It was always the budget option for people who couldn't afford Intel/Nvidia. Hardware is much cheaper now relatively speaking, so there's no need for AMD to exist. A five year old Intel/Nvidia setup is better than a brand new AMD one.
also AM4 has been getting new CPUs for 8 years at this point so you can save a ton of money on motherboards if you do an in-place upgrade.
using the numbers from
AMD has the best gayming CPUs right now if that's what you're into. 7800 X3D is the best bank for your buck and then the 7950X3D is the best overall
>have a 5600x >138fps average >can get a 31% bump for $229 (5600x3d) >can get a 41% bump for $289 (5800x3d) >a 36% bump would cost $319 for intel (13600k) >a 43% bump would cost $417 for intel (13700k) >add $100-$250 for a motherboard, depending on how cheap you want to be
>A five year old Intel/Nvidia setup is better than a brand new AMD one.
Maybe a very high-end system then vs a low-budget option now. At least if you don't care about Vulkan performance or Linux support.
>>no full ray tracing acceleration
Wrong >no real wmma acceleration
Wrong >garbage hardware encoder
Wrong >chiplets connected with copper
How is this a problem? There aren't too many options and alternatives are much more expensive for no real benefit >garbage memory controllers
ROFL, the funny part is that memory controller on Zen4 is better than controller on Raptor Lake.
Best bang for buck for both CPUs and GPUs. CPUs are way more energy efficient than the competition. Actual have good support and FOSS drivers for GNU/Linux.
Nothing went wrong. They are only behind in AI and ray tracing, which isn't a concern for most consumers.
>They are only behind in AI and ray tracing, which isn't a concern for most consumers.
These are just general compute tasks they're only behind by the same margin that they're behind in overal gpu power.
Word on BB in one of the OEM groups is that Zen5 is 20% faster ST per clock compared to Zen4, multicore scaling slightly higher, much more energy efficient. Apparently the int side of the core is a couple ALUs wider, the FPU supports native AVX512, and the front end has some ultra complex fetch and branch predictor they're calling AI accelerated now
Grain of salt as always since these dudes, while confirmed work at some big OEMs, are poor as shit and the possibility of spreading rumors to fish for donations from western journos to get an "exclusive" is very real
Ikr, and then some castrated embedded ai chip, that I don't really see a use-case other than aiding telemetry.
2 months ago
Anonymous
Stop being moronic for a second here, kiddo
There is no other "embedded AI chip." Learn what a branch predictor is before you post again
2 months ago
Anonymous
So the advertising homies did really call a pentium4 tech an AI?
2 months ago
Anonymous
There have been people calling things AI since the 1960s.
Some early ultra basic algos for controlling the timing of traffic lights in the 1980s were claimed to be AI.
Nothing. Best CPU in the market. Budget GPU in the market. Also the best GPU for Linux, yet.
Intel ia only good for certain workloads like design which required high frequencies. AMD here compete near.
Unless you're fricked by a contract or you're a 2d designer, AMD is the only white man choice.
They stopped fighting for market share and started just taking whatever they get. They could sell the 7900 XTX for $399 and BTFO Nvidia, but they're worried that Nvidia will just lower their prices and AMD will get left with still no market share and also 5% of the profit from what few sales they do get.
AMD has faced several challenges recently. Discussions in the tech community suggest that the RDNA3 graphics cards may have been overhyped, leading to disappointment with the actual performance compared to expectations. Additionally, AMD’s stock has seen fluctuations due to various factors, including market reactions to competitor Nvidia’s earnings and anticipation of economic events. There have also been technical issues, such as problems with clock speeds and bugs affecting Windows 11 PCs with AMD CPUs. These factors combined have contributed to a perception that AMD is facing setbacks, despite the company’s overall growth and success in the semiconductor industry.
RDNA3, in the discrete cards at least, brought ALU count per CU up to 128. This is the ultimate dream of GCN which was 4x16 SIMD, RDNA changed it to 2x32SIMD, now they have dual issue instruction potential to increase throughput. The future CDNA iterations for enterprise will be based on this. What AMD did was bring these two differentiated architectures come back together.
Big performance gains, improving scalar perf, improving inferencing and RT perf comes with RDNA4.
AMD has faced several challenges recently. Discussions in the tech community suggest that the RDNA3 graphics cards may have been overhyped, leading to disappointment with the actual performance compared to expectations. Additionally, AMD’s stock has seen fluctuations due to various factors, including market reactions to competitor Nvidia’s earnings and anticipation of economic events. There have also been technical issues, such as problems with clock speeds and bugs affecting Windows 11 PCs with AMD CPUs. These factors combined have contributed to a perception that AMD is facing setbacks, despite the company’s overall growth and success in the semiconductor industry.
Lisa Su creating the entire financial plan for the restructuring and financial recovery of AMD while Rory Read was still in charge
She literally did it all
Jim Keller said in an interview that AMD already had everything (power efficiency team + high performance team), his job was just to coordinate the two
2 months ago
Anonymous
its almost as if multi-billion company has access to best engineers on the market huh
Nothing, AMD is at a historical high. The actual truth is performance gaming market is a fading to small, but vocal niche that makes small money. Nvidia and AMD have draw their attention elsewhere. Nvidia is riding entirely on sheer momentum and wouldn't care if gamers en mess stop buying discrete GPUs.
Min/maxs types are going to look like 1960-1980s motorheads who pretend that muscle cars are still ubiquitous and are massively profitable.
Chinese hardware
Indian software
The match made in hell
this will be intel in next gen since they're gonna piggy back tsmc's chips
>Chinese hardware
>Indian software
sums up every piece of tech nowadays.
Not enough funding for RTG
It was always the budget option for people who couldn't afford Intel/Nvidia. Hardware is much cheaper now relatively speaking, so there's no need for AMD to exist. A five year old Intel/Nvidia setup is better than a brand new AMD one.
AMD has the best gayming CPUs right now if that's what you're into. 7800 X3D is the best bank for your buck and then the 7950X3D is the best overall
>tfw basic 7600
Should have bought something better, didn't know CPU matters that much.
also AM4 has been getting new CPUs for 8 years at this point so you can save a ton of money on motherboards if you do an in-place upgrade.
using the numbers from
>have a 5600x
>138fps average
>can get a 31% bump for $229 (5600x3d)
>can get a 41% bump for $289 (5800x3d)
>a 36% bump would cost $319 for intel (13600k)
>a 43% bump would cost $417 for intel (13700k)
>add $100-$250 for a motherboard, depending on how cheap you want to be
Why the frick do games need that much cache to perform reasonably well? Glad AMD could throw more cache at the problem but it's still fricked up
>A five year old Intel/Nvidia setup is better than a brand new AMD one.
Maybe a very high-end system then vs a low-budget option now. At least if you don't care about Vulkan performance or Linux support.
>9900k $279
>5600 $135
Have you even built a computer since 2018?
rumao he doesnt know
Shalom!
Nothing went wrong, the same way I am israeli and there is nothing wrong with that.
unironically nvidia fanboys, why would they spend R&D on products that nobody will buy anyway
Good hardware with no software support.
Only GAI can save them.
>Good hardware
>no full ray tracing acceleration
>no real wmma acceleration
>garbage hardware encoder
>chiplets connected with copper
>garbage memory controllers
Still, most of problems come from shit software.
>>no full ray tracing acceleration
Wrong
>no real wmma acceleration
Wrong
>garbage hardware encoder
Wrong
>chiplets connected with copper
How is this a problem? There aren't too many options and alternatives are much more expensive for no real benefit
>garbage memory controllers
ROFL, the funny part is that memory controller on Zen4 is better than controller on Raptor Lake.
Best bang for buck for both CPUs and GPUs. CPUs are way more energy efficient than the competition. Actual have good support and FOSS drivers for GNU/Linux.
Nothing went wrong. They are only behind in AI and ray tracing, which isn't a concern for most consumers.
>They are only behind in AI and ray tracing, which isn't a concern for most consumers.
These are just general compute tasks they're only behind by the same margin that they're behind in overal gpu power.
Word on BB in one of the OEM groups is that Zen5 is 20% faster ST per clock compared to Zen4, multicore scaling slightly higher, much more energy efficient. Apparently the int side of the core is a couple ALUs wider, the FPU supports native AVX512, and the front end has some ultra complex fetch and branch predictor they're calling AI accelerated now
Grain of salt as always since these dudes, while confirmed work at some big OEMs, are poor as shit and the possibility of spreading rumors to fish for donations from western journos to get an "exclusive" is very real
What's the use case of that local embedded AI chip? Other than ratting on you ofc.
"AI" is a meaningless buzzword at this point
Ikr, and then some castrated embedded ai chip, that I don't really see a use-case other than aiding telemetry.
Stop being moronic for a second here, kiddo
There is no other "embedded AI chip." Learn what a branch predictor is before you post again
So the advertising homies did really call a pentium4 tech an AI?
There have been people calling things AI since the 1960s.
Some early ultra basic algos for controlling the timing of traffic lights in the 1980s were claimed to be AI.
Branch predictors have been called AI here and there since like 2012, its just basic preemption for instructions to be dispatched
even jim keller predicted zen 5 to have 20-25% ST per clock increase
literally doesnt matter anymore if you go amd or intel
Nothing. Best CPU in the market. Budget GPU in the market. Also the best GPU for Linux, yet.
Intel ia only good for certain workloads like design which required high frequencies. AMD here compete near.
Unless you're fricked by a contract or you're a 2d designer, AMD is the only white man choice.
They stopped fighting for market share and started just taking whatever they get. They could sell the 7900 XTX for $399 and BTFO Nvidia, but they're worried that Nvidia will just lower their prices and AMD will get left with still no market share and also 5% of the profit from what few sales they do get.
AMD has faced several challenges recently. Discussions in the tech community suggest that the RDNA3 graphics cards may have been overhyped, leading to disappointment with the actual performance compared to expectations. Additionally, AMD’s stock has seen fluctuations due to various factors, including market reactions to competitor Nvidia’s earnings and anticipation of economic events. There have also been technical issues, such as problems with clock speeds and bugs affecting Windows 11 PCs with AMD CPUs. These factors combined have contributed to a perception that AMD is facing setbacks, despite the company’s overall growth and success in the semiconductor industry.
RDNA3, in the discrete cards at least, brought ALU count per CU up to 128. This is the ultimate dream of GCN which was 4x16 SIMD, RDNA changed it to 2x32SIMD, now they have dual issue instruction potential to increase throughput. The future CDNA iterations for enterprise will be based on this. What AMD did was bring these two differentiated architectures come back together.
Big performance gains, improving scalar perf, improving inferencing and RT perf comes with RDNA4.
Thanks ChatGPT
Eat shit you nonwhite
I have more first hand technical knowledge than you could ever dream of
Know your place, machine
what went right?
Lisa Su creating the entire financial plan for the restructuring and financial recovery of AMD while Rory Read was still in charge
She literally did it all
homie that's nuts
She also brought back the engineer(s) from the original Thunderbird team iirc.
Jim Keller said in an interview that AMD already had everything (power efficiency team + high performance team), his job was just to coordinate the two
its almost as if multi-billion company has access to best engineers on the market huh
Did something happen?
Ive been out of the loop for like a year or so
>What went wrong?
Their stock only went up 750% over 5 years instead of 2200% like Nvidia. So really, nothing went wrong.
Nothing, AMD is at a historical high. The actual truth is performance gaming market is a fading to small, but vocal niche that makes small money. Nvidia and AMD have draw their attention elsewhere. Nvidia is riding entirely on sheer momentum and wouldn't care if gamers en mess stop buying discrete GPUs.
Min/maxs types are going to look like 1960-1980s motorheads who pretend that muscle cars are still ubiquitous and are massively profitable.