People don't understand how INCREDIBLE ZEN4 is
ฝัง
- เผยแพร่เมื่อ 1 ต.ค. 2024
- Support me on Patreon: / coreteks
Buy a mug: teespring.com/...
Bookmark: coreteks.tech
My channel on Odysee: odysee.com/@co...
I now stream at:
/ coreteks_youtube
Follow me on Twitter: / coreteks
And Instagram: / hellocoreteks
Footage from various sources including official youtube channels from AMD, Intel, NVidia, Samsung, etc, as well as other creators are used for educational purposes, in a transformative manner. If you'd like to be credited please contact me
#AMD #zen4 #rdna3
AMD is executing its product development better than Intel and Nvidia, that’s why it continues to gain market share. Both Intel and Nvidia missed revenue expectations in the recent quarter, but AMD exceeded it.
Any bets on the next quarter.
AMD is leading the processor industry now, and Zen 4 is the industry-first processors using 5nm process. AI chip ban to China will impact Nvidia Q3 revenue by about $400 millions, but AMD said it does not have any material impact on its revenue forecast.
I think we all know what you meant @Charles Chen, but you should correct that "AND" to "AMD" of course so a non-native speaker may understand
Xilinx helped
I know I've had no regrets on my 5950x and 6900xt system... seems every driver update lately unlocks more performance... crazy gains in some games.
06:52 This is entirely wrong. L3 is shared among all cores on the CCD. That is not a bad thing. That's a good thing. The 4MB figure in that chart is just a number, indicating how much cache would be effectively available per core if all were doing completely different workloads. A multi-threaded load in the same process is using the same memory, where having that L3 shared is important. The big reason Zen 3 was faster than Zen 2 in gaming is because the total L3 amount was shared among all cores rather than split in two and used by two 4-core CCX's. That imposed a latency penalty any time a core needed data from the L3 cache in another CCX.
1) Apple A15 core has 62% higher IPC than Zen 3 ..... this means A15 will have 42% higher IPC than Zen 4
2) ARM Cortex X3 (new this year) has 29% higher IPC than Zen 3 ... means X3 will have 14% higher IPC than Zen 4
3) both Apple and license Cortex X3 cores are pure 64-bit only .... no 32-bit compatibility, saves transistors and simplifies development for max 64-bit performance
4) both Apple and X3 are monster wide cores: 6x ALU + 2x Branch Units = 8-wide scalar integer engine
5) Intel and AMD are just 5-wide speed demons (Zen 3 has 4x ALU + 1x Branch, Intel Golden Cove has 5x ALU with branch units shared on them)
6) L1i cache for Apple A15 is humongous 192 kB, Cortex X3 has 64 kB ....... AMD Zen 3 and Intel has 32 kB (necessary tax for high GHz)
7) Zen 3 is great because it's last Jim Keller"s design (basically it's improved K12 core, the famous taped out but canceled core)
8) In the video the idea of passive cooled 16-core 7950X is nonsense. 170W TDP / 230W PPT CPU to be passive cooled to fight Apple Studio? C'ommon. Wet dream of AMD fanboys. Even Apple cannot passive cool 16-core M1 Ultra (Apple core uses 5 W each in compare to 20 W for AMD and Intel). Technically impossible.
BTW do no forget that first SERVER CPU at 5nm TSMC is 64-core monolith AWS Graviton 3 based on license ARM Neoverse V1 core (derived from consumer core Cortex X1). Also Graviton 3 uses much advanced packaging than AMD - G3 uses 7 chiplets close tight together to minimize power consumption. Phoronix benchmarks shows higher performance than 64-core Zen 3 Epyc while having TDP 100W only (EPYC has 280W TDP). Pretty amazing for such an old V1 core from 2020. New server V2 core based on X3 will massacre any x86 server CPU.
www.phoronix.com/review/graviton3-amd-intel/9
semianalysis.substack.com/p/amazon-graviton-3-uses-chiplets-and
@@arm-power that’s a lot of IPC! Can it play my games? 🙃
@@conorabc Sure, ARM is great for gaming. More than 50% gaming revenue comes from mobile gaming platforms = ARM CPU. Also Nintendo Switch uses ARM Cortex A57 core (64-bit ARMv8 ISA).
MS and Sony want release mobile consoles too = ARM only. x86 is no go for mobile.
@@conorabc Also, GPU Immortalis released this year supports RAY-TRACING.
AMD Radeon 6000 with ray-tracing was released March 2021. Only one year later ARM GPU Immortalis got ray-tracing support too. Pretty cool stuff.
@@arm-power 1) No, it doesn't.
Not worth responding to the rest.
Gonna wait for the 3D cache versions. I already have a 5900X and its more than enough at the moment. Probably will hold off until Zen 5 next year. Will be looking more into RDNA3 or Lovelace.
That's a good idea, especially with DirectStorage, DDR5, and other stuff falling into place.
You think version,it will be only one 7700x3d /7800x3d because others don't have point because of high price snd lower productivity performance and gaming will be same amongst them
it's so tempting but even my 11900 will last me years to come. Will wait for sale or 3D one's. with clock speeds that high I don't have to OC anymore.
@@Mr11ESSE111 we most likely will get a 7800x3d
And don't forget Navi32 which will also launch 2023 Q1 and will be very interesting price performance wise.
People don't understand how INCREDIBLE Coretex is.
⭐🌄🌅🤔
texcore
I on the other hand find him very credible :-D
Grass is green, water is wet. Now give me my coretex points/likes.
people dont understand how INCREDIBLE the ddr5 rams and am5 motherboards prices is.
The Barry White of TH-cam tech does it again
Best romance tech on TH-cam.
Not a dry seat in the house.
Finally, someone else noticed that not all 13th gen Intel CPUs are based on the new Raptor Lake die! Good video, I agree with your points on Zen 4 pricing and the efficiency advantage for mobile.
I am your subscriber 🥳
14:30 valid point
I'm more interested in the Zen 4 APUs than the CPUs and eager to find out how RDNA2 desktop APUs perform paired with DDR5!
Probably similar performance to that of the Xbox series x
hopefully
RDNA 3?
RDNA 2 is already on the steam deck
@@ireallyreallyreallylikethisimg I am expecting at least RX 5500 XT / RX 590 performance.
@@cdoublejj It is, but the Deck is a mobile device with limitations. It's performance is likely just a small fraction of what we can expect from an RDNA2 APU on an AM5 desktop PC.
I think they've always known how much of an advantage they have from having GPU and CPU under the same roof, their fusion initiative is many many years old now. Their issue is that in purchasing ATI back in the day they so badly stretched their resources that they weren't able to actually bring those advantages to fruition. It too so long for the CPU dept to recover from its stumbles that the GPU side had basically been sucked dry and was itself in desperate need of investment then leaving it to Ryzen to shoulder the burden of carrying the GPU dept for a few years as AMD found its equilibrium and ultimately thrived.
Console contracts basically saved them from bankruptcy. I remember when it was announced I basically fistpumped in the air and predicted with relative accuracy that in 5 years it'll pay off, which it did with Ryzen coming out with Lisa Su at the helm.
I remember being ridiculed and called delusional in every comments section, forum post or discussion. I still keep in touch with some of them and don't hesitate to rub it in their faces that blind fanboyism (of Intel at the time) completely blinds their perspective of the bigger picture and how companies operate and how certain decisions can make a night and day difference. Having that permanent steady income stream parallel to PC components is that cushion that will absorb heavy blows and feed steady income giving them more money to invest aggressively into R&D when they couldn't before.
Fusion didn't work except for some nice low end APU like Llano, Richland and X Box & PS which saved the company.
There's basic problems:
1) performance gfx likes bandwidth but a CPU with support for quad channel RAM is expensive. So we suffer split address space with data copied over the PCIE bus.
2) power requirements make heat, it's easier & cheaper to cool 2 seperate 100W & 200W chips, than one 300W die.
@@egonieser I remember the relief, Intel had tried to force Itanic and build an absolute monopoly by the partner program locking AMD out of OEMs.
The general public didn't realise they were being offered inferior products or why.
@@RobBCactive TL;DR all I really should have typed is that Intel followed AMD into unification territory and did fine with it for years. From adopting UHD into their socket while AMD still needed external graphics, to experimenting with layers of hybrid storage during a period when RAM cost too much and HDDs were too slow. AMD is only now making integrated graphics a standard feature despite having a head start on APU design. You can skip everything else that I typed below.
Xbox One X sure drew plenty of power, something averaging about double what the Series X pulls. Approaching 150-180W package power, not total system power. Fused resources have been in development this entire time, features like BAR and peer-to-peer DMA and storage accelerators. And now that we are seeing 3D cache layering finally come out, and renewed interest in small HBM caches, it could soon be possible to have an APU that uses GDDR6 without the large penalties seen on those 4700S systems. All they need is enough onboard cache and aggressive prefetching to cover for those crippling latencies, that's why Radeons have Infinity Cache.
They took a forward-looking approach that required a crazy amount of changes in the ecosystem, and somehow survived long enough to start gaining some benefits from the approach. Their new compute cards work on the APU model, and the new Ryzen 7000s will all follow basic APU principles because the graphics are part of system IO now.
Llano was absurdly good if you undervolted, you could maintain 3GHz all-core load on a good A8 APU. Trinity and Bulldozer specifically were the things that almost killed the company, call it their Pentium 4 spiral of mistakes. And just as Pentium M led to the development of Core, AMD went back to the drawing board and tried to undo a platform that had no future. What we're seeing now are diminishing returns from higher wattage being weighed against the increased price of a larger die area. If people in certain market segments pay for the larger die, AMD can deliver 40-50% more performance without increasing wattage. Intel is still scaling in a more linear way at the high end, but their performance drops like a rock going into lower wattage. The newly launched U series proves that everything is the same fundamental technology and Intel *can't* optimize better for low wattage. Question now is whether Ryzen 7000 scales enough at the high end, not the low end. For anything below the most extreme performance segments, APUs will probably take over just as full-tour SoCs are now the standard for mobile devices. Consolidation is always a win for the system integrator, fewer external lanes and traces.
@@passintogracegoldenyearnin6310 The Intel iGPU were not the Fusion vision, some media encode/decode blocks and weak gfx with added dGPU wasn't the Fusion idea.
It was unified address space for gfx, yes;. but it was offloaded GPU compute. Intel added evermore AVX instructions instead and failed to scale their iGPU with improved GPU architecture. Even the Quicksync success became less important as quality requirements changed.
The only Intel APU are those Iris with the expensive EDRAM caches.
2:30 yes, support for OSes that aren't windows, open source drivers , etc
yeah amazing. but I will wait for my mini pc with an APU with RDNA2/3 and DDR5. Like you said, 45-65W is yet to be revealed.
I will demystify the slide at 6m29s which you see from the Retired Engineer.
ROB = reorder buffer , this has to do with out of order execution (so executing based on availability of data, not in the sequence of the program, this to decrease data-latencies/hazards)
BTB=branch target buffer , it has to do with branch prediction.
That slide bascially shows you why the 'IPC' increases.
ROB sizes of competitors:
- Apple A14/M1 .... 620 entry
- Intel Alder Lake ... 512 entry
- ARM Cortex X3 ... 320 entry (same as Zen4)
IPC = GB5 score per GHz:
- Apple A14/M1 .... 42% higher than Zen 4
- Intel Alder Lake ... 3% lower than Zen 4
- ARM Cortex X3 ... 14% higher than Zen 4
- ARM has support for 2048-bit vectors (SVE2 SIMD)
- both Apple and X3 are 64-bit only (no single transistor wasted for old 32-bit garbage)
1st 5nm consumer CPU was Apple A14 (ARMv8 ISA)
1st 5nm server CPU was 64-core Graviton 3 (ARMv8 ISA + SVE SIMD, license Neoverse V1 core, similar to Cortex X1 core)
1st monolithic 128-core was Ampere Altra Max at 7nm TSMC (ARMv8 ISA, license Neoverse N1 core, similar to Cortex A76 core)
1st super-computer CPU-only which was beating GPU-based super-computers: ARM based Japanese Fugaku (52-core Fujitsu A64FX CPU, ARMv8 ISA + 2x512-bit SVE, 7nm TSMC)
--------------------------------------------
Make your own conclusion.
@@arm-power Ultimately it is all about the performance for the much used applications. If anything we should be happy that with all the progress which AMD made it pushed Intel to finally make some progress again too. However it pans out, there is competition and the prices can decrease. Win for us! Though right now I am not in the market for Intel for the simple reason that Intel does not support Linux yet with Alder Lake, Intel made a mess by not using acpi the way it is intended to be used. I am sure that that will get solved before I buy a new CPU though. :)
I have a 3700X and 5700 XT, first I want to upgrade the graphics card. I consider buying a 5800X3D if the price comes down (like €200 in the next 1-2 years), if not then it gradually becomes time for a new platform and I will choose whatever works best between Intel and AMD. Zen4 is not interesting for me anyway, AM5 is too new, DDR5 is too new, PCIe5 drives up the cost. Better wait at least one year. As much as I would like ARM to become a competitor, gamers are stuck to X86 in the next few years.
@@arm-power By the way, I would love Intel and AMD dumping superfluous old instruction sets. That really holds X86 back, you are 100% correct about that!
@@peterjansen4826 x86 cannot drop old 16-bit and 32-bit instructions. x86 uses variable encoding where each instruction can be 1 to 15 bytes long. All (16/32/64-bit) uses same variable CISC encoding, backward compatible.
ARM is RISC with fixed 4 byte encoded. 32-bit AArch32 (ARMv7) is backward compatible to ARMv5 and v6.
64-bit AArch64 (ARMv8 / 9) is clean sheet design from scratch in 2010 and is not backward compatible (uses different encoding). For backward compatibility there must be separate 32-bit decoder. Different names and different number of instructions, take a look at page 15 here:file:///C:/Users/User/Downloads/Arm_Cortex-X1_Core_Software_Optimization_Guide-1.pdf
So ARM by abandoning 32-bit saved a lot of transistors for second decoder.
64-bit AArch64 has very advanced and flexible memory model with fencing similar to DEC Alpha (x86 uses old simpler mem model). Programmer in AArch64 can access BTB and cache status. All this magic helps to extract IPC with as low resources as possible (it's more difficult for assembly programmer or compiler though).
ARM is moving forward faster because:
- 4 different CPU designers (ARM, Apple, Nuvia/Qualcomm, Ampere)
- MANY chip manufacturers via license
- new clean sheet ISA design
- anyone can buy license for ISA or CPU
- same is RISC-V royalty free / open (x86 is closed proprietary now, but back in 90 there were 7 designers because IBM forced Intel so)
It'll be nice having a good split between Intel and AMD again, when they're both fighting each other, we as the consumers, win.
I will say that AM5 is looking like the last traditional x86 socket by AMD, the next one looks like AMD's working on a RISC solution with a dedicated software stack to support excellent x86 emulation. Intel is probably doing the same if they're smart.
RISC or ARM. I'm leaning towards an ARM hybrid. I'm guessing partnerships with Google, Microsoft and Samsung
Are there any concrete indications of that? Genuinely curious.
@@chbrules RISC and RISC-V are not the same thing. RISC-V is one of several architectures that implements the Reduced Instruction Set Architecture design philosophy. ARM is a company that designs processors including ARM-v8, which is another architecture built on RISC principles. But, here's the interesting part, Zen and Core are both RISC designs inside the CPU core, but they use large decoders on the uncore to support the whole x86-64 ISA.
@@chbrules Yes, I said that at the end of my comment.
I kinda need to see reviews and thorough testing before I buy that it's so amazing..
NDAs lift about four days before launch, I believe, so expect a deluge just before the 27th. ;)
If this is true then it's amazing. But I'm going to wait for a third party benchmarks before making any conclusion. All of the big 3 pulled so much crap in the past I don't believe any of them.
Hey Coretex, since when did you get a new mic? sounds good, less audible gain, still punchy low end & plenty of clarity.
Could use a pop guard though.
I don't understand why he always overprocesses his voice
@@genebeidl4011 hearing him say Ps and Bs always grate my ears
Always thought this guy is unnatural to listen to. It's like he speaks very quiet, very close to the mic at high gain to get a low and growly quality to his voice.
I bet when he speaks normally it sounds a lot different.
Could be wrong but just how it sounds to me.
@@genebeidl4011 Won't help when you're deepthroating the mic lmao
Zen 4's biggest competitor is their own 5000 series. people with older amd boards like myself not gonna go to x670 and DDR5. we'll just go with 5000 chips selling at big discounted prices from MSRP. i'll wait for Zen 5
I am looking forward to the upcoming dragon range of mobile products. I need a new notebook, so hopefully, there will be at least 16 OLED 1440p 144 Hz+ with Ryzen and Radeon 7000 with an AMD Advantage program.
I recently upgraded my tired old 2600 on X370 to a 5700X on B550 and i'm very happy with the uplift. As promising as the performance of Zen4 looks the cost of the overall platform is a little prohibitive for me. Maybe next year if the price of the motherboards and DDR5 memory come down it will be more appealing for the average PC mortal.
AMD is following up with Zen 4D quite soon actually and I think it will definitively crush Intel. Of course RaptorLake will likely close the gap quite a bit, but with much bigger power envelope and much bigger die making it hard for Intel to sell it low enough prices to compete with AMD.
I mean it's not like both parties don't have a plan, but AMD is simply executing it better. However it's not Intel's architecture that performing bad it's their process technology and while both do affect one another (it's not like you can just print out a CPU with two different nodes just by saying so) I think that if Intel had any success with their process technologies to get to even TSMC 7nm level or slightly better they would give AMD hard time.
I wouldn't expect this to crush Intel. Most likely they'll do like they did previously and pay systems integrators not to use AMD chips or find some other illegal business tactic to steal AMD's wind like they did the last couple times.
i am so glad AMD focusses more on efficiency than pushing their components to the power limit!
There were some leaks stating to reach advertised speeds the new ryzen chips were pulling 220-230 watts and reaching 95c temps and throttling back as a result and this was on like a 320-360mm water aio cooler.
Same test had Intel at only 83C temps.
Would be funny to see AMD's new chips now become the power hog space heaters as compared to Intel and see how the AMD fanboys reacted then!
@@billwiley7216 yes intel has efficiency cores but Ryzen only has the normal cores
@@billwiley7216 Perhaps, but historically only one company has had to pay other companies not to use competing chips.
Zen x3d will be insane. A0 to A0 silicon, it's showing 30% better than Zen 4 right now. That's so fast that our next gen gpus might not be enough
@@AntiFurry927 People dont forget about it. It just presents a massive cost to benefit analysis engineering wise, and is very conditional for it to help. It requires alot of very advanced prediction of execution to fully take advantage of, and its very costly, not only in terms of production, but of die space. it is an energy sap, as it is constantly refreshed even if not in use. It takes careful design considerations to decide on the amount too.
It's part of why double stacking cache is such a nasty business and why AMD doesn't allow OC-ing of the 3d chips, because it makes it very hot, which not only causes issues with performance and error occurrence, but it also stresses the structure and can cause failures. but it has huge performance implications if it is able to be taken advantage of due to the extremely low latency.
@@AntiFurry927 Unless Intel has managed to get E cores much better, the fact that they have 24c on i9 and 14 on i5 will not change things. One Zen 4 core will be most likely as fast as two E cores and in AVX512 code it's probably even more. Also the insane power draw of Raptor Lake in all core productivity workloads starts to have an effect. Here electricity is starting to be close to $1 KWh and if CPU takes 100W more and it's running on close to 100% five hours per day, then it's one dollar more electricity every other day. That adds up to $180 per year.
AMD might lose on mid range if Intel can price their CPUs well. But I think it's going to be a tough job as the yields on these monolithic dies is going to be much less than TSMC 5nm chiplets. Though they own the fab so it helps, but I doubt they can drop the pricing. Adding more cache increases the die even further as does the extra E cores.
Though I absolutely love Intel going with E and P cores. It will the best way to tackle Amdahl's law, and optimize how much performance you get per watt and per die area. In a decade I think there will be 3 different cores, one where they just do everything to get the single thread speed as high as possible for running threads that are the bottleneck, then common fast cores where energy efficiency is somewhat taken into account, and then the highest performance per watt cores. And CPUs could be like 2+8+64.
"That's so fast that our next gen gpus might not be enough"
Current-gen GPUs aren't enough for current-gen CPUs (at 1440p and 4K in most AAA games), but this problem is fixing itself. GPUs gain performance faster than CPUs. Next-gen GPUs should be about twice as fast as current-gen, while the CPUs are less than 50% faster.
Saying that next gen GPUs might not be enough for a next-gen CPU is just a weird thing to say. It's much less likely for that to be the case then than it is now.
@@juzujuzu4555 "in AVX512 code it's probably even more"
I think you're hedging your bets a bit too much there. Gracemont does not support AVX512. Obviously Zen 4 is going to be more than twice as fast as E-cores at AVX512: It's going to be infinitely faster, because AVX512 can't run on E cores.
Usain Bolt can probably run faster than a brick.
@@r100curtaincall The heat issue is something AMD has debunked time and time again. The lower clocks are due to a _voltage_ problem.
The 5800X3D used a beta-test version of V-cache that couldn't handle normal levels of voltage. Expect the 7000X3D chips to fix this.
this is my fav channel for cpu, gpu and related architecture/tech info.
and moores law is dead, but the accent here is way sexier.
How does it feel being hands down the greatest silicon tech tuber?
I've been religiously watching your videos for the last few years and your presence has softened the blow of losing Jim from AdoredTV.
I miss Jim :( ALRIGHT GUYS HOWS IT GOING?
zen4+ rdna3 on one chip laptop with total power of 90w would be crazy like a good 6 core apu!
The laptop chips are gonna DESTROY intel’s offerings due to the power efficiency
Sure just destroy!
It's going to be such a massacre no one could justify an Intel laptop 💻
heard that for the last 3 years...
How long have you been repeating this for? Intel laptop offering are fast... but power hungry just because something uses less power doesn't make it faster lol.
rip and tear until it is done!
The cold reaction is due to 2 factors, that are actually 1 single factor: most people don't have money for PCs anymore. The covid hit many of us and we lost a lot, and this is not just a CPU upgrade, you'll have to change CPU, motherboard, and RAM, and DDR5 RAM is expensive. And then honestly the GPU situation with the mining and all was just disgusting in the most literal sense of the term, it really made PC gaming as a whole look very bad.
Yep, if they don't release a backport to AM4 (even with a performance hit) Zen 4 will sell way less than Zen 3 imo.
Crypto destroyed the image of the gaming industry in the eyes of actual rec gamers. who spent ALOT of money as a hobby. We didnt give a shit about "investing" in junk bonds. We already had income, to burn. When they turned their backs on us, we found other avenues to entertain ourselves. Oh well. DJI is doing smashingly these days ;)
Pretty much this. These products might not even exist to me. Even if I have the money to buy then, why would I at those idiotic prices ?
Alternatively the insane demand for Ryzen 5000 and RTX 3000 means a lot of people have only recently upgraded. I built my PC in late 2020 and snagged my 3070, I won't even consider a new build until 2025. I just follow tech news for the fun of it but 2020 also saw the release of next gen consoles which were extremely competitive, essentially we're between big generational leaps right now for consumers.
@@gorana.37 To be fair 700$ for the absolute top of the line CPU is a fair deal. Intel "extreme" processors always used to be priced at 999$.
and not to forget the integrated gpu, hopefully there will be mobos equipped with multiple video outputs and/or usb4 = need 3/4 monitors but no demanding video card, if in need will have to go for a single slot amd radeon pro wx 2100 or similar...or intels A40 if it ever gets out, not everybodys priority are games....
AMD's execution is pretty damn amazing at this point!
That's what a boat load of money does for a company.
@@badass6300 false, Intel back in 2015 had way more money than AMD, better marketing, and more engineers but their management got too complacent.
It's not money that you need for good execution but great leadership.
Lisa Su came in while AMD had 3.57 Billion dollars in debt(was worth 2.3B) now they are worth 129 billion dollars. For reference Intel is currently worth 128B.
@@Thesinistereyes1 Dude as important as the leadership is, it's the engineers that made it happen, Jim Keller at the front of Zen. Jim Keller always makes great products. He was also the person who was leading K7 and K8 back in the 2000s when AMD was again very successful.
Well intel sells 70x times more chips than AMD, the stock market is so fake nowadays it's unreal. AMD from 2016-2021 sold fewer chips than intel sells in 1 quarter.
@@badass6300 Jim keller also Developed Alder Lake S for intel but if they had good leadership this should have been out in 2019 but instead they refreshed the same arch and node again.
You realize why AMD is worth more because they acquired big companies like Xilinx. AMD sell more chips to Tesla, all consoles except switch, nearly all the new server upgrades and new server builds are AMD. Most Intel sells are cheap embedded solutions and prebuilts from companies like Dell. Aside from that their sales had gotten weaker and weaker even with good engineering on their new cores.
@@Thesinistereyes1 Jim Keller wasn't in intel in 2019 or had just arrived at intel in 2019.
When I say Intel sells more chips than AMD I mean all chips from both intel and AMD.
The APU announcements are always my favorite. CPUs are still interesting but the APUs are where its at. I always want to know if indiscrete can beat out discrete and then people can just focus their budget on one AiO compute system that is portable.
My last 5 builds have all been APU based due to GPU costs, even now with GPU prices being somewhat normal, I would still go the APU route. The 3400G that im currently using is fine for the games I play, even at 1440p, and my sons 5700G based systems are quite simply amazing, though one of my sons did buy a 6500XT as a placeholder GPU.
Im hoping that desktop APUs with ZEN4 and RDNA are a major step forward in the market.
@@VeggieManUK Or you could just join the mining bandwagon and not have to worry about all of this
I always want to know if indiscrete can beat out discrete
I don't understand how anyone could not understand how great Zen 4 is. If anyone is watching a seminar or presentation on microprocessors, you'd imagine they have to have some comprehension of the technology.
It's simple, Zen 4 is using 6,000MHz DDR5, It has great levels of cache, it has massive frequency jumps ie: 5800X @ 4.8 GHz to 7700X @ 5.3 GHz. Zen 4 also has much lower TDP and super stable TJMax exceptions vs Intel's 12th and 13th gen processor lineup and SO MUCH MORE.
the most improvement is coming from the die size , more cpu for less space. sounds incredible. The gold coating to make the indium stick better on the inner IHS hopefully actually makes a significant difference.
I am getting the sense that everyone is acting like gaming is the only metric ? Multi core already destroyed alder lake, and even if gaming is your main focus as you said, you surely focus on other things occasionally too? Even if it’s just multitasking or streaming. Who desperately needs those extra frames and isn’t willing to take all the other benefits in exchange.
Also cheaper chips will come. I don’t understand people acting like amd has betrayed us for not pricing things extremely low forever. DO YOU WANT THEM TO BE A COMPETITOR LONG TERM??? Or just for a few years? Because Intel still owns 70% of the market and they WILL get their crap together eventually and they WILL get back to paying their way to success the moment they have a lead across the board and then AMD will be stomped out fairly quickly if they haven’t filled their coffers and capitalized on this opportunity. I want comptetitoon to be restored for good, if “you” do too we should stop acting like AMD has stabbed us in the back and intel is our friend now because they have so much supply available they can instantly offer cheaper sku’s , but people really need to consider their priorities and ask why they are willing to give up long term competition to save a few dollars.
It should be said, 4K gaming. None of this is even remotely necessary for 1080p gaming. I'm using a 5600x/5700xt and I am crushing 1080p gaming with no desire to upgrade anything at all.
@@doublecrossedswine112 yep that’s a fair point and also why I’m saying the 150 vs 170 fps or whatever the case may be at 1080p becomes a moot point
Exactly what I needed, so interested in the announcement but didn't have time to analyse it
6 cores for 300$ still sounds very expensive to me I'll be here having fun with Ryzen 5600s for 140$ Ryzen 7 5700x for 200$ , i3s 10100f for 60$ and i5 10400 for 112$ at Newegg and Amazon. You don't really need the latest stuff to game or work .
"omg new generation is more expensive than old stuff"
Duh
"Omg amd is becoming Intel $300 for a 6 core is a ripoff"
How quickly we forget 10 years of quad core Intels being $330 until 2015. Which is $400 or more inflation adjusted
If the gaming benchmarks against 12900K are accurate, there just isn't any reason to sell it cheaper. It's new platform and cheaper CPU is not going to get people with small budget to buy it. Not that I like the price but I understand why it is what it is. Also I think most of those CPUs are actually working 8 core CPUs that just have some cores that are not clocking that well, or that need higher voltage, so AMD decided that instead of lowering the 8 core CPU clocks they would just cut down those chiplets.
$249 for the six core, $379 for the eight core, $519 for the twelve core, and $699 for the sixteen core CPUs would have created much better overall feeling. But they will sell ridiculous amounts of those six core CPUs anyway, and dropping the price $50 would have dropped the profits probably by 35% per unit.
@@l0s78 yeah I'm aware but amd is doing the same as Intel now we should trust the product and the value it brings not become fanboys of either company I have used intel and ryzen CPUs in the past I always pick the one who has the best value.
@@NemusDark no we shouldn't because AMD chip were almost always dropping in price a lot faster than Intel's. The only exception happened last year because of global shortage issues and demand on TSMC for playstation5's cutting in
The 5600 was $300 for a long time, and successful. $140 will take awhile, or a very strong Intel showing.
I think they are bad at marketing. That bold guy there at AMD should either step up his fuckin game or Lisa needs to pay for people better qualified for the job.
I stayed up until 2am to watch the live stream and even though I was blown away by the performance figures of the 7600X beating the 12900K in gaming, overall I was disappointed at the show. And this has happened many times over the past few AMD keynotes. Just overall underwhelming presentation.
For example they could have had some rigs with people playing games on stage or doing some production workloads or demonstrate AVX512 or the build in RDNA 2 graphics or USB4.
They could've demonstrated a Gen 5 NVMe SSD working on a motherboard. Or even RAID 0 it.
They could've showed perhaps a game using not only Direct Storage but direct storage access or whatever they call it.
They could've shown SAM working.
A real RDNA 3 demo.
etc. etc.
This would have made the show so much more exciting, as opposed to it being the opposite, which was even boring at some point. And I'm just a bloke on the internet. There are people being paid to do this stuff and they're sitting on their asses! smh
CPU speed doesn't matter anymore. guess what? Ryzen 8000 will be EVEN BETTER and 9000 BETTER THAN THAT, and so on. CPU speed isn't the bottleneck anymore
Going by the title of your video, you don't understand either.
Nor did you understand for Zen3
Nor did you understand for Zen2.
In fact you have not bothered your backside to learn ANYTHING about the Ryzen CPUs since the 3950X came out in 2019.
F1 2022 7600X was much faster than 12900KS and they used 6950XT the fastest Gaming GPU on the Market only Raytracing is faster on NV
I'm skipping This Generation in favor of Zen 5 Even though I''m sure you're right but The news is that Zen 5 is going to redo literally the entire architexture and redesign it from scratch all over again and so Zen 4 is great but Zen 5 is going to kick its lily white ass all over town. So I'm like I'll wait.
I should stop watching videos about this--I just built a new box with a 3700x a bit over a year ago, and can't justify upgrading again so soon...if ever. That's ok, the one it replaced was nine years old at the time. I hate being poor.
5.7ghz with an IHS, I wonder if PBO could take it higher with some extra work done to cooling, like something extreme like direct die+liquid metal
I am going to wait for the 3D V Cache version. Will be a mega upgrade for me coming from a 6700K. I hope by that time there will be decent real PCI E 5 PSU's from e.g. Seasonic/Corsair etc
You won't gain much with a 3d version. Focus on gpu bottleneck if you want to upgrade with budget limit. Don't sacrafice gpu for 3d version of the cpu!
@@WimiBussard he will gain a ton with 3d version lol
@@WimiBussard How do you know?
3D V cache is already looking really good for gaming.
@gewel @Dawi I am ofc talking about having a budget. I might be wrong, but I assume, that if you have to choose between
option 1: 5800X (EUR 300) + rtx3060ti/3070 (EUR 480/560) = EUR 780-860
option 2: 5800X3d (EUR 455) + rtx3060 (EUR 380) = EUR 835
(today's prices in Germany)
I assume option 1 will give you more performance. If you don't have to choose, then ofc you can buy the best of the best.
@@WimiBussard Since I'm putting aside 1K/year for my PC hobby, and I was not able to buy a new system during the last 2 years, I have enough money to go high end both for CPU and GPU.
If you have 2 products that are borderline the same but 1 of the 2 has vastly better power efficiency then it becomes a no brainer.
For years the power efficiency argument boiled down to heat output really, most people werent hurting their wallet by having less efficient parts but now due to a certain eastern european debacle the entirety of Europe is going down the shitter in terms of electricity prices thereby making power efficiency that much more needed.
From the leaks regarding RDNA3 it seems AMD is gonna have massive wins on both sides in terms of performance per watt.
Your comment of the 7900X probably not being much faster than the 7700X at gaming.. it seems that way until you look at the cache. It has nearly double which will certainly raise it's performance in comparison.
in 5900X vs 5800X there's almost no difference in games. Same will happen with new gen
As an audio engineer and voice actor... The idea of a passively cooled silent 7950x small form factor pc gave me the chills.
The 7950x draws roughly 170W and can sustain up to 230W. That's about as much as Intel's 12900k. Passiv cooling is a dream and nothing more.
Zen 4 is still impressive though.
Or just get an M1 Max Mac Studio...
@@DaCarnival But you need a second gaming machine.
@@Plasma_Mobile Sure - having a work machine and a play machine isn't unreasonable.
intel better for creative professional due to BigLittle cores CPU ..
The 12900K and 7950X used the same DDR5 in the testing showed only the 5950X used DDR4, it says so in the notes.
13:40
You make it sound like they did this comparison with the 12900K using DDR4 which is absolutely not the case.
Ryzen 7000 L3 Cache is most likely shareable, at least between 8 cores. 2x32 MB, So each 8 core have access to 32 MB of cache, similiar to intel.
Yeah, he made a really bad read there. He also didn't call out that this slice-for-slice comparison leaves out the much smaller E-cores. A better comparison is total die area of Intel's single complete package vs all three dies in AMD's chiplet package. Cheaper, more reliable small chiplets in exchange for the extra packaging complexity. And AMD both consumes less power AND spreads it to three different dies instead of concentrating it inside a single die. Completely different strategies. In heavy multithreading all of Intel's cores are on one package, while in the worst case one of AMD's cores may need to go through the IO die to get data from the other CCD's cache. This is why there's been talk about whether the L3 tile may become standard on Zen 5, and maybe an L3 tile could even span across two compute dies and join them together. This was also discussed for the largest of the upcoming Radeons. Imagine moving from 2x32MB for 16 cores, to 96MB or more unified across 24 cores. Two CCDs, one L3 tile, and performance approaching a 28-core Xeon W in as little as 1/3 the watts.
@@passintogracegoldenyearnin6310 I am not familiar with E-cores from Intel, are though on the same die, so Intel uses one die, I can understand your argument, but you may have worded it a bit confusion... what did you mean with "leaves out the much smaller e-cores" ? Like not mentioning that they are included in Intel's package/die... Hmm well you have an interesting point there ! =D But I am for spreading of heat... so AMD's solution seems interesting. Though intel probably packages it in a smaller surface area, not sure though.
@@passintogracegoldenyearnin6310 By the way, the shared L3 cache of intel, reminds me of IBM's special L1 cache on it's latest mainframe/Z chips/processors have you seen it ? There CPU's can share each others L1 caches ! Quite amazing ! I think Intel may have had help from IBM, designing it's L3 cache to function in a similiar fashion, or maybe it was IBM that learned a trick or two from Intel's L3 cache design lol.
@@skybuck2000 On almost every traditional CPU, the L1 and L2 caches are exclusive per core and the L3 cache is the "unifying cache". Up through Zen 2, on desktop there were two 4-core CCDs with 16MB L3 per each, inside one CCX. Zen 3 changed to a single 8-core CCD with 32MB shared L3. The 7950X will be two separate CCDs with eight cores and 32MB L3 (for a total of 16 cores and 64MB L3), plus 1MB L2 per core instead of 512KB L2 (A total of 80MB L2 + L3 cache).
Raptor Lake's best part the 13900K is expected to have 24 cores, 8 P-core and 16 E-core. It is a gigantic chip on their most advance process, needing 257 mm2 and being very expensive to make (bigger chip = higher chance of defect per chip). It is 25% larger than the 12900K but the extra eight E-cores only needed 13mm2 and the rest is more cache. It is believed that L2 cache increases from 14MB to 32MB and L3 cache increases from 30MB to 36MB. On the 12900K each P-core had 1.25MB exclusive L2 and a 3MB 'local' L3 (with the rest of L3 shared on an internal network), plus each GROUP of four E-cores got 2MB shared L2 and a 3MB 'local' L3. On the 13900K there is now 2MB L2 per P-core, 4MB shared L2 per cluster of 4 E-cores, and two extra slices of 3MB L3 for the two extra clusters of E-cores. 13900K has 68MB of cache vs 80MB of cache on 7950X and only 44MB on the 12900K.
AMD's comparison image was one P-core plus one 3MB L3 slice. They say they can fit one Zen 4 core and 4MB of L3 into half as much space. However Intel can seemingly fit four E-cores into the same space as one P-core, with L3 being the same size. Or one would say two E-cores, 2MB L2 and 1.5MB L3 in about as much space as a Zen 4 core with 1MB L2 and 4MB L3. With 24 cores, the 13900K is not as inefficient in space per core across the entire chip. It will still use more physical area and cost a lot to make, and Intel is already possibly LOSING money on the 12700 and 12900 chips. Their earnings and profits have almost collapsed compared to before, these chips are just too big and too expensive compared to how much they sell for. AMD's chiplets are much smaller and have far fewer defective tiles. And by selecting only good chiplets to put into a package, they can ensure something like 99.9% success rate in packaging. They are using 5nm and 6nm TSMC process, and the total manufacturing cost is still less than a 12900K manufactured on the Intel 7 process... 13900K will have a 50% higher defect rate for being 25% larger, and you get fewer chips per wafer. They'd need to charge $800-900 probably to justify the investment and manufacturing costs.
@@skybuck2000 IBM making a shared L1 would not surprise me. Their special high-density eDRAM is laid out vertically in the wafer instead of horizontally and yielded phenomenal cache density on only the 22nm node!
Again with the usual echo chamber in the comments section where people are heaping praise of a product that hasn't even launched yet. Why not wait til the reviews come out first and then give your feedback instead of over hyping something that might not meet your expectations😄
creepy sound and voice
12:48 I think the 7600X is still overpriced. When Zen 2 was launched, the 6 core model started out at $199 , the 3600X was $249 . Charging $299 for a 6 core 7600x is too much considering i5-12600KF is only $264 . In multithreaded workload, the Intel chip will beat Zen4 due to extra core count. Plus for Intel platform, you can go for the cheaper DDR4 boards, no such option for Zen 4. According to Steam Hardware Survey, Intel still command 70+% of the CPU percentage. AMD should aim for higher market share , not cash grab so soon.
Steam charts lmao
Unless 7600X has some dark magic, the price will be lowered after RPL releases, AMD will need to reevaluate their cpus position in the market at least until X3D versions arrives.
12600 won't beat the 7600 for sure lol
@@doggSMK Yea since the 12600 non k has no e core. I just hope their 6 core and 8 core offerings are cheaper.
Why do you think a 21 minutes and 17 seconds long sentence without any punctuation in a highly compressed format is a good idea?
Just fyi the phrase is "knocked it out of the park" - from baseball. Not "thrown".
TFW you realize that these companies develop the technology years before and deliberately release incremental increases in performance in order for you to think its a crazy leap/keep buying.
I've been thinking that since the Intel "you'll get 5% increase gen over gen and be happy about it" days and I developed a habit to only buy new hardware when I see a 50%+ improvement over what I have now.
You need a pop filter and or stay a bit further away from the microphone. too many p, ch, t and ss sounds jump out as too loud on my monitors.
I’m super curious about the XDNA platform, maybe it’s an ARM SoC with RDNA?
Salam brother
It's Xilinx with some other stuff
It’s just Xilinx IP rolled into AMD’s other portfolios
Yeah, that was weird. They showed it and never talked about it. wtf?
@@SNESChalmers147 sounds abt right, forgot about the acquisition
The incredible thing here... is how much AMD starts to look and act like Intel. They forgot how they started and what they stand for. We don't need another Intel, one of those is enough. I will pass. I just need a working decent procesor, I'm not building a NASA computer at home. Also, why are you kissing their arse so much? 😘
You make close to zero sense. What does AMD stand for?
If you just need a decent processor you can buy a processor a generation or two old and be fine.
If you have been waiting for 2+ years to get your gaming rig, you aren't really gaming are you?
Please god no whiney chipset fans on these motherboards
we have seen the footage of event, stop making videos if you dont have anything new to say. instead of rephrasing
your voice and pacing makes this video incredibly hard to watch
Sounds cool but I'll be excited over real benchmarks
You're right, because it's not even out yet and zero real world reviews.
You should wait for RPL (13900k) and then compare ZEN 4 ( 7950X ) , taking pricing and availability into consideration , i think 7950X will be retail at somewhat more inflated price .
@Based Madara nor this , nor that .
@Based Madara yes that is नीती ।
@Based Madara नेति-नेति and नीती, two different things .
Pat from Team Blue has to wait much longer for what he said is the intel Zen moment. LoL
Do I go for Intel or AMD for the most powerful gpu, price aside
No cooler, PC world confirmed with the AMD marketing guy. Way too much wattage for a cooler with any of these models
The voice seems so TTS, but sometimes doesn't it looks like judging the inflections in voice. Is it TTS or not, would anyone tell me?
Amd have been fantastic over the year's...making intel work hard
We do understand i can't go anywhere without seeing RIP INTEL HUEHUEHUE
21mins drooling over AMD marketing.
Cool.
Now that lenovo have z16 laptop with AMD apu I will definitely buy it when the zen4 comes to laptops.
Killer Chip! One Platform to Rule them all! AM5
Can't stand computer generated voices, it could be the greatest information to be heard but if you're not using a human voice you can shove it.
He's real
Temperature seems to be very high, and that is disapointing. Will keep my money till later information on the 3D cache units.
It's higher, but more powerful. Definitely not as bad as Intel.
I don't know what they mean when they it's more efficient when zen 4 consumes almost the same power as raptor lake loses in both single thread and multi threaded performance and runs at least 10 degrees C hotter practically throttling every now and then.
@@Yusufyusuf-lh3dw exactly what I have been saying it’s funny almost intel is still better in actual power of the cpu and why a 6 core cpu for $300 the my 12700k has 12 core 8 power and performs better then a 7700x I don’t get it currently amd overhyped up.
The AMD shilling will be amped up to uncharted terrritory.
I doubt it will change much tbh.
Then being quiet on the gpu at that show could be because they are waiting to see what Nvidia shows as I get the impression that both are playing games with each other on the specs and capabilities of their cards and it does look like AMD have something special, at least Nvidia likes to think so with how hard it looks like they are pushing their next gen gpu's, basically, this is the first time I've seen Nvidia feeling worried about what AMD has up their slaves with the new gpu and AMD might be playing on that by being cryptic and keeping Nvidia guessing.
What I am interested to know is the ray tracing performance of AMD's new gpu's, we seem to know nothing about that at the moment and being as that was the major weakness of the current gen gpu's, I would think AMD would put a lot of focus in that area to boost performance a lot.
So long as Minecraft looks amazing, who cares :)
That lady's voice is way deeper than I expected
If we can actually afford it this time around.
Why does this sound like a creepy pasta video? 💀
I am interested in 7600X vs 13400 in gaming, but probably not going AMD right at launch because new platforms always have problems. Also, AMD needs to make a maxed out APU, it is thier forte
In terms of maxed out APU, I am hoping the AM5 platform allows for this, but I think we may end up waiting for AM6 or whatever it's going to be called as I think AMDs 3D vCache is a prelude to memory being included on the substrate, they did it with VEGA afterall. All this drive for power efficiency is going in the right direction for this.
I would be suprised if we did not get a killer APU by 2026-28, 16c/32t minimum , its how I see AMD going, with decent if not top performance in 1440p gaming.
@@VeggieManUK What we need is quad channel memory. Because CPUs are always one generation and double the bit rate behind GPUs. System Memory Bandwidth is always the limiting factor for APUs. Dual Channel DDR5 most likely won't allow for anything above 16CUs, unless AMD sticks 512MB L3 cache(Infinity Cache) to the iGPU.
there sandbagging
There sandbagging?
Where sandbagging?
Here sandbagging?
Why is everything goated all of a sudden 😂
seems like Coretex sponsored by AMD
Your voice 😳😳😳😳😳
you and I are on the same page about Zen 4 :D
The voice is pretty AI Robert.
Man, I would sure love to undervolt amd's mobile gpu's and cpu's. Would love lower temps.
Sadly in this regard amd is a complete scum. Worse than scum because ngreedia allows mobile gpu tuning and intel used to at least allow undervolting, whereas amd never allowed it on mobile.
I agree with you on the 7600x and even the previous 5600x when it came out .
Methinks AMD is gonna go with IBM's cache approach - massive L2, maybe some type of virtual/logical L3 and HUGEASS physical L3/L4 stacked + possible offloaded to IO supercache or smth. AMD really digs into memory subsystem if you look clse enough and they're doing tremendous advances. Intel's caching system looks like an abomination compared to AMD's which is especially sad, cuz i always perceived Intel as a memory company firstly.
AMD is basically getting to the same process node as M1/M2 and it’s going to be interesting to compare efficiency between ARM and x86 (although AMD has a significantly newer architecture)
same nodes doesnt mean a thing, ARM cores are low power designed for mobile. Full fat X86 Zen4 cores do not have the same power constraints thus comparing the two architectures make no sense.
@@ssaini5028 when both are being used in mobile first designs and form factors, how does it not make sense to compare them? AMD Phoenix will be going directly against M1 or M2 Pro laptops
@@Ha-ue9kg AMD Phoenix Point will be based on Zen 4 yes, but on TSMCs 4nm in 2023
@@ssaini5028 M2 Pro is rumored for TSMC 4nm too
zen 4 is priced to high.
stop overhyping zen4 and am5 already. its not gonna be so great you make people believe. dont waste your money on the overpriced first generation of boards and cpus. am4 is still plenty good, no reason to jump onto early am5 with the first gen teething issues and greedy cpu pricing. overall platform cost is gonna be insane.
Why are you so butthurt? Bought Intel with no upgrade path after raptor lake?
Exciting times, especially as it's about time for me to upgrade (from 1500X), and I finally have the money to do so. Haven't decided if I'd go Raptor Lake or Raphael (or even Vermeer or Alder Lake), but the longer lifespan of AM5 is certainly very enticing. Guess I'll just have to wait and see performance and prices...
So cringe voice editing
Reduction of Watt per performance is the most exciting part of this announcement for me, the recidivous increase in power consumption shown by both Intel and NVidia in recent models is simply unsustainable
@b gg My electric bill also agrees 💸💸
Steakm deck 2 on cut down phoenix? Sounds delish!
Product, performance, frequencies etc. are incredible. What's not incredible is them not giving Zen 4 a $100 price cut. Zen 3 was forgivable because it's the same platform, a drop in upgrade. Zen 4 requires a new, costly motherboard and expensive DDR5 memory (6000 MT/s CL32) to run as advertised. This way, they'll just get undercut by Intel's 13th gen which is also compatible on Z690/B650 with DDR4...
Oh I understand how incredible, but we have reached a point where unless you are a content creator or doing some serious heavy lifting the question becomes "what EXACTLY" are you going to do with all that power? I'm still on the R5 3600 because even with gaming its pushing more frames than my monitor can show so I don't even need to move to a 5600x so if I were to get a 7 series (and blow an insane amount on replacing my 32Gb of DDR4 with DDR5) it would probably spend 70% of its time with half the cores doing nothing and I doubt it would feel any different than the 3600.
incredible ?? buddy you miss spelled Average
1. Its EXPENSIVE
2.Average performance gains
3.No Stock Cooler
4.Have to get new platform
5.HOT
6.PowerHungry
7.No APU
8. No X3D
9.Average RAM speed scaling
10.Latency
6:52 Zen 4 uses shared L3 cache as well, just a lot more of it. In Zen 1, each 4 Core Complex (CCX) had 8 MB of shared L3 cache, in Zen 2, this was doubled to 16 MB. Zen 3 and Zen 4 use a shared 32 MB L3 cache per 8 Core die (CCD).