?
Anyone’s guess. Probably they haven’t done it because they haven’t needed to. with intel making better gpus and Qualcomm trying, they need to now to get ahead
I am looking forward a mini PC based on that Strix Allo. A mini PC with a large silent fan that will outperform a lot the previous mini PCs. Meantime i am enjoying my 7840HS based mini PC.
@@vladimirljubopytnov5193 when I said this I meant that by them being so close, they have a unified memory interface, which allows them to be much faster and more efficient
Camera and lighting is kinda weird. makes me think it has a 50% chance of being an AI face. I do appreciate the video content, I am not sure if its because you are sick, but I would like to see you write a script for the videos, refreshing takes on things. Maybe do a recording without the face came then run the audio through descrypt and work the script from there then record the the face cam then overlay it on the material recording from the first take.
lol I look like a AI? I was having a hard time with this one because I was sick/drugged on NyQuil (I definitely drink more than 1 serving) as for the lighting, I thought that was 1 of the only good things I had…. I film this with my iPhone so I know it won’t always look the best 🤷♂️
No. Well, they will likely put it in mini pcs. like I said in the vid, because they are using a 256bit bus, you can’t use a reg motherboard. The ram will be soldered on this one
>Swan song to mobile gamers swan song (noun) -- a farewell appearance or final act or pronouncement (Merriam-Webster) Do you believe that AMD is going to leave mobile space or something? Also Strix Halo is fun, but MI-300 was a much more interesting product even if it's not for mortals, just because of how complex it is. P.S. people pretend like RDNA 2 to RDNA 3 jump per core is weak and that's somehow bad. Sure, But if you look for example 3060 (3584 cores) vs 1080ti (3584 CUDA cores) they often have similar performance (obviously discounting ray tracing). That's just the feature of GPUs, very weak cores that manufacturers pack more and more of
29:52 ROFL not really a fair shot there, and you should know better. He bases his predictions on the information he is getting from his contacts and that was 3 years before release like, it not possible that AMD can't make changes in the interim and then not tell him right?
Also included one of his recent predictions. Nothing wrong with reporting leaks or making predictions, just gets annoying when he never admits when he is wrong. Seems to have a lot of pride going on
ddr speed equals bits even it is quad time mhz real ramclock is 1000mhz times 8 this is why transfer per second equals mhz because of 1 bit per mhz this is a ram standard since sram was invented therefore 256bit dual channel +- 240gb/s depends on cpu clockspeed and latency Igp needs inf cache to reduce latency from socket cpu if later am6 will have also 256bit Si and camm2 ddr 6 at 12gt/s Now with strix halo the biggest apu ever we will see if the design works i think it will
click bait title, strix halo will be mediocre. the best variant uses way too much power for a handheld and performs worse than a 4070laptop. the low end variant will be ~10% better with a gtx 1060 which isnt good enough to play modern videogames and will really suck a handheld dry of power in no time, so you wont really be able to game off a charger in either case. I don't see the point of strix halo at all. its not considerably cheaper than buying brand new hardware either. I built a 700$ 14th gen intel 4060 ti desktop pc with 32gb of ram last year that outperforms strix halo's best variant. Its really just a novelty toy at this point. A worth while apu is still 5 years out minimum.
In the vid I was talking about how it wouldn’t make sense if it used too much power, and I estimated it to be slower than the 4070M. The point is to get more gpu perf in a smaller space with less of a power envelope. This has always been the point of APUs. If it doesn’t do this, it’s a flop IMO
Also, I appreciate your comment. This channel is all about creating an active discussion. If I’m wrong tell me and we can talk about it, I am not bringing an ego into this. I’m just a nerd who loves new technology
@@SiliconSteak Yeah, i agree with your talking points, just not the title. but i understand it gets views. also why i commented critically, alot of the time people engage more
@@topelite666 exactly, also I mean. I think that if strix halo did succeed, maybe it could be the beginning of the end. Eventually we won’t need more performance, we will just want more power efficiency, and apus may give us that
bro what mlid said was all during development of zen 5, you never know what the real performance of the final product will be, maybe they ran into problems in development and mitigating those issues lost all the performance or something, and apparently they pushed out an unfinished microcode or something, so in a few months we may se e 10 to 15% better performance
This is my real take, he came up with a lot of different excuses why zen 5 was a flop. He is coping hard rn, i tuned into his recent vids and his on his most recent broken silicon he was bragging about how good zen 5 is. He has the need to be right and when he is wrong instead of admitting it, he makes excuses. I still like his content, that's just the way i see it
@@SiliconSteak His posts are also extremely negative against Intel. While he can justify it personally, his anti-intel narrative gets annoying real fast when you're just there for some news. He also has the need to be right like you said, therefore he keeps bashing ARC, since they somehow keep defying his prophecy of ARC going to disappear. I find it revolting when there might be like thousands of people working on it now, and he just comes out and gives out this negative perception on a genuine effort. They did very well for their first foray into discreet graphics, and for productivity they hit it out of the park. Wish they prove him and the naysayers wrong and stick it out till the end.
@@SiliconSteak In Hardware Unboxed's most recent podcast 3 days ago, they talk about this. The chapter in the middle of the video titled "What role did leakers play". YT watch ID: lUk7rfY5KQQ Like Tim says there, if he just takes a more reportorial stance as opposed to trying to be a prophet, then he wouldn't even have to make any excuses. Maybe he thinks his channel becoming prominent only because of the accuracy of his "Leaks" are but most people are just there for the discussions & information [I mean accuracy isn't the be all and end all for unreleased products anyways]. Which you are doing very well. Maybe your channel might become bigger than his one day.
@@SiliconSteak He has a ton of copium going on right now with Zen 5. He was wrong and has been wrong about a lot of things. Only problem is he doesn't admit it when hes wrong.
@@Tom-bt8eg do you really think AMD is that far of during their development? It was obvious from the beginning that improvements will be rather small (no significant node change, few changes to the overall design, most notably still using the long overdue infinity fabric). If the product wasn’t ready, they wouldn’t launch it, they knew exactly what they’re delivering. MLID has an extremely pro AMD community and positive leaks about AMD are received way better than positive leaks about Intel or NVIDIA. He’s telling his community what they want to hear. When you look at leaks, there is a tendency that AMD improvements are usually way below what leaks promised while Intel improvements are usually better than expected (see Rocket Lake, Alder Lake and Raptor Lake). Then, AMD usually manipulates their IPC improvement claims to be better than they’re in reality. But that doesn’t bring much clicks. About MLID: there are statistics on his accuracy on Reddit. He often deletes videos that didn’t age well and in total all his leaks that he’s >80% sure about have an average accuracy of around 50%. Note that this doesn’t differentiate in how exact his leaks are. Most of his leaks which are true (in the 50%) are not really precise or pretty obvious like process nodes, naming, TDP etc. or rough estimations (like at least 10% better).
has your grandma not taught you not to cough freely? even coughing into the camera, so rude! do it into the inside of the elbow if you can't reach for a tissue. jesus, kids these days.
I forgot the camera was to my side, ( I was basically drunk on NyQuil for this vid) so I accidentally turned to my side where the camera was and coughed 🫥
They're basically releasing what they've been doing for PS and Xbox for years.
Zen 5 is not for gamers or Windows machines. It does deliver promised performance on Linux, which shows it's for datacenters and AI/workstation uses, not gamers. This is the trend now in the space created by AI and HPC.
Mainly because their zen5 core die is designed and used to make server product 😂
It just a server driven design surprisingly not bad for gaming task
Replace the IO die with consumer grade and vola
Nah man. That’s not it, not quite. Linux’s scheduler is better aware of the latency between the CCDs which doubled this generation. It’s faster than Windows on cross platform benchmarks. Windows mostly needs patches but that could be a while. 24H2 release maybe. Zen 5 is still _incredible_ in terms of AVX-512 which I look forward to using for emulators and video encode. There’s big architectural changes that should pay off down the line, but they regressed or just traded blows with Zen 4 in other microbrenchmarks. Software issues too. I’ll probably wait till Zen 6 for the next I/O controller and memory controller.
Me, playing all of my game library on Ubuntu: ok
Microsoft fixing the ZEN5-problem now. Windows 11 Update 24H2: up to 35% more FPS now on AMD Ryzen-9000
honestly, i kinda agree. i wouldn't need a gpu with this thing and it on a laptop, man, it's going to be perfect
Yea, as long as the power envolope is significantly lower than a cpu + gpu combo. I think for mini pcs (like the ones that minisforum makes) this would be really cool aswell
@@SiliconSteak i suppose we could use the apple M chips as reference for that. it's probably going to be ballpark the same stuff as an m3, if a bit higher
You cant run 110-140w on a laptop. How are you going to cool it and it would drain the battery under 1 hour.
@@impuls60 meteor lake can go 150w on laptop
Although AMD's recent handling of Zen 5(and a lot of things) as been pathetic, to say the least, I'm still stoked about getting new super powerful APUs. It would be absolutely market shattering if a CPU is able to get similar performance almost that of a Laptop RTX 4070, while consuming less power overall, especially if it's a lower overall cost. I hope Intel also follows suit, cuz having a CPU + GPU + NPU in one package is really useful.
100%. AMD is actually uniquely positioned to do this before any else in the x86 space. Let’s hope they don’t screw it up 🤞
@@SiliconSteakAMD is well positioned but definitely not 100% certain to be the first as it’s not the only one. Intel’s got all the required processor architecture design and advanced packaging technology to bring lower power high performing platform in the form of a SoC with CPU + next-gen improved 1080p capable iGPU + NPU with super small data type support. Also the iGPU has XMX bringing extra TOPs which is ideal for software that aims to utilise multiple AI accelerators as such.
Interesting to see both releasing, very exciting for the future industry direction nontheless.
Can't wait for the Ryzen 4070
It’s going to be the best price/performance!!!
It all depends on price. Since it´s an integrated chip it will depend on what the whole product costs. I think it´s cool that we will get this level of performance in integrated chips. Will probably make the product cheaper as a whole as well. It will probably be enough for most people needs without breaking the bank.
Manufacturing cost may be cheaper but final end product will be more expensive than a regular Strix Point + RTX 4060/4070 laptop. It’s efficiency tax.
It also requires new integration from laptop manufacturers which add up.
If this APU comes out with these specs I’m genuinely going to touch it
APUs are not monolithic due to bandwidth and latency issues. Chiplets do not cause bandwidth or latency issues significant enough to be noticed. The reason for APUs being monolithic is power consumption. It takes more power to drive signal lines between chiplets. As such, devices that must be power efficient (mobile/handheld/laptop) are all going to want to use parts based on monolithic designs. But as soon as power is not a concern, chiplet based designs are just as fast and can be manufactured at a lower price.
You’re wrong about this. I mean you’re partially wrong
@@SiliconSteak Sorry, you're the one who is COMPLETELY wrong. Stating "previous APUs are monolithic due to bandwidth and latency constrains caused by chiplet design" shows you have pretty much zero knowledge about chip design and inability to research or even logical thinking...
@@yibozhao1012 Remember, APU include both consoles APU, PlayStation and Xbox, where power isn't much of a concern. So, it's because of both bandwidth/latencies and power consumption.
And you can expect AMD to use a next-generation interconnect on Zen 6 to drive down latency and increase the bandwidth.
@@807800 There were interviews with the product manager of Zen 4 APUs and he stated very clear: they chose the monolithic design due to power consumption, nothing eles mentioned. Unless you think he was intentionally misleading the public and you have enough knowledge about chip design on silicon level to debunk his claims, you must take his words i guess. Fact is moveing data between chips (through copper wires as the current AMD methodes) consums about 100 times the energy compared to moving same amount data on the same chip. And with your logic, laptops, handhelds and consoles had a bigger concern about latency and bandwidth than high-end PCs and servers... and lastly: with consoles power consumption is not a big concern, but heat dissipation IS.
@@yibozhao1012 "Zen 4 APUs", Yes, for Zen 4 APU, but, as I said, there's consoles APUs.
The fact is, monolithic design offer both bandwidth/latencies and efficiency. In fact, as I mentioned above, you can expect Zen 6 to use next-gen interconnect on Ryzen desktop which would bring the improvement in bandwidth/latency and efficiency. I'm sure you are probably aware that one of the issues with Ryzen dual CCD CPU like 9950X is the latency between the CCD, which is why for a gaming CPU, you would be better served with single CCD like 9700X.
About memory bandwidth:
It's basically just bus width * transfer rate.
On a 256bit bus it's 256bit=32B => 32B/T*8000MT/s=256000MB/s=256GB/s
At 7500MT/s it would be about 240GB/s
What about timings? My understanding is this makes a difference for ddr memory
@@SiliconSteak doesn’t have any effect on theoretical throughput. Latency is pretty much independent from bandwidth.
It should influence real world throughput though.
@@lukas_ls gotcha
Yes, this product is welcome, I myself looking for this because of AMD's Linux support, but there is something which many people are not aware, as he mentioned everybody wants a PS5 SoC into a laptop, and here we are with obvious reasons that AMD could to that for a long time, especially when they license to others GPU IP and other things under unknown to the public royalties fees.
Lastly don't forget that AMD have had for a long time IP for both CPU/GPU usage especially when compared to intel who has to license from Nvidia to be able to sell you a product with an iGPU on board.
If its way too expensive its DOA. APUs are suppose to be cheap
The main problem with APUs is low ram bandwidth they aim to solve that with higher bus width in halo's case 256bit and also infinity cache
That and the power/heat
@@SiliconSteak intel has been makin super high power draw cpus for years . Unless you're using it for handhelds power draw isn't the first problem
@@50H3i1 for a apu, I think keeping power down is really important
Amd might have or might have had deals with MS, Sony to not bring those powerful console level APUs to desktop, pc. I can't think of any other reason for them not doing it till now.
It would be pretty cool if AMD would get their Instinct MI300A as a single unit in a desktop workstation. Now *that* is an APU!
I wanted to believe the hype about Strix Halo. Then Zen 5 came out. So... no. I don't believe it anymore.
It will probably be 10 percent better than the 780m lol
@@mexreax4493 that would be crazy bad performance for pushing to 40CUs and a 256bit bus
@@SiliconSteak well we will see.
on linux productivity zen 5 actually had about 15~25% uplift. it's games that it's disappointing for now
Well they also had power efficiency go up.c
When this thing comes out I get my first laptop in ~2 decades xD
Been waiting a hell of a long time watching. Switching GPUs sucks for Nvidia or AMD, they both have issues.
Just found your channel and subbed
Also this new APU is what I have been wanting for a long time
I am a x360 user who needs good graphics like not max graphics but like a mobile 4060 and it might or might not reach that level but if it isn't that much slower i would still love it for the weight and battery life improvements
I'd like to point out that Strix Halo is chip for "mobile" laptop, not desktop. laptop gamer (gaming laptop with discrete GPU) have quite lower performance expectation than desktop gamers. They don't expect recent AAA game run 60 FPS at ultra setting+1440p. More modest setting and performance is OK for them. Expensive laptop models with high performance GPU like 4080 doesn't sell well. Gaming performance of gaming laptop is largely limited by its battery or cooling. So most laptop gamers are OK with laptop with 4060 mobile or 3060 mobile GPU. So performance of RX6550 (or 4060 mobile) is not satisfactory for desktop gaming user, but enough for satisfying gaming laptop gamer/buyer.
For sure. The expectation for mobile performance is much lower and that’s justified with having smaller screen sizes and lower power consumption
Too bad it will be priced for the AI / Big Data market.
Has nobody noticed that the uplift of Zen 5 is huge in the mobile space? the AI 9 370H and such? There was never going to be a super uplift in gaming, that's your GPU now anyways. The performance uplift at desktop and high power is more modest, and the gamers have so bellyached about their niche that they haven't paid attention to what people reviewing laptops in the low-power mobile niche think of Zen 5. There it is hugely successful ...
I didn’t know zen 5 was out in laptops yet
After quick google search it seems like the Ryzen 300 series is a monolithic design. This is likely the reason there is actually an increase in performance.
@@SiliconSteak I rejected the ASUS G16 because everyone reported it was still too hot, but the new version ...
@@SiliconSteak most of AMD's designs are monolithic. Only dragon range (7045) were chiplet designed. These cpus were probably intended to replace ryzen 8000 series mobile and ryzen 7040 series instead. Which are mono designs. The main performance boost from this generation likely stems from the big little design from the zen c cores instead.
Moore's Law is Dead was tricked in April fool's about Zen5 performance and couldn't even make a video to repair his credibility, what a clown xD
asus predator xenomorph gamers republic laptop gonna cost 3000 dollars with the ryzen halo chip
Ooof. This is what I was talking about in the vid. If amd wants to make a product like this, they should use the cost savings to lower the price. They never utilize a opportunity to not use the advantage they have
3:18 183 now :)
Why haven't AMD done what we all know they can do? One word...sandbagging ;)
Don't put too much hope on Strix Halo. It definitely will NOT reflect what we all hope for, and definitely will NOT reflect what AMD could easily produce. Additionally, expect Strix Halo to be expensive...very expensive, especially if AMD includes decent CPU cores and speed. Unfortunately, it is what it is.
Why do they sandbag though if that is the case?
Just a matter of not getting into an actually competitive struggle with nvidia? I guess this kind of "implicit cartel" can happen with a market of only 2 or 3 players.
2:20 While yes, on mainstream desktop Zen 5 only offer little to gains specially in gaming, on server/AI workload it's actually impressive. It offers on average 35% uplift in cryptography (server workload), 30% in machine learning, 22% in HPC (simulation stuff).
Basically EPYC stuff. And Ryzen desktop is basically just EPYC die combined with mainstream I/O die. So, his leak is actually accurate. EPYC is where the money goes after all.
If Mlid said Desktop and your speaking of Server, and now the uptic is like 6-10% he was wrong. Avx512 is what made chip better but desktop rarely uses it so no the leak was highly misleading. The Amd fanboys took the leak to heart and talked down the 14900k based on this copium. Because of Mlid wrong information I almost didnt buy the 7800x3d on offer. It turned out this was the best purchase since the new cpus sucks without overclocking. If Mlid really new what he was talking about he would know that its the Soc die using like 17w and core ccx uses 55w. I found this out by looking at Hwinfo64 after buying the 7800x3d. So even with an marginal improvement on a ccx that wouldnt help much because of the improvements would only apply to the 55w, and IMC we knew would stay the same since its on the SoC die. To me it looks like Mlid doesnt use any of the cpus he is talking about at all.
Really hope this comes out as a mini pc variant or socketed. Like others have been saying though - why have they not done this before now? What is the business strategy?
Anyone’s guess. Probably they haven’t done it because they haven’t needed to. with intel making better gpus and Qualcomm trying, they need to now to get ahead
I am looking forward a mini PC based on that Strix Allo. A mini PC with a large silent fan that will outperform a lot the previous mini PCs. Meantime i am enjoying my 7840HS based mini PC.
No, gpu does not need to sit near cpu to be fast. Wide memory interface is sorely needed since GPU shares memory bandwidth with CPU.
@@vladimirljubopytnov5193 when I said this I meant that by them being so close, they have a unified memory interface, which allows them to be much faster and more efficient
I judged you too fast by your face, you ate actually knowing a lot.
What’s wrong with my face 😂
@@SiliconSteak ahah nothing, just more clean than general techtubers :p no offense bro !!
@@nab-v1w haha none taken. Was just curious :)
Camera and lighting is kinda weird. makes me think it has a 50% chance of being an AI face. I do appreciate the video content, I am not sure if its because you are sick, but I would like to see you write a script for the videos, refreshing takes on things. Maybe do a recording without the face came then run the audio through descrypt and work the script from there then record the the face cam then overlay it on the material recording from the first take.
lol I look like a AI? I was having a hard time with this one because I was sick/drugged on NyQuil (I definitely drink more than 1 serving) as for the lighting, I thought that was 1 of the only good things I had…. I film this with my iPhone so I know it won’t always look the best 🤷♂️
Appreciate the advice, I will start trying some of this
Ryzen 4070 is Canon :0
Is it for desktop?
No. Well, they will likely put it in mini pcs. like I said in the vid, because they are using a 256bit bus, you can’t use a reg motherboard. The ram will be soldered on this one
>Swan song to mobile gamers
swan song (noun) -- a farewell appearance or final act or pronouncement (Merriam-Webster)
Do you believe that AMD is going to leave mobile space or something?
Also Strix Halo is fun, but MI-300 was a much more interesting product even if it's not for mortals, just because of how complex it is.
P.S. people pretend like RDNA 2 to RDNA 3 jump per core is weak and that's somehow bad. Sure, But if you look for example 3060 (3584 cores) vs 1080ti (3584 CUDA cores) they often have similar performance (obviously discounting ray tracing). That's just the feature of GPUs, very weak cores that manufacturers pack more and more of
I def used that wrong then lol
29:52 ROFL not really a fair shot there, and you should know better. He bases his predictions on the information he is getting from his contacts and that was 3 years before release like, it not possible that AMD can't make changes in the interim and then not tell him right?
Also included one of his recent predictions. Nothing wrong with reporting leaks or making predictions, just gets annoying when he never admits when he is wrong. Seems to have a lot of pride going on
But yes, you are totally right, skus and specs change internally without word getting out.
ddr speed equals bits even it is quad time mhz real ramclock is 1000mhz times 8 this is why transfer per second equals mhz
because of 1 bit per mhz
this is a ram standard since sram was invented
therefore 256bit dual channel +- 240gb/s depends on cpu clockspeed and latency
Igp needs inf cache to reduce latency from socket cpu if later am6 will have also 256bit Si and camm2 ddr 6 at 12gt/s
Now with strix halo the biggest apu ever we will see if the design works i think it will
If am6 natively supports 256bit memory, I’d be very surprised
get ublock origin bro
Oki
Strix is faster than PS5. No point to buy console
Eh casuals will still like the simplicity of consoles. I prefer PC myself so I get where you are coming from
click bait title, strix halo will be mediocre. the best variant uses way too much power for a handheld and performs worse than a 4070laptop. the low end variant will be ~10% better with a gtx 1060 which isnt good enough to play modern videogames and will really suck a handheld dry of power in no time, so you wont really be able to game off a charger in either case.
I don't see the point of strix halo at all. its not considerably cheaper than buying brand new hardware either. I built a 700$ 14th gen intel 4060 ti desktop pc with 32gb of ram last year that outperforms strix halo's best variant.
Its really just a novelty toy at this point. A worth while apu is still 5 years out minimum.
In the vid I was talking about how it wouldn’t make sense if it used too much power, and I estimated it to be slower than the 4070M.
The point is to get more gpu perf in a smaller space with less of a power envelope. This has always been the point of APUs. If it doesn’t do this, it’s a flop IMO
Also, I appreciate your comment. This channel is all about creating an active discussion. If I’m wrong tell me and we can talk about it, I am not bringing an ego into this. I’m just a nerd who loves new technology
@@SiliconSteak Yeah, i agree with your talking points, just not the title. but i understand it gets views. also why i commented critically, alot of the time people engage more
@@topelite666 exactly, also I mean. I think that if strix halo did succeed, maybe it could be the beginning of the end. Eventually we won’t need more performance, we will just want more power efficiency, and apus may give us that
bro what mlid said was all during development of zen 5, you never know what the real performance of the final product will be, maybe they ran into problems in development and mitigating those issues lost all the performance or something, and apparently they pushed out an unfinished microcode or something, so in a few months we may se e 10 to 15% better performance
This is my real take, he came up with a lot of different excuses why zen 5 was a flop. He is coping hard rn, i tuned into his recent vids and his on his most recent broken silicon he was bragging about how good zen 5 is. He has the need to be right and when he is wrong instead of admitting it, he makes excuses. I still like his content, that's just the way i see it
@@SiliconSteak His posts are also extremely negative against Intel. While he can justify it personally, his anti-intel narrative gets annoying real fast when you're just there for some news. He also has the need to be right like you said, therefore he keeps bashing ARC, since they somehow keep defying his prophecy of ARC going to disappear. I find it revolting when there might be like thousands of people working on it now, and he just comes out and gives out this negative perception on a genuine effort. They did very well for their first foray into discreet graphics, and for productivity they hit it out of the park. Wish they prove him and the naysayers wrong and stick it out till the end.
@@SiliconSteak In Hardware Unboxed's most recent podcast 3 days ago, they talk about this. The chapter in the middle of the video titled "What role did leakers play".
YT watch ID: lUk7rfY5KQQ
Like Tim says there, if he just takes a more reportorial stance as opposed to trying to be a prophet, then he wouldn't even have to make any excuses.
Maybe he thinks his channel becoming prominent only because of the accuracy of his "Leaks" are but most people are just there for the discussions & information [I mean accuracy isn't the be all and end all for unreleased products anyways].
Which you are doing very well. Maybe your channel might become bigger than his one day.
@@SiliconSteak He has a ton of copium going on right now with Zen 5. He was wrong and has been wrong about a lot of things. Only problem is he doesn't admit it when hes wrong.
@@Tom-bt8eg do you really think AMD is that far of during their development?
It was obvious from the beginning that improvements will be rather small (no significant node change, few changes to the overall design, most notably still using the long overdue infinity fabric).
If the product wasn’t ready, they wouldn’t launch it, they knew exactly what they’re delivering.
MLID has an extremely pro AMD community and positive leaks about AMD are received way better than positive leaks about Intel or NVIDIA.
He’s telling his community what they want to hear.
When you look at leaks, there is a tendency that AMD improvements are usually way below what leaks promised while Intel improvements are usually better than expected (see Rocket Lake, Alder Lake and Raptor Lake).
Then, AMD usually manipulates their IPC improvement claims to be better than they’re in reality.
But that doesn’t bring much clicks.
About MLID: there are statistics on his accuracy on Reddit. He often deletes videos that didn’t age well and in total all his leaks that he’s >80% sure about have an average accuracy of around 50%.
Note that this doesn’t differentiate in how exact his leaks are.
Most of his leaks which are true (in the 50%) are not really precise or pretty obvious like process nodes, naming, TDP etc. or rough estimations (like at least 10% better).
has your grandma not taught you not to cough freely? even coughing into the camera, so rude! do it into the inside of the elbow if you can't reach for a tissue. jesus, kids these days.
I forgot the camera was to my side, ( I was basically drunk on NyQuil for this vid) so I accidentally turned to my side where the camera was and coughed 🫥