AMD on the road to recovery.

[gpuerrilla]

I dunno, the facts seem to be out there, I know what I think happened...

This is one of the reasons why I critique them and nvidia. I may buy their products from time to time but if another vendor offers what I need then I have no loyalty to them.

 

[Bencher]

if another vendor offers what I need then I have no loyalty to them.

Good, that's a great attitude, and im on the same boat. Problem is amd doesn't usually offer what other companies do, so you are kinda forced to go with intel and nvidia. Especially in the GPU department, those cudas are invaluable to so many people

 

[humbug]

AMD are a semi-conductor engineering firm, i would argue the best of the best.

What they are not is a software firm, historically they have not been good at that, but improving all the time, they are aware of their shortcomings in software and are tying to do something about it, with increasing success.

Nvidia enjoy being both a very good semi-conductor engineering firm and one of if not the best software firms, i would argue the latter is true.

Intel are not as good at hardware as Nvidia and certainly not as good as AMD, but still very good, in software i think they are second only to Nvidia.

 

[Bencher]

Intel are not as good at hardware

That part is definitely not true. They are much better than AMD, it's the node battle that they are losing against TSMC. Just looking at how long it took for AMD to reach the ST performance of the 14nm 2014 skylake architecture, lol.

 

[humbug]

That part is definitely not true. They are much better than AMD, it's the node battle that they are losing against TSMC. Just looking at how long it took for AMD to reach the ST performance of the 14nm 2014 skylake architecture, lol.

From the moment AMD started designing their own X86 architecture they beat Intel, yes they made a pigs ear of it with Bulldozer but before and since they have beaten Intel at every turn.

 

[4GB means 4GB]

That part is definitely not true. They are much better than AMD, it's the node battle that they are losing against TSMC. Just looking at how long it took for AMD to reach the ST performance of the 14nm 2014 skylake architecture, lol.

I forgot, which company has more personnel, resources, money, etc.? And who was on the verge of collapse? . It's quite impressive what AMD has achieved, considering they were almost bankrupt, and despite their limitations, they managed to surpass the mighty Intel. What does that tell us? If someone with fewer financial resources can achieve more and perform better, then who is the superior hardware company? I suppose Intel would need ten times more money just to be on par with AMD. Intel is indeed a highly efficient company. .

 

[humbug]

I don't like making excuses for Bulldozer, in my view its better for AMD to just own that calamity and move on.....

However.

I forgot, which company has more personnel, resources, money, etc.? And who was on the verge of collapse? . It's quite impressive what AMD has achieved, considering they were almost bankrupt, and despite their limitations, they managed to surpass the mighty Intel. What does that tell us? If someone with fewer financial resources can achieve more and perform better, then who is the superior hardware company? I suppose Intel would need ten times more money just to be on par with AMD. Intel is indeed a highly efficient company. .

Exactly that ^^^

 

[Bencher]

From the moment AMD started designing their own X86 architecture they beat Intel, yes they made a pigs ear of it with Bulldozer but before and since they have beaten Intel at every turn.

It took 3 iterations of Ryzen and TSMC's 7nm to beat 2014's skylake in ST performance on 14nm.

The node plays a huge role and you can see that historically, everytime AMD was on the ropes was because they were using an inferior node and had slow execution, which is kinda what Intel is going through right now by trying to use their own nodes. Think about phenom or phenom II, those would be great CPUs - would definitely beat or at least tie Intel Quad cores but they launched too late having to go up against the i7s.

 

[4GB means 4GB]

No one is preventing Intel from creating a much better product considering the finances they have, and then compensating for a slightly inferior node. Architecture is the most crucial factor; no node can turn a poor architecture into a good one. If Intel were an efficient company, given their size and workforce, they should have produced a product that is four times better, rendering any node advancements useless for AMD. However, Intel is so inefficient and in disarray that they are incapable of achieving anything. They have lost their way in both the CPU and GPU segments, becoming the subject of ridicule across the internet.

 

[humbug]

It took 3 iterations of Ryzen and TSMC's 7nm to beat 2014's skylake in ST performance on 14nm.

The node plays a huge role and you can see that historically, everytime AMD was on the ropes was because they were using an inferior node and had slow execution, which is kinda what Intel is going through right now by trying to use their own nodes. Think about phenom or phenom II, those would be great CPUs - would definitely beat or at least tie Intel Quad cores but they launched too late having to go up against the i7s.

Zen 1 was on an immature Global Foundries 14nm node, a node one might consider a budget node, Zen 1 was at least as efficient as Intel whilst also being on a very mature 14nm.

Separate from that amazingly Zen 1 IPC was only a few % behind Coffeelake.

 

[Emlyn_Dewar]

This thread

 

[Bencher]

Zen 1 was on an immature Global Foundries 14nm node, a node one might consider a budget node, Zen 1 was at least as efficient as Intel whilst also being on a very mature 14nm.

Separate from that amazingly Zen 1 IPC was only a few % behind Coffeelake.

IPC is irrelevant, one architecture might be designed to run at high clocks by sacrificing IPC etc. It does't make sense to compare IPC like that. Cometlake has the same IPC as skylake from 2015 and it beats amds 2018 cpus in ST performance. They even increased ST by 18% with rocketlake on the same ancient 14nm catching up to the latest and greatest that AMD had at 7nm TSMC. That is impressive engineering, getting more from less. Are you telling me that if AMD was using Intel's 14nm and Intel was using 7nm TSMC we would see similar results? I mean come on, we both know the answer to that.

80% of AMD success is TSMC's ability, or Intel's inability in process nodes. Even Apple on their first try made a great CPU cause they were just a node ahead of everyone.

 

[humbug]

I think Zen 1 would have clocked a lot better on Intel's more mature 14nm node, yes, that is what i'm saying.

 

[4GB means 4GB]

Nvidia and AMD are examples of excellent hardware companies. Nvidia consistently remains at the top and effectively utilizes the resources and funds they have. Despite having fewer resources, AMD surpasses Intel and if they had the same resources as Intel, they would bury Intel permanently. Therefore, Intel is an inefficient company, and their products are also inefficient, consuming a lot of power and offering lower performance. This pattern repeats across all their products, including desktop and server processors that are inferior despite larger chip sizes and lower performance. The same inefficiency is observed in their GPUs, with large chip sizes and high power consumption relative to performance. The pattern persists, and their overall efficiency is generally zero.

 

[humbug]

Nvidia and AMD are examples of excellent hardware companies. Nvidia consistently remains at the top and effectively utilizes the resources and funds they have. Despite having fewer resources, AMD surpasses Intel and if they had the same resources as Intel, they would bury Intel permanently. Therefore, Intel is an inefficient company, and their products are also inefficient, consuming a lot of power and offering lower performance. This pattern repeats across all their products, including desktop and server processors that are inferior despite larger chip sizes and lower performance. The same inefficiency is observed in their GPUs, with large chip sizes and high power consumption relative to performance. The pattern persists, and their overall efficiency is generally zero.

Its also about getting complex engineering challenges to work.

AMD was last to 64Bit and first to crack it, AMD were the only ones to crack it full stop, we have true multicore X86 CPU's now thanks to AMD.
MCM and 3D stacked semi conductors had been the final fronter of semi conductor companies again for decades, IBM was the first to crack that, so while AMD was not the first once they did put their hand to it they were the ones who made it work properly, to this day no one does it as well as AMD, AMD was first to 3D stacking, HBM memory, and first to do it on logic dies, Ryzen 5800X3D, still the only ones doing that. AMD was first in MCM GPU's, again no one else has managed that yet.

AMD are a machine pumping #### out where everyone else fails.

 

[Bencher]

I think Zen 1 would have clocked a lot better on Intel's more mature 14nm node, yes, that is what i'm saying.

Sure, but how about 7nm TSMC vs 14nm Intel. I mean come on, every time you dodge the actual point im convinced you know the truth, you just don't want to admit it

 

[humbug]

Sure, but how about 7nm TSMC vs 14nm Intel. I mean come on, every time you dodge the actual point im convinced you know the truth, you just don't want to admit it

It just made it that much better, my point was we have a like for like comparison. if even that.

 

[4GB means 4GB]

Its also about getting complex engineering challenges to work.

AMD was last to 64Bit and first to crack it, AMD were the only ones to crack it full stop, we have true multicore X86 CPU's now thanks to AMD.
MCM and 3D stacked semi conductors had been the final fronter of semi conductor companies again for decades, IBM was the first to crack that, so while AMD was not the first once they did put their hand to it they were the ones who made it work properly, to this day no one does it as well as AMD, AMD was first to 3D stacking, HBM memory, and first to do it on logic dies, Ryzen 5800X3D, still the only ones doing that. AMD was first in MCM GPU's, again no one else has managed that yet.

AMD are a machine pumping #### out where everyone else fails.

That's right, AMD is doing a great job. They are innovative and introducing new things that are not yet available in the market, and it's all the work of AMD engineers, not TSMC's. Intel is not helped by the fact that they can't create a good node; they are unsuccessful in that aspect too. Instead of lifting Intel up, Bencher is burying it, achieving the opposite effect. But that's because there's no person who can save that unsuccessful and cumbersome structure called Intel. No matter how good of a troll Bencher is, Intel is equally bad, and you have to be a wizard to find something positive in Intel.

 

[gpuerrilla]

 

[4GB means 4GB]

Intel innovation .