Why would/did you buy an AMD CPU?

[nwo]

I just picked up an OEM FX 4130 for $40 at MC. I think that speaks for itself considering the only other CPUs in that price range are dual core Intel celerons or a single core Sempron.

 

[sefsefsefsef]

Interesting is that both processors that used that shared units approach, Rock and Bulldozer, needed high clock speeds and had a huge problem with power consumption.

These are totally separate issues. CMT is a die area saving technique, probably reduces power consumption, and it's totally separate from frequency scaling. It is not a fundamentally flawed approach, it's just that the real world examples we have of it so far have other shortcomings, and we tend to blame every problem with the design on every aspect of the design, which is not logically sound.

 

[Zor Prime]

I buy AMD gear to help keep them in business so that Intel trolls can pay lower prices.

 

[ColonelBlimp]

I'm using a Phenom X4-965be at 3.8ghz only because I had an AM3 board with a X3-435 on it. Couldn't afford a new mobo at the time and the socket won't take an FX.
Next Xmas though, I'm looking to get Mobo/CPU and SSD so I'm curious how the land will lie at that time.
Would still like to go AMD if possible, plucky underdog and all that, but I'm not stupid, it's looking like an i5 K for me this far out.

 

[mrmt]

These are totally separate issues. CMT is a die area saving technique, probably reduces power consumption, and it's totally separate from frequency scaling. It is not a fundamentally flawed approach, it's just that the real world examples we have of it so far have other shortcomings, and we tend to blame every problem with the design on every aspect of the design, which is not logically sound.

I cannot say much about Rock's die size, but AMD's chips aren't exactly stars here. Richland CPU part alone is bigger than the entire Sandy Bridge 2C die (GPU + CPU). How's CMT can be remotely considered die area saving?

 

[ElFenix]

I cannot say much about Rock's die size, but AMD's chips aren't exactly stars here. Richland CPU part alone is bigger than the entire Sandy Bridge 2C die (GPU + CPU). How's CMT can be remotely considered die area saving?

comparing what intel is able to do to what AMD is able to do is worse than apples to oranges. intel is selling fruit and AMD is closer to froot loops.


it's die saving compared to what AMD might have been able to do with 4 fully fleshed-out cores offering similar performance (maybe). intel might be able to have made it work well, but intel doesn't need to do novel solutions to problems. AMD does, but AMD doesn't have the engineering resources to pull it off.

 

[bgt]

I prefer the fx8350 for general computer use , just feels erm more snappy. It uncompresses WinRAR files quicker than the i7 for sure
I am often amused when people say AMD make rubbish CPU's. Not for gaming they don't. I mean the i7 is better but it's not £100 better

This:thumbsup:

 

[sefsefsefsef]

it's die saving compared to what AMD might have been able to do with 4 fully fleshed-out cores offering similar performance (maybe). intel might be able to have made it work well, but intel doesn't need to do novel solutions to problems. AMD does, but AMD doesn't have the engineering resources to pull it off.

Actually, you have the reasoning a little backwards. Since AMD has fewer R&D resources than Intel, they are forced to try crazy stuff every now and then, and hope that one of these stunts pays off and they get huge returns. They have to behave somewhat like a startup in this regard.

 

[SolMiester]

Not even AMD use there own CPUs for GPU benchmarks....Nuff said really!

 

[Elixer]

Not even AMD use there own CPUs for GPU benchmarks....Nuff said really!

Pretty much this, if they had a decent high-end CPU, then, they would be using it.

AMD has basically abandoned the high-end CPU, heck, some will argue mid-range as well, and are squarely in the low-end area.

Their whole business model has changed.
Back in the good ole days, they actually put up a fight against intel's flagship, and they even won lots of those battles.
Sadly, this hasn't been the case for many, many years now.

There is nothing on their roadmap that even suggests they will get out their disastrous bulldozer micro-architecture roll out, and until they wake up and drop their own NETBURST fubar, AMD will not have a high-end CPU that can compete to intel's flagship ever again.

Look at what they are doing, building ARM+GPU now. They see the writing on the wall, so they figure they will do stuff they are good at (GPU/APU), and marry it to ARM.

Sad days.

 

[Headfoot]

Why are there so many threads like this right now? What does that possibly achieve?

 

[ElFenix]

Actually, you have the reasoning a little backwards. Since AMD has fewer R&D resources than Intel, they are forced to try crazy stuff every now and then, and hope that one of these stunts pays off and they get huge returns. They have to behave somewhat like a startup in this regard.

i didn't say there was any order involved in the reasoning.

 

[dennilfloss]

I did my build in summer 2009 at the time the 955BE came out and it was a pretty decent CPU at the time. I had been using AMD since I got a 1GHz tbird from Compuwiz1 at the turn of the century so I was more familiar with AMD. Ditto for Radeon, I've been using those since the original Radeon 64 replaced my Diamond Viper 770 nonUltra.

 

[mrmt]

comparing what intel is able to do to what AMD is able to do is worse than apples to oranges. intel is selling fruit and AMD is closer to froot loops.

Ok, what about comparing Trinity/Richland against Llano? Trinity CPU part isn't significantly smaller than Llano. In fact, Trinity's CPU part is slightly larger. So where's the die-saving space that should have been provided by the CMT concept?

 

[dave_the_nerd]

Why are there so many threads like this right now? What does that possibly achieve?

Trolling, probably.

 

[ElFenix]

Ok, what about comparing Trinity/Richland against Llano? Trinity CPU part isn't significantly smaller than Llano. In fact, Trinity's CPU part is slightly larger. So where's the die-saving space that should have been provided by the CMT concept?

maybe you should read the rest of my post.


trinity is faster than llano, for the most part.

 

[coercitiv]

Some data to chew on, transistor counts at the end. GPU die area percentage is estimated, on Llano and Trinity I had to compare area myself. Some of the transistor counts may differ (schematic vs layout)

32nm
Intel Sandy Bridge (4C) 216mm2 17% GPU - 1.16B
AMD Llano 228mm2 35% GPU - 1.18B
AMD Trinity 246mm2 37% GPU - 1.3B

28nm
AMD Kaveri 245mm2 47% GPU - 2.41B

22nm
Intel Ivy Bridge (4C) 160mm2 27% GPU - 1.2B
Haswell GT2 (4C) 177mm2 33% GPU - 1.4B
Haswell GT3 (4C) 260mm2 65% GPU - 1.7B

 

[mrmt]

maybe you should read the rest of my post.


trinity is faster than llano, for the most part.

I did read. Trinity is slightly faster than Llano, but not always and not that much in most cases (we should exclude the benches that use instructions not supported by Llano, correct?) and in others there are performance regressions. Not good for a chip that should be faster, if anything because AMD is throwing more hardware at the problem (more transistors in a bigger die size) and is a core that received the bulk of the R&D budget in the last 5 years of the company, while stars was then a 10 years old core bolted with new instructions and power gating.

This thinking of "CMT will save you die size" seems more and more like IBM claims that gate first will result in smaller dies. IDC showed that in AMD's case gate first worked the other way around, it warranted a bigger die than it would warrant with gate last.

 

[SammichPG]

I built a fairly economical rig for a guy recently. I put an i3 in it. I really couldn't find anything better for the money on the AMD side.

And once you go high-end, AMD wouldn't enter the thought process at all.

So what are they offering now? The last time I had an AMD rig was when they were offering a chip that really competed.

Virtualization extensions. Amd has enabled them even on zacate, it's awesome.

Pentiums and celerons are up to the task, but running a vm without vt-x? Just too slow.

I'm running a debian host and a monowall guest (and planning to add another guest to experiment with dydns and webhosting) on an amd laptop.
I'm so glad I didn't pick the model using an intel pentium, sure it's been hotter (45nm mobile phenoms heh) but the igp was faster and now that the laptop is no longer useful as a mobile machine it can be repurposed as a small home server.

Feature segmentation from Intel can bite you in the ass down the line expecially on budget machines.

 

[Enigmoid]

comparing what intel is able to do to what AMD is able to do is worse than apples to oranges. intel is selling fruit and AMD is closer to froot loops.


it's die saving compared to what AMD might have been able to do with 4 fully fleshed-out cores offering similar performance (maybe). intel might be able to have made it work well, but intel doesn't need to do novel solutions to problems. AMD does, but AMD doesn't have the engineering resources to pull it off.

Phenom x6 - 45nm - 346 mm^2
8150 - 32 nm - 315 mm^2

Considering how the 8150 is only minimally better than the Phenom x6 the decrease in die size does not make up for the possible gains from the process advantage. A phenom x8 on 32 nm would consume similar amounts of power and be faster for a similar die size.

I did read. Trinity is slightly faster than Llano, but not always and not that much in most cases (we should exclude the benches that use instructions not supported by Llano, correct?) and in others there are performance regressions. Not good for a chip that should be faster, if anything because AMD is throwing more hardware at the problem (more transistors in a bigger die size) and is a core that received the bulk of the R&D budget in the last 5 years of the company, while stars was then a 10 years old core bolted with new instructions and power gating.

This thinking of "CMT will save you die size" seems more and more like IBM claims that gate first will result in smaller dies. IDC showed that in AMD's case gate first worked the other way around, it warranted a bigger die than it would warrant with gate last.

Its pretty bad when you consider you went from 2.7 Ghz to 3.8/4.2 Ghz. IPC most definitely went down.

 

[TeknoBug]

Oh yeah later Phenom II's (Denub and Thuban) had faster IPC than Bulldozer, but about on par with Piledriver. I still ask myself today to why I didn't stick to my X6 1090T when I picked up an 8320 then a 4350.

One nice advantage with the 4350 over 1090T is it runs much cooler.

 

[Deceneu]

I buy AMD because is cheaper ,simple as that.And it gets the job done for less money.

If on an online store in my country(a UE country) i choose to buy a quad core ,i would have to choose between a AMD quad an a Intel quad that has double the price.

As in most casual applications the extra mega Intel superiority is not observable i think choosing a AMD CPU + a quality mobo for the same money ,that i would pay for an Intel CPU counter part ,is the logical choice.

I can give examples at any moment linking to some online stores if needed

PS:I hope Mr Stahlhart will not warn me again for expressing my own opinion in a civilized manner.

 

[ElFenix]

I did read. Trinity is slightly faster than Llano, but not always and not that much in most cases (we should exclude the benches that use instructions not supported by Llano, correct?) and in others there are performance regressions. Not good for a chip that should be faster, if anything because AMD is throwing more hardware at the problem (more transistors in a bigger die size) and is a core that received the bulk of the R&D budget in the last 5 years of the company, while stars was then a 10 years old core bolted with new instructions and power gating.

This thinking of "CMT will save you die size" seems more and more like IBM claims that gate first will result in smaller dies. IDC showed that in AMD's case gate first worked the other way around, it warranted a bigger die than it would warrant with gate last.

Phenom x6 - 45nm - 346 mm^2
8150 - 32 nm - 315 mm^2

Considering how the 8150 is only minimally better than the Phenom x6 the decrease in die size does not make up for the possible gains from the process advantage. A phenom x8 on 32 nm would consume similar amounts of power and be faster for a similar die size.

do i really have to set this out for you guys?

it's die saving compared to what AMD might have been able to do with 4 fully fleshed-out cores offering similar performance (maybe). intel might be able to have made it work well, but intel doesn't need to do novel solutions to problems. AMD does, but AMD doesn't have the engineering resources to pull it off.

 

[CHADBOGA]

Looking at AMD's current line up, the circumstances where AMD's offerings make any sense, are such tiny niches, that I can only imagine that anyone going with AMD CPU's today, are doing so either out of sheer ignorance or a warped sense of "morality".


First, thread crapping is not allowed. Second, you insulted a large number of users, myself included.
Knock it off
Markfw900
Anandtech moderator

Because I hate Intel and because I believe in the direction AMD is headed, and I'm willing to kick them a little coin on their way to getting there.

Can't handle the TRUTH, huh?


The mod callout gets you an infraction
Markfw900

 

[Remobz]

I buy AMD because is cheaper ,simple as that.And it gets the job done for less money.

If on an online store in my country(a UE country) i choose to buy a quad core ,i would have to choose between a AMD quad an a Intel quad that has double the price.

As in most casual applications the extra mega Intel superiority is not observable i think choosing a AMD CPU + a quality mobo for the same money ,that i would pay for an Intel CPU counter part ,is the logical choice.

I can give examples at any moment linking to some online stores if needed

PS:I hope Mr Stahlhart will not warn me again for expressing my own opinion in a civilized manner.

I bought a 14 inch laptop with an AMD A6 mobile chip. Bought it for basic computing and internet use and it gets the job done. Saved a lot of money as well.