What do AMD CPUs have over Intel?

[vm'N]

Intel's IPC is only about 20% better than AMD's. In Haswell it's in .86 and Piledriver it's .68, when fully unlocked the 8350 will perform AT LEAST on par with with a 4770k, as the video by Logan shows.

That is VERY wrong. First of all, forget EVERYTHING about IPC on cisc processors. It is a meaningless number. You have no ideas of how many uops are been executed per cycle on CISC processors.

Second; How do you calculate the IPC on piledriver? It features CMT, which makes IPC an even more meaningless number.

The FX 8350 is on par (when overclock slightly) with the core I7 3770k in CERTAIN workloads. NOT ALL. Might be 2/10 workloads where it will match or be better.

I don't care what logan have mixed together, the architecture doesn't lie.

[patrickjp93]

Intel's IPC is only about 20% better than AMD's. In Haswell it's in .86 and Piledriver it's .68, when fully unlocked the 8350 will perform AT LEAST on par with with a 4770k, as the video by Logan shows.

This would be true if not for AMD's other architectural shortcomings when it comes to context switching or handling float-heavy tasks (1 float scheduler per 2 cores) which aren't reflected in the raw IPC numbers.

[patrickjp93]

That is VERY wrong. First of all, forget EVERYTHING about IPC on cisc processors. It is a meaningless number. You have no ideas of how many uops are been executed per cycle on CISC processors.

Second; How do you calculate the IPC on piledriver? It features CMT, which makes IPC an even more meaningless number.

The FX 8350 is on par (when overclock slightly) with the core I7 3770k in CERTAIN workloads. NOT ALL. Might be 2/10 workloads where it will match or be better.

I don't care what logan have mixed together, the architecture doesn't lie.

There are only 4 microinstructions basically consisting of fetch, decode, execute, and return. And you know the exact NUMBER which are executed, but you do not know the order. 

 

Don't listen to him he's not very well versed in what microcode buffering actually is. You can compare IPC apples to apples, and then any other architectural considerations (like the oddball float scheduling on AMD's modules) afterward.

http://en.wikipedia.org/wiki/Micro-operation

 

There is some hidden reshuffling of microinstructions to shrink the workload, but this shuffling takes up a lot of clock cycles.

[Bloodkil933]

That is VERY wrong. First of all, forget EVERYTHING about IPC on cisc processors. It is a meaningless number. You have no ideas of how many uops are been executed per cycle on CISC processors.

Second; How do you calculate the IPC on piledriver? It features CMT, which makes IPC an even more meaningless number.

The FX 8350 is on par (when overclock slightly) with the core I7 3770k in CERTAIN workloads. NOT ALL. Might be 2/10 workloads where it will match or be better.

I don't care what logan have mixed together, the architecture doesn't lie.

I've no idea what you are trying to prove, Logan is not sponsored by AMD nor is he a regular user of it. If you listen all the way through, then you may actually understand how he got his numbers. Unless you actually have something to show me, then please do, otherwise what you're actually saying means nothing. Also any application/game that can utilise all 8 cores with the overclock that Logan has, then as he proved, beats out the 3770k 9/10.

 

This would be true if not for AMD's other architectural shortcomings when it comes to context switching or handling float-heavy tasks (1 float scheduler per 2 cores) which aren't reflected in the raw IPC numbers.

Of course there is more to it than that, I've seen what you have said in other threads and seem to actually know what they are talking about

 

Also IPC is generally irrelevant, it's about the actual numbers, as Logan showed. If you are gaming, I would and still will recommend an 8350/8320 and to bump it up to 4.8Ghz.

[vm'N]

There are only 4 microinstructions basically consisting of fetch, decode, execute, and return. And you know the exact NUMBER which are executed, but you do not know the order. 

 

Don't listen to him he's not very well versed in what microcode buffering actually is. You can compare IPC apples to apples, and then any other architectural considerations (like the oddball float scheduling on AMD's modules) afterward.

http://en.wikipedia.org/wiki/Micro-operation

 

There is some hidden reshuffling of microinstructions to shrink the workload, but this shuffling takes up a lot of clock cycles.

Well, you have completely misunderstood me. CISC processors have complex instructions, which are translate doing the decoders stages, translated into RISC-a-like instructions also known as uops (micro operation). Intel and AMD use custon uops, and the uops are only known to the company. Once the complex instruction have been decoded, we would have no idea of what instructions will be executed. That is why IPC is meaningless to CISC processors.

What you are thinking of is the pipeline stages. Where you would be looking at branch > prefetch > fetch > queue > decode > execute > store / or load.

IPC is used on RISC processors, as there you have full vision of what instructions are been executed.

[vm'N]

I've no idea what you are trying to prove, Logan is not sponsored by AMD nor is he a regular user of it. If you listen all the way through, then you may actually understand how he got his numbers. Unless you actually have something to show me, then please do, otherwise what you're actually saying means nothing. Also any application/game that can utilise all 8 cores with the overclock that Logan has, then as he proved, beats out the 3770k 9/10.

 

Of course there is more to it than that, I've seen what you have said in other threads and seem to actually know what they are talking about

 

Also IPC is generally irrelevant, it's about the actual numbers, as Logan showed. If you are gaming, I would and still will recommend an 8350/8320 and to bump it up to 4.8Ghz.

Piledriver will only have the same or more throughput as haswell in workloads than is well multithreaded, almost purely integer commands, been more predictable.

You will need to successfully keep all 8 ALU clusters busy, instead of 4 ALU cluster on haswell, which is only possible on very few workloads.

At that point, and only that piledriver will have the same if not more (due to higher clockspeeds) than haswell. But this is only at certain workloads, that are not used in much everyday workloads.

Haswell will burn piledriver in anything that doesn't fully use every 8 ALU cluster, and don't even mention SIMD, as Intel is completely dominated in that field.

Also you would need to be carefully not to put to much pressure on the frontend as it will end up starve the backend for resources.

It is quite late, and I will head for bed now. But tomorrow I will make sure to post lots of pictures and benchmark and other lovely things. However that is first tomorrow. I will also answer all the questions you might ask over the night.

[patrickjp93]

Well, you have completely misunderstood me. CISC processors have complex instructions, which are translate doing the decoders stages, translated into RISC-a-like instructions also known as uops (micro operation). Intel and AMD use custon uops, and the uops are only known to the company. Once the complex instruction have been decoded, we would have no idea of what instructions will be executed. That is why IPC is meaningless to CISC processors.

What you are thinking of is the pipeline stages. Where you would be looking at branch > prefetch > fetch > queue > decode > execute > store / or load.

IPC is used on RISC processors, as there you have full vision of what instructions are been executed.

 

Piledriver will only have the same or more throughput as haswell in workloads than is well multithreaded, almost purely integer commands, been more predictable.

You will need to successfully keep all 8 ALU clusters busy, instead of 4 ALU cluster on haswell, which is only possible on very few workloads.

At that point, and only that piledriver will have the same if not more (due to higher clockspeeds) than haswell. But this is only at certain workloads, that are not used in much everyday workloads.

Haswell will burn piledriver in anything that doesn't fully use every 8 ALU cluster, and don't even mention SIMD, as Intel is completely dominated in that field.

Also you would need to be carefully not to put to much pressure on the frontend as it will end up starve the backend for resources.

It is quite late, and I will head for bed now. But tomorrow I will make sure to post lots of pictures and benchmark and other lovely things. However that is first tomorrow. I will also answer all the questions you might ask over the night.

And if it's a hyperthreaded 4-core Haswell, you're talking about 8 ALU clusters.

[Bloodkil933]

Piledriver will only have the same or more throughput as haswell in workloads than is well multithreaded, almost purely integer commands, been more predictable.

You will need to successfully keep all 8 ALU clusters busy, instead of 4 ALU cluster on haswell, which is only possible on very few workloads.

At that point, and only that piledriver will have the same if not more (due to higher clockspeeds) than haswell. But this is only at certain workloads, that are not used in much everyday workloads.

Haswell will burn piledriver in anything that doesn't fully use every 8 ALU cluster, and don't even mention SIMD, as Intel is completely dominated in that field.

Also you would need to be carefully not to put to much pressure on the frontend as it will end up starve the backend for resources.

It is quite late, and I will head for bed now. But tomorrow I will make sure to post lots of pictures and benchmark and other lovely things. However that is first tomorrow. I will also answer all the questions you might ask over the night.

Of course, anything that can fully utilise all 8 cores of the 8350 will match a 4770k. Agreeably in programs that use 2 or 4 cores, the Haswell will win, but that can also depend on bottlenecks, be it GPU or CPU.

[CreativeTHC]

As i see it i havent heard one good thing about amds cpus besides value. anyone disagree and why? growing up ive always loved amd but being a gamer im leaning towards intel nowadays

As technology goes, some say AMD it was and still have it. Not having money to evolve that tech, is a problem. If is to talk your question, as a daily use gamer bla bla pc, AMD is the way to go just becuz of the price. If is to take 2 cpu's there are choices and money ofc. Now I do think tha an AMD FX 820 which is just a FX 8350 but the 820 is $50 cheaper, is the bst cpu out there when it comes to work, gaming, etc. But if you rly are a rly heavly desinger, video editor and such,  ofc Itel has some nasty expensive good thing which are the best right now. There are tons of AMD haters, i have never used on myself an AMD, just an fx 6350 and it's a rly amazing cpu. You will never get this straight untill you teste and see it with your own eyes. What is the cpu you need for, why, budged etc.

[ZeroBreaker]

Still my fault for not getting enough people on my side to sway my congressman's opinion.

 

First, way to address all the points i brought up.

Second, please go and live in the real world for a while. It's a whole different place from the fantasy land you seem to live in, and it might help you to understand these issues a little better and maybe even help you to stop coming off so concieted and self-righteous in your posts.

[patrickjp93]

First, way to address all the points i brought up.

Second, please go and live in the real world for a while. It's a whole different place from the fantasy land you seem to live in, and it might help you to understand these issues a little better and maybe even help you to stop coming off so concieted and self-righteous in your posts.

I addressed all your points in previous posts. You had nothing new. Second, I already live in the real world and am politically active. Any mis-step of the U.S. government is directly my fault for not holding them accountable as it is the fault of every U.S. citizen. You're the conceited one who doesn't take responsibility.

 

The greek people brought this on themselves with their opulence and arrogance. End of story.

[PhilBoisvert]

MOAR CORES!!! Get all them cores!! Btw I have an AMD build and i love it

[ZeroBreaker]

I addressed all your points in previous posts. You had nothing new. Second, I already live in the real world and am politically active. Any mis-step of the U.S. government is directly my fault for not holding them accountable as it is the fault of every U.S. citizen. You're the conceited one who doesn't take responsibility.

 

The greek people brought this on themselves with their opulence and arrogance. End of story.

 

Sorry, but you didn't provide any evidence for your claims intel was innovating and aiming for phones, which was the main thing i took issue with.

 

Are you skimreading my posts or something? I specifically said that in greece there were many protestors, many people doing ALL THEY COULD in the legal system, holding the government accountable to the extent they could. They physically could not do any more. So you are saying they should be blamed because they could not physically do any more? Read properly next time mate.

 

If the government takes action, all the citizens are accountable? Funny, that's the same logic the taliban use when demonising US citizens, and the same logic (though not as outspoken) as the US government uses when confronted about the drone strikes that kill far more innocent civillians than actual terrorists. If you actually believe in what you are saying, I hope I never meet you, I hope you never get into a position of power, and I hope you never pass these views on to any future children you may have.

[patrickjp93]

Sorry, but you didn't provide any evidence for your claims intel was innovating and aiming for phones, which was the main thing i took issue with.

 

Are you skimreading my posts or something? I specifically said that in greece there were many protestors, many people doing ALL THEY COULD in the legal system, holding the government accountable to the extent they could. They physically could not do any more. So you are saying they should be blamed because they could not physically do any more? Read properly next time mate.

 

If the government takes action, all the citizens are accountable? Funny, that's the same logic the taliban use when demonising US citizens, and the same logic (though not as outspoken) as the US government uses when confronted about the drone strikes that kill far more innocent civillians than actual terrorists. If you actually believe in what you are saying, I hope I never meet you, I hope you never get into a position of power, and I hope you never pass these views on to any future children you may have.

I provided strong inductive reasoning based on what their recent directional pushes have been. If that's not enough for you then go back to college and learn what logic actually is. Take an algorithms course, or informal logic. My inductive reasoning is sound.

 

As per Greece having many protestors, so? Clearly there weren't enough/a majority. If the government doesn't follow the will of the people they need to be held to swift and grueling account by the people. That goes back to the blame falling on the people, not just the government. It doesn't matter what you suggest if you can't articulate the argument and persuade people you're right. That is the essence of politics. SOME people should have convinced MORE people so the government was facing majority opposition.

 

And the Taliban is right to blame the U.S. citizenry. We don't hold our government to account. We don't do enough to save innocent lives in war. We are incompetent and we need to take responsibility for that. We are citizens of the world, not just our own country. 

[vm'N]

And if it's a hyperthreaded 4-core Haswell, you're talking about 8 ALU clusters.

Hyper-threading doesn't duplicate or split up an ALU cluster. A core I7 4770k still only have 4 ALU clusters.

[Faa]

You know how much of an idiot you look saying something like that? The 8350 is more future-proof than a i7 4-core processor as the current DirectX version only utilizes about 2 cores, 4 at the most. With the 8350 and 8320 for that matter being 8 core processors they current are at the disadvantage by the poor multi-core load balancing in DirectX 11. DirectX 12 will likely improve the performance a fair amount or those using 6 and 8 core processors in general as it has MUCH better multi-core load balancing, thus making the 8350 (and other 6 - 8 core processors too) future-proof. Please use your brain before you write crap like that.

 

EDIT: Mantle pretty much does what DirectX 12 will do, to back myself up i'm going to link to my DirectX11 vs Mantle thread regarding Thief. I get a 29.2% improvement in performance thanks to the optimizations made with Mantle. Thread: http://linustechtips.com/main/topic/163329-mantle-vs-dx11-thief/

People ignored your post because you automatically looked like an idiot to them when you brought futureproof up. Your level of understanding CPU's perfectly explains my statement that why Intel has more competition with amd fanboys than their cpu's. Dude paid like 250$ for his olddozer cpu who thinks its futureproof with its 8 years outdated IPC getting wrecked even at 5GHz by a i5 at stock when 2 games are properly utilizing more than 4 cores. Before Directx12 comes out, you'd be asking advice here for a new cpu so I'd advice you to shut it until DX12 comes out.

Shut it about Mantle, nvidia's new drivers are reducing cpu bottlenecking way better. Linking the evidence for this:

http://pclab.pl/art58123.html

http://i.imgur.com/taEVTSg.png

http://cdn.wccftech.com/wp-content/uploads/2014/02/battlefield-4-mantle-benchmarks.gif

Google amd vs nvidia cpu overhead

[Faa]

Intel's IPC is only about 20% better than AMD's. In Haswell it's in .86 and Piledriver it's .68, when fully unlocked the 8350 will perform AT LEAST on par with with a 4770k, as the video by Logan shows.

Rofl? Youre referring to single core performance with IPC.

http://vr-zone.com/articles/amd-fx-8350-vs-intel-core-i7-3770k-4-8ghz-multi-gpu-gaming-performance/17494.html

http://www.anandtech.com/show/7189/choosing-a-gaming-cpu-september-2013/8

http://pclab.pl/art58123.html

http://www.pcgameshardware.de/screenshots/original/2013/09/SC2HotS-pcgh.png

http://www.hardwarepal.com/best-cpu-gaming-9-processors-8-games-tested/

http://i.imgur.com/4rU8fJF.png

http://hardwarepal.wpengine.netdna-cdn.com/wp-content/uploads/2013/11/Battlefield-4-1920-x-1080-Low-settings-GTX-770-vs-7970.jpg

http://cdn.overclock.net/1/1f/500x1000px-LL-1fa7a24a_3350949334973.png

FX 9590:

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/62166-amd-fx-9590-review-piledriver-5ghz-11.html

http://www.xbitlabs.com/articles/cpu/display/amd-fx-9590-9370_5.html#sect0

 

[Priller]

Shut it about Mantle, nvidia's new drivers are reducing cpu bottlenecking way better. Linking the evidence for this:

http://pclab.pl/art58123.html

http://i.imgur.com/taEVTSg.png

http://cdn.wccftech.com/wp-content/uploads/2014/02/battlefield-4-mantle-benchmarks.gif

Google amd vs nvidia cpu overhead

 

 

PClab: If you look at some of their benchmarks the results don't change. 

Picture: Done at 1280x720. Plus it's Watch Dogs on release. Total Biscuit had issues with his TITAN. 

WCCFTech: Very few people believe what comes from WCCFTech, surprised you would go that low.

 

 

How about a trusted source? 

 

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-33750-Driver-Analysis-Single-GPU-and-SLI-Tested

 

It does so good it got 5.2% max improvement. The only way 30-35% was seen was by fixing an SLI issue. 

 

Mantle is better.

[Akolyte]

They have more cores, more GHZ but less speed and they run a bit hotter.  Esspecially the FX 9 series they are freaking hot! Hotter than Haswell! 

[Priller]

They have more cores, more GHZ but less speed and they run a bit hotter.  Esspecially the FX 9 series they are freaking hot! Hotter than Haswell! 

 

More cores and more core speed yes. They actually run cooler then Haswell due to a larger die size and being soldered.

 

AMD tried something new and it didn't work that well performance wise. Phenom IIs have over a 1Ghz advantage in per core performance. The issue is they spent so much money on it they couldn't go back so they had to produce and sell it but they are going back to SMT reportedly.

[Faa]

PClab: If you look at some of their benchmarks the results don't change. 

Picture: Done at 1280x720. Plus it's Watch Dogs on release. Total Biscuit had issues with his TITAN. 

WCCFTech: Very few people believe what comes from WCCFTech, surprised you would go that low.

 

 

How about a trusted source? 

 

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-33750-Driver-Analysis-Single-GPU-and-SLI-Tested

 

It does so good it got 5.2% max improvement. The only way 30-35% was seen was by fixing an SLI issue. 

 

Mantle is better.

You're missing my point; pcper didnt test the cpu overhead just the actual raw performance at max load. The sources I included are testing the overhead, not just 780ti at 99% vs 290x at 99%. Mantle isn't better atm in terms of reducing cpu bottlenecking, 780ti's outperform 290x's in Star swarm and plenty of more games its the same thing.

[ZeroBreaker]

I provided strong inductive reasoning based on what their recent directional pushes have been. If that's not enough for you then go back to college and learn what logic actually is. Take an algorithms course, or informal logic. My inductive reasoning is sound.

 

As per Greece having many protestors, so? Clearly there weren't enough/a majority. If the government doesn't follow the will of the people they need to be held to swift and grueling account by the people. That goes back to the blame falling on the people, not just the government. It doesn't matter what you suggest if you can't articulate the argument and persuade people you're right. That is the essence of politics. SOME people should have convinced MORE people so the government was facing majority opposition.

 

And the Taliban is right to blame the U.S. citizenry. We don't hold our government to account. We don't do enough to save innocent lives in war. We are incompetent and we need to take responsibility for that. We are citizens of the world, not just our own country. 

 

You still havn't provided any logical evidence that intel is innovating for and pushing into mobile.

 

If you cannot fathom that there are plenty of situations, such as we see all the time in the middle east, where  a small group of people cannot convince the government or the people with their limited power and resources, then nothing I say can educate you on such matters. Politics isn't simple, people aren't simple, just because what you say is right does not mean that people will listen, care or act. A revolution is not so easily strung together.

 

Seriously, your ideology falls right in line with extremists. The Taliban are therefore justified in killing innocent americans, innocent pakistanis, innocent people in general, because by your logic they are responsible for the wrongdoings in the middle east? Is the government therefore justified in killing the innocent afghanis, iraqis and pakistanis with drones, because they all hold that responsibility for that loss of american life? Is it the fault of the jews in nazi germany they were subjected to genocide?

 

Please, don't just type whatever bull comes out of your rear end, because that's all i can see in your posts.

[Faa]

*

Just don't bother arguing with him. Last time he claimed there's a fully interactive interface for the gpu bios, like accessing a motherboard bios. So before you boot in windows, you can enable ECC change fan profiles of your discrete gpu. He claimed to have a video that shows it, didnt post it but instead in a different thread he claimed that he PM'ed everyone with evidence. Dude was silly enough to just google "access gpu bios" and link this http://www.ehow.com/how_8206263_access-gpu-bios.html

He knows he was wrong but kept going on using any kind of random technical BS

[thebennyboy]

Threads like this just make me want to slap some ppl, the amount of fanboys on this forum are terrible. Both Intel and AMD have fanboys, although it is more common for people to talk shit about AMD. If your going to have an argument, prove your point instead of talking out of your ass.

[Linus_torvalds]

NOTHING, NOTHING, NOTHING!!!

 

intel is better in every single aspect!!!