Is AMD Ever Gonna Make A Great CPU?

[melikepie]

well first off look at this
http://www.zdnet.com/blog/computers/intel-says-future-atom-processors-will-match-amd-phenom-ii-performance-in-2015/6453

well i guess intel is getting better every year but amd is getting worse every year but i really dont know what there gonna do next like they have a great gpu buisness and they have a lot of people helping them by buying a lot of GPUs and i think there great priced and there 7000 series offers tons new things and i think there doing well but while there doing that the CPU stuff is really gonna make them go bankrupt. amd does best at going bankrupt and there spending more and more money on bulldozer when basicly there gonna be spending more money then they are getting and i think some people should get fired. i really think that they hired rats for there marketing or intel workers got hired at amd either way they are probaly gone soon and im wondering i amd will EVER make a cpu that will work as well as intels but if it is as good as ivy bridge... they better be sure thats when ivy bridge is the best intel can offer at that time.

 

[amuffin]

Now what if you are an intel enthusiast^^

 

[fazers_on_stun]

AMD will not go bankrupt .. they bought ATI for like a billion or more dollars not long ago, they have some cash, Intels scare mongering just seems to effect the less knowledgable

Try $5.4BN for the ATI purchase. With the senior notes interest, it came out to nearly $6BN. Of that, AMD has written off nearly the entire amount as "goodwill" loss. Not that they are planning to donate ATI to Goodwill 😛, but that the actual current book value of ATI is much, much closer to zero than to $5.4BN.. If AMD had waited as little as 6 months they could likely have paid half that amount..

PCs are a hobby of mine and thats one reason why I choose AMD - to getter better results with my AMD compared to the equivalent Intel product. I choose to fiddle with o/c and voltage/cooling etc.. and .. I really really like doing that and it doesnt cost a great deal

So AMD do make a great cpu, you just have to be an enthusiest to get the most out of them

Happy Computing !

Most of us here are enthusiasts, as are all the benchmarking sites like Anandtech, Tom's, and a couple dozen other sites. And most of those reviews show signficantly worse results for BD than for the lower priced 2500K in gaming and the higher-priced 2600K in video/audio/photoshop stuff.

 

[melikepie]

The company has all but secured the gpu market for all next gen consoles and they are putting a ton of focus into their APU. You have to think about something than the high end power user which is the minority. I was quite critical of AMD after the BD release, but if you look at what they are doing it all makes sense and could lead to some pretty sweet solutions in the not so distant future.

No NVIDIA seems to be winning in the gpu stuff.

 

[jimmysmitty]

No NVIDIA seems to be winning in the gpu stuff.

Right now AMD has the crown with the HD7900 series. It may change with Keplers release. Hell we had a customer wanting a system built with a GTX680, which isn't out till at least next month.

On DT nVidia is winning but AMD is supposed to be getting the next gen consoles and I can see why. Its cheaper to have a single company make the parts than multiple. And it will save power if you can have it as a SoC instead of multiple parts. Right now Intels IGP is not powerful enough and nVidias CPUs are ARM based and very weak. AMD is the only logical solution for consoles.

 

[melikepie]

Right now AMD has the crown with the HD7900 series. It may change with Keplers release. Hell we had a customer wanting a system built with a GTX680, which isn't out till at least next month.

On DT nVidia is winning but AMD is supposed to be getting the next gen consoles and I can see why. Its cheaper to have a single company make the parts than multiple. And it will save power if you can have it as a SoC instead of multiple parts. Right now Intels IGP is not powerful enough and nVidias CPUs are ARM based and very weak. AMD is the only logical solution for consoles.

Who says AMD will be used for the new consoles?

 

[jimmysmitty]

Who says AMD will be used for the new consoles?

Its just a rumor for now as MS wont confirm anything but as I said, it makes sense. Also considering that IBM is no longer producing the Cell CPU, I can see why Sony would also go with AMD.

Logically they are the best to go with for a overall decent system. Sure they could go with a AMD GPU and Intel CPU or some other but why when you can have a SoC that costs less and is easier to work with than something like Cell?

 

[melikepie]

Its just a rumor for now as MS wont confirm anything but as I said, it makes sense. Also considering that IBM is no longer producing the Cell CPU, I can see why Sony would also go with AMD.

Logically they are the best to go with for a overall decent system. Sure they could go with a AMD GPU and Intel CPU or some other but why when you can have a SoC that costs less and is easier to work with than something like Cell?

Wait... IBM made the 360 and ps3 cpu?

 

[jimmysmitty]

Wait... IBM made the 360 and ps3 cpu?

Yes. IBM made both. PS3 uses Cell, 360 used a PowerPC based 3 core CPU.

 

[melikepie]

Yes. IBM made both. PS3 uses Cell, 360 used a PowerPC based 3 core CPU.

wow why did'nt ibm flaw the cell so 360 would be epic
which is better cell or powerpc x3?

 

[grumbledook]

The cell is a more powerful cpu, but harder to program for.

 

[jimmysmitty]

wow why did'nt ibm flaw the cell so 360 would be epic
which is better cell or powerpc x3?

The cell is a more powerful cpu, but harder to program for.

Actually that can be argued. You see the PowerPC 3 core is three PPe CPUs while the Cell is one PPE with 7 SPEs, one in reserve and one as a backup so in reality its one PPE with 5 SPEs.

Overall a PPE is much more powerful than a SPE, a SPE can be though of like SMT just weaker than a full core.

i've actually visited the facility IBM has to make them.

Fun. I have only toured a Ti facility here in town. And Raytheon. And Hughes before it became Raytheon.

 

[nforce4max]

Actually that can be argued. You see the PowerPC 3 core is three PPe CPUs while the Cell is one PPE with 7 SPEs, one in reserve and one as a backup so in reality its one PPE with 5 SPEs.

Overall a PPE is much more powerful than a SPE, a SPE can be though of like SMT just weaker than a full core.



Fun. I have only toured a Ti facility here in town. And Raytheon. And Hughes before it became Raytheon.

Cool then we might be living in the same state, I had family that worked for Raytheon.

 

[zyzz]

why isn't AMD as good anymore? When I bought this AMD X2 6000+ weren't they the time?

I also use to have an AMD 64 3400+. I think it was one of the best during it's time.

How did Intel become the best?

 

[melikepie]

why isn't AMD as good anymore? When I bought this AMD X2 6000+ weren't they the time?

I also use to have an AMD 64 3400+. I think it was one of the best during it's time.

How did Intel become the best?

Well the core 2 could have more cores at higher clock speeds and so the old athlons were not as good and amd kept the same performence per core while intel kept getting better and better chips so intel is much better, amd is currently at bulldozer which the fx-4100 is almost the same as the core 2 quad so there kinda late. Who knows if amd will ever catch up to intel since to make something the same as IB at amd's rate it's gonna take years, by then intel will still be the best.

 

[aqe040466]

well they might not do so well with all the money they spend on there CPUs and i think they need to fix that before they can move on and im hoping for an am3+ cpu or all the sad people like me who baught one

I believe in the saying: "You GOT what you've PAID for"

 

[jimmysmitty]

Cool then we might be living in the same state, I had family that worked for Raytheon.

You could be in Cali. Or Arizona. Thats all I know of after Hughes got bought out.

why isn't AMD as good anymore? When I bought this AMD X2 6000+ weren't they the time?

I also use to have an AMD 64 3400+. I think it was one of the best during it's time.

How did Intel become the best?

They were. But as to how, Intel took the Pentium III Coppermine arch, redesigned it a bit and got higher clocks with higher IPC and lower thermals. The thermals was mainly due to a better process.But the IPC allowed for Core 2 to take back the performance crown and Intel has been just tweaking and enhancing the arch ever since.

Sandy Bridge is the biggest change from Core 2 since core 2 hit.

 

[nforce4max]

You could be in Cali. Or Arizona. Thats all I know of after Hughes got bought out.



They were. But as to how, Intel took the Pentium III Coppermine arch, redesigned it a bit and got higher clocks with higher IPC and lower thermals. The thermals was mainly due to a better process.But the IPC allowed for Core 2 to take back the performance crown and Intel has been just tweaking and enhancing the arch ever since.

Sandy Bridge is the biggest change from Core 2 since core 2 hit.

Coppermine had issues with it's L1 cache when going over 1.133 ghz. Tualatin on the other hand was overclock able to around 1.6ghz or more on a good board when fsb didn't become an issue.

 

[sykozis]

well at this rate i think that intel gonna do just that and then lets see what happends

i really think amd is a failure they really need to look at what intel is doing with ivy bridge and future CPUs and just put all that in a nice really overclockable smaller then 22nm cpu and i will be happy but what exactly matters how small like what does it change and whats 22nm all the prossesers are the same size

You really don't know enough about the computer industry to make the statements you have. Due to laws, Intel has very little control over the total cost of a computer (Anti-Trust and consumer protection laws). Intel can't "completely destroy" AMD as they are the only processor companies competing in the desktop market. Intel also competes against VIA in the embedded market, where AMD used to compete as well, but VIA has no effect on the desktop market. Until ARM processors or VIA enter the desktop market, Intel legally can't do anything to "completely destroy" AMD. In other words, Intel has to make sure AMD can compete in some capacity....

Really? Then why is this chip almost $30 more than what I paid for it? It doesn't look like Intel wants to keep prices low at all.

http://www.newegg.com/Product/Product.aspx?Item=N82E16819115070&Tpk=i7%202600k

Also, if I'm not mistaken (& please correct me if I am) this chip used to be $200: http://www.newegg.com/Product/Product.aspx?Item=N82E16819115072&Tpk=i5-2500k


Personally I don't believe for a second that prices would lower if AMD was gone, since they seemed to have gone up since Bulldozer's release.

AMD will stick around because they can make a good cpu. In addition to that they make a very good apu, & a great gpu. Plus they seem to have a pretty realistic game plan. It doesn't look like they will be disappearing any time soon.

I've never seen the i7 2600K below $300.....and the i5 2500K has never been $200. The i7 2600 was $300 and the i5 2500 was around $200 at one point, but the "K" series processors are at the same price they've been at for the last year.

Well the core 2 could have more cores at higher clock speeds and so the old athlons were not as good and amd kept the same performence per core while intel kept getting better and better chips so intel is much better, amd is currently at bulldozer which the fx-4100 is almost the same as the core 2 quad so there kinda late. Who knows if amd will ever catch up to intel since to make something the same as IB at amd's rate it's gonna take years, by then intel will still be the best.

The Conroe architecture had shorter pipelines and higher IPC than the K8 (9-12IPC) or K10 architecture. Conroe had lower clocks speeds compared to the Northwood or Prescott architectures based on NetBurst. NetBurst had long pipelines, which catered to high clock speeds but suffered from low IPC (4-6) as a result. AMD's "Bulldozer" architecture is very similar to Intel's NetBurst architectures in that is has long pipelines and thus, low IPC.

 

[aqe040466]

You really don't know enough about the computer industry to make the statements you have. Due to laws, Intel has very little control over the total cost of a computer (Anti-Trust and consumer protection laws). Intel can't "completely destroy" AMD as they are the only processor companies competing in the desktop market. Intel also competes against VIA in the embedded market, where AMD used to compete as well, but VIA has no effect on the desktop market. Until ARM processors or VIA enter the desktop market, Intel legally can't do anything to "completely destroy" AMD. In other words, Intel has to make sure AMD can compete in some capacity....


I've never seen the i7 2600K below $300.....and the i5 2500K has never been $200. The i7 2600 was $300 and the i5 2500 was around $200 at one point, but the "K" series processors are at the same price they've been at for the last year.


The Conroe architecture had shorter pipelines and higher IPC than the K8 (9-12IPC) or K10 architecture. Conroe had lower clocks speeds compared to the Northwood or Prescott architectures based on NetBurst. NetBurst had long pipelines, which catered to high clock speeds but suffered from low IPC (4-6) as a result. AMD's "Bulldozer" architecture is very similar to Intel's NetBurst architectures in that is has long pipelines and thus, low IPC.

You can buy i5 2500K for $179.99 in Micro Center, but they only sell in-store only.

 

[sykozis]

You can buy i5 2500K for $179.99 in Micro Center, but they only sell in-store only.

Their closest store is about 3hrs from me....although, I do get a laugh from them comparing their in-store prices to Newegg's more convenient online price. Only way I could possibly get a processor from them is on our yearly family trip to McHenry, MD.

 

[MU_Engineer]

You really don't know enough about the computer industry to make the statements you have. Due to laws, Intel has very little control over the total cost of a computer (Anti-Trust and consumer protection laws). Intel can't "completely destroy" AMD as they are the only processor companies competing in the desktop market. Intel also competes against VIA in the embedded market, where AMD used to compete as well, but VIA has no effect on the desktop market. Until ARM processors or VIA enter the desktop market, Intel legally can't do anything to "completely destroy" AMD. In other words, Intel has to make sure AMD can compete in some capacity....

Intel actually has a pretty large amount of control over the price of a standard computer as the CPU is one of the most expensive parts of the computer, and it is the only piece of computer hardware that is not essentially a commodity. You have Microsoft to thank for that one as the Windows monopoly locked in Intel's x86 architecture as THE only CPU architecture anybody wanted for consumer machines. It takes access to Intel's patents to make an x86 CPU and Intel holds onto those with a death grip. Only Intel, AMD and VIA make x86 CPUs, and AMD and VIA were embroiled in massive, recurrent lawsuits to be able to do so. And even with that, VIA's licensing term will be up soon and the only reason Intel let VIA make x86 CPUs is that Intel needed some of VIA's patents. Apparently those VIA patents will expire soon and I predict VIA will be up a creek without a paddle when the current x86 licensing term is up.

What will break Intel's monopoly is the Windows-x86 hegemony breaking. As much as they suck, smartphones and tablets running non-x86/non-Windows and/or Web-based programs is what is going to do this. Once "Windows" and "computing" are no longer synonymous, Intel will be forced to compete with the multitude of non-x86 CPU makers instead of just AMD and VIA. I'd love to see Intel take on somebody more their own size like Samsung. That would be a good battle, I'll bring the popcorn!

The Conroe architecture had shorter pipelines and higher IPC than the K8 (9-12IPC) or K10 architecture. Conroe had lower clocks speeds compared to the Northwood or Prescott architectures based on NetBurst. NetBurst had long pipelines, which catered to high clock speeds but suffered from low IPC (4-6) as a result. AMD's "Bulldozer" architecture is very similar to Intel's NetBurst architectures in that is has long pipelines and thus, low IPC.

Conroe actually had a longer pipeline than K8 and K10. Conroe and Nehalem had 14 stages and K8/K10 had 12 stages. We don't know how many Bulldozer has, except that it is more than 12. I'd bet somewhere around 20 based on its frequency and behavior in certain applications. I've also seen no figures given for Sandy Bridge, except that it has more than Conroe/Nehalem. My guess is 15-17.

Also, the actual IPC of any of those CPUs is well under 3, generally on the order of 1.0-1.5. You never get to keep the pipeline 100% full all of the time and retire the maximum theoretical number of instructions every clock cycle. Pipeline stalls due to dependencies, branch mispredictions, and cache misses prevent you from reaching maximum performance. Otherwise, we'd all be using massively wide, in-order units like Itanium. It has a maximum IPC of 8 compared to 3-4 for current x86 CPUs. The reason that long pipeline can hinder performance is that you may have to completely flush it and then refill if you mispredict a branch, etc. Your 20-stage 4.0 GHz CPU takes longer to refill the pipeline and finish operations than a 10-stage 2.5 GHz CPU. However, pipelining is not all bad, you can run the CPU much faster and get much greater throughput with a full pipeline with a long-pipelined, high-clocked CPU than a short-pipelined, low-clocked one. It's just that delicate balance that you have to reach, and it keeps changing. The better you get at minimizing stalls and misses (better branch prediction, bigger/better caches), the longer you can make the pipeline without seriously degrading performance in less than optimal code and the higher peak performance your CPU can attain. That's why we are seeing the number of stages slowly increase, from 10 in the PIII/Athlon to closer to 20 in Sandy Bridge without going the way of the P4's poor performance.

 

[Chad Boga]

Sandy Bridge is the biggest change from Core 2 since core 2 hit.

Why do you say this?

I would have thought the IMC and point-to-point processor interconnect of Nehalem is a bigger jump than anything being done in SB.

 

[jimmysmitty]

Why do you say this?

I would have thought the IMC and point-to-point processor interconnect of Nehalem is a bigger jump than anything being done in SB.

The IMC was a big change, but I meant in terms of the actual core arch. Overall Nehalem was the Cor 2 arch with tweaks and a IMC. SB on the other hand was redesigend, the OoO engine was changed a lot, same with the cahce system and the new ring bus CPU connection.

Either way though Intel is still using the base Core design and I doubt they will change that unless they find something beter.

 

[MU_Engineer]

Why do you say this?

I would have thought the IMC and point-to-point processor interconnect of Nehalem is a bigger jump than anything being done in SB.

The IMC was a big change, but I meant in terms of the actual core arch. Overall Nehalem was the Cor 2 arch with tweaks and a IMC. SB on the other hand was redesigend, the OoO engine was changed a lot, same with the cahce system and the new ring bus CPU connection.

The ring bus actually made its debut in the Nehalem-EX Xeon 6500/7500 CPUs. I do agree though, Nehalem is essentially Core + IMC + HyperThreading, while Sandy Bridge makes a few more tweaks to the underlying microarchitecture.

Either way though Intel is still using the base Core design and I doubt they will change that unless they find something beter.

It goes even farther back than that. Core (as in Core 2) is the Pentium M/Core (Core as in the original "Yonah" Core Duo) P6+ design modified a bit to handle 64-bit operation, beef up FPU power, and the ability to issue a fourth instruction per clock. P6+ is the original Pentium Pro/PII/PIII P6 design changed to add updated SIMD, op fusion, and the P4's QDR FSB. So in essence, Sandy Bridge is a (highly) tweaked Pentium Pro.

As far as going with something entirely new, I wouldn't bet on it. The last three attempts Intel had at that turned out fairly poorly. Itanium sunk. NetBurst flamed out. Atom wasn't intended to be a very powerful chip, but it didn't even hit those targets very well and gets its butt kicked by AMD's Bobcat and isn't even all that much better than far cheaper and cooler-running ARM CPUs. I'd suggest Intel stick with continuing to modify their tried and true P6+++ architecture 😉

 

[Chad Boga]

The IMC was a big change, but I meant in terms of the actual core arch. Overall Nehalem was the Cor 2 arch with tweaks and a IMC. SB on the other hand was redesigend, the OoO engine was changed a lot, same with the cahce system and the new ring bus CPU connection.

Either way though Intel is still using the base Core design and I doubt they will change that unless they find something beter.

The ring bus actually made its debut in the Nehalem-EX Xeon 6500/7500 CPUs. I do agree though, Nehalem is essentially Core + IMC + HyperThreading, while Sandy Bridge makes a few more tweaks to the underlying microarchitecture.

It is a strange thing though when the clock per clock advantage Nehalem has over C2Q, is bigger than the clock per clock advantage SB holds over Nehalem.