What is Architecture and what does it affect?

[Random32]

My friend and I are in a heated argument over architecture and what it brings to the table. I'm saying that architecture affects performance and is the basis of almost all things yielded from a CPU/GPU because it's what allows the part to use its resources efficiently/inefficiently, and such. He's saying that architecture has nothing to do with performance and only affect power draw and heat output. According to my friend, having a Haswell CPU and a Piledriver CPU with the same everything (cache, speed, power, cooling) would yield the same performance in games, and the only difference would be that one would give more heat than the other.

 

No matter what I say, he will not admit that architecture is more than just power draw and heat output.

 

Any input?

[Agazed]

He is wrong lol

 

Pentium 1 @ 2GHz = i7-4790k @ 2GHz

[TechDookie]

I don't know the technical details of it, but I'm 99.9999999% positive that architecture affects the CPU's performance. Turbo Boost and Hyperthreading all are based on the CPU's architecture and power consumption and heat output are affected by the architecture.

[Sam Z Man]

Your friend is stupid

[Curufinwe_wins]

Tell him that the 290x has 5.6 TFLOPs compute performance (SP), yet somehow gets beaten 24/7 by the 980 and trades blows with the 970 (4.6 and 3.5 TFLOPs respectively).

 

 

There is your definitive proof of the badness of architecture. 

[Random32]

Your friend is stupid

o damn

[Random32]

Tell him that the 290x has 5.6 TFLOPs compute performance (SP), yet somehow gets beaten 24/7 by the 980 and trades blows with the 970 (4.6 and 3.5 TFLOPs respectively)

If only a maxwell chip could use the resources that a 290X gives to its welding iron of a chip...

[79 Potatoes]

The word Architecture defines itself....

It is basically the design and organization of the logic circuits so it does affect performance hell a lot.

[Frankie]

He's dead wrong. Show him this thread.

[Random32]

He's dead wrong. Show him this thread.

I told him to make it and to prove me wrong and he kept saying things like he's busy or he's in the car or he's in the bathroom lol

[79 Potatoes]

I told him to make it and to prove me wrong and he kept saying things like he's busy or he's in the car or he's in the bathroom lol

Ask him then why CPU die shots are different  (pls get some cringe worthy answers)

[Frankie]

I told him to make it and to prove me wrong and he kept saying things like he's busy or he's in the car or he's in the bathroom lol

 

So when you can prove him wrong, he will rather dodge it and have his own truth. Not much you can do sadly. It will bite him sooner or later.

[SansVarnic]

My friend and I are in a heated argument over architecture and what it brings to the table. I'm saying that architecture affects performance and is the basis of almost all things yielded from a CPU/GPU because it's what allows the part to use its resources efficiently/inefficiently, and such. He's saying that architecture has nothing to do with performance and only affect power draw and heat output. According to my friend, having a Haswell CPU and a Piledriver CPU with the same everything (cache, speed, power, cooling) would yield the same performance in games, and the only difference would be that one would give more heat than the other.

 

No matter what I say, he will not admit that architecture is more than just power draw and heat output.

 

Any input?

ROFLMAO (at your friend)

 

(Sorry but his ignorance has reached a 'level of epic proportion')  :rolleyes: 

[ANewFace]

You are right

[SImoHayha]

Dear dummy. Please go educate yourself. Love Ramamataz.

[Agent181]

Lmfao this thread

[skyress3000]

Architecture is the reason why AMD CPUs with a butt-ton of cores don't perform as well as Intel chips with a butt-ton of cores (i.e. why this is cheaper than this)

[Naruke Alpha]

If only a maxwell chip could use the resources that a 290X gives to its welding iron of a chip...

but that's the thing! you can't give the maxwell chip the resources of a 290x because of the difference in architectures !

[Random32]

Ask him then why CPU die shots are different  (pls get some cringe worthy answers)

 

So when you can prove him wrong, he will rather dodge it and have his own truth. Not much you can do sadly. It will bite him sooner or later.

[Random32]

Ask him then why CPU die shots are different  (pls get some cringe worthy answers)

[79 Potatoes]

Ask him then why CPU die shots are different (pls get some cringe worthy answers)

Lol that direct burnt butthurt way to get out of the conversation

[Godlygamer23]

Ask him where the differences of performance come from within various architectures, even at the same clock speeds with the same number of cores.

[Glenwing]

Architecture defines what a cycle is, and how much computation is done during each cycle. So... yeah, it is kind of important for determining performance. There's a reason a 3.4GHz 6MB L3 quad-core like the i5-3570K beats the crap out of a 3.4GHz 6MB L3 quad-core like the Phenom II X4 965.

 

I can run a direct test for you if you'd like. I have an i5-2500K (3.3GHz Quad-Core, 6MB L3 cache) and a Phenom II X4 965, they're basically the same specs on a spec sheet, I can set them to the same frequency and run any test you'd like

 

Also, there is no fundamental difference between a CPU and a GPU, there's nothing at a basic level that applies to CPU architecture which doesn't apply to GPU architecture. Even that distinction is misleading, since they are both the same thing. You COULD do graphics processing on a CPU, there's no reason you couldn't. It'd just be unbearably slow. The only difference is the type of calculation the architecture is optimized for.

[79 Potatoes]

^- @Random32

[Pomfinator]

Direct quote from the Nvidia blog talking about Maxwell.  He thinks there is some difference between the actual cores and the gates used to regulate workloads and such.  But each core design requires overhauls of that system, since changing things in the core must be changed in the overall chip.

 

"The new SMM, shown in Figure 1, includes all of the architectural benefits of its first-generation Maxwell predecessor, including improvements to control logic partitioning, workload balancing, clock-gating granularity, instruction scheduling, number of instructions issued per clock cycle, and more."