AMD Bulldozer Finally!

[Martini1991]

Well people obviously aren't going to agree, fair enough. No one said "multi-threading is niche", rather the FX8150, I think that'd be evident by the lacklustre performance in pretty much everything real world (Except for when you need that many cores).

words cannot even describe the ammount of non-factual amd fanboy talk in that post haha

don't get me wrong, i love amd as well, but seriously? xD

?

 

[aln]

Well people obviously aren't going to agree, fair enough.

?

People don't need to agree, but if you somehow manage to see value in the other side of the argument then maybe things aren’t so bad for AMD. As I mentioned in the other thread, I'm a pretty hardcore competitive gamer, and I like to play with stupidly high FPS. I crave a 2500k for this reason, but I'm just not convinced it's a worth while upgrade, when anything that matters will be better on BD. On the other hand I don't like the idea of buying a chip with such relatively poor ST performance. I can't justify my reasoning with any actual facts, but I really don't want to buy the CPU with the current state of things.

words cannot even describe the ammount of non-factual amd fanboy talk in that post haha

don't get me wrong, i love amd as well, but seriously? xD

Is this referring to me? I gave my opinion on the matter, I'm happy to be proven wrong, I just don't think the answer is actually definitive to achieve that.

 

[bhavv]

I looked at the reviews.

I thought ... 'Meh, dammit AMD ... Intel I7 Hex Core prices arent going to come down at all with that crap'.

Multithreading is "niche"? Are you guys still running MS-DOS or something?

Well, 6+ cores is a niche, and so is hyperthreading.

99% of people dont need anything more than an I5 2500k atm, and the Bulldozer doesnt even beat that.

 

[kitch9]

There is one program which I find infuriating: MS Outlook.
I have about 10 different email accounts, which I do an automatic send and receive for, every 5-10 minutes. During the send/receive, the entire program locks up. This is a single threaded program. My dual core cpu goes to 50% during the send/receive, while the other 50% is untouched.

I'm developing a major program right now and before I started coding it with multiple threads, the GUI would lock up regularly, while some cpu intensive operations were being processed. Multi-threading is definitely the way to go. I can't understand why other developers dont make use of heavy multi-threading. I'm stunned that many games don't use more than 2-4 threads. By now, I would've thought that most games would be using 100+ threads.

At the minute, games tend to have a main thread and then seperate lighter threads for audio, physics, etc etc.

There is a technique being worked on that rather then allocating specific jobs to a specific thread you instead have one master thread that is purely for distributing calculations regardless of what they are for to cores that are least loaded to spread the entire workload evenly across the CPU.

That in theory would utilise as many cores as you had, but I've not heard much about it for a while.

 

[Predicament]

Multithreading is "niche"? Are you guys still running MS-DOS or something? I can't be bothered to post a screenshot but I just took a peek in the process tab of my task manager. You know how many threads Chrome alone is running?

54!

One does not need to spend £300 on a FX motherboard/CPU combo in order to run google chrome.

 

[Zuban]

Problem is that apps that need that sort of threading will often need FP precision to match and BD lacks in that department, so other than web serving I can't see what good it is even in highly 6+ threaded apps.

I think this is the case with what i do, i was hoping bulldozer with its 8 threads and presumably higher overclocking headroom would be a decent increase over my 1090t, but if you look at this chart, even though it has an extra 2 cores over the 1100t, its barely faster at all, rendering is the only one that really matters there, lightwave 3d's unusual in that its seem to always have been good on amd chips, which is why i bought an x6 at the time. I would have much prefered an optimized die shrink to 32nm of the phenom II than bulldozer at this point!..

http://benchmarkreviews.com/index.p...k=view&id=831&Itemid=63&limit=1&limitstart=11

 

[Destination]

One does not need to spend £300 on a FX motherboard/CPU combo in order to run google chrome.

And if one does spend the cash, bulldozer will ensure it doesn't run any better than it did before, it might even run chrome more slowly, good work BD

Just as well the chip turned out to be terrible for the general user, else the issue of yields and availability would really have come to the fore.

 

[bhavv]

And if one does spend the cash, bulldozer will ensure it doesn't run any better than it did before, it might even run chrome more slowly, good work BD

Just as well the chip turned out to be terrible for the general user, else the issue of yields and availability would really have come to the fore.

As terrible as it may be for the 'general user', those are the people that will think that 8 cores must obviously be better than 4 or 6, simply because its a higher number

Looks like I will be using my I7 920 for a lot lot lot longer, I will probably be waiting until 16 nm CPUs until there is something significantly better that is worth getting a new mobo out.

 

[eddyr]

Thats just it , it isn't terrible at all for the general user. They won't notice any difference in perhaps anything but the electricity bill, and it will be slight at that. The bottlenecks that affect usability or 'experience' are elsewhere.
They just need to sort out their pricing a little, particularily the x4/x6 but yeah, the marketing will work with the core count.

 

[manveruppd]

One does not need to spend £300 on a FX motherboard/CPU combo in order to run google chrome.

Everyone here knows that, and of course only 2-3 of those threads are actually using any significant CPU time at any given time, but I thought that in a forum full of techies and enthusiasts I wouldn't need to point out the obvious:

Do you know how Anandtech, Tom's Hardware etc. run benchmarks? Install Windows, install drivers, install benchmarks (whether that's a synthetic benchmark or a game), run benchmark. So if they're testing Crysis, all that's running is Windows and Crysis.

Now that's as it should be of course, as it gives us an unadulterated view of a system's capabilities, but is it real world use? We all have 4+GB RAM these days, do you exit Chrome when you launch Crysis? Do you exit MSN? Dropbox? Evernote? uTorrrent? The little Gmail notifier thingy? The billion little other things you've got running there (many of which you don't even know are running, like Adobe's virulently annoying autoupdater)?

No of course you don't - our systems aren't the pure clean installs that people bench on, they're a torrid mess of crapware all checking things online occasionally, generating random threads, downloading stuff, indexing your hard drive so advertisers can spy on you, updating themselves, popping up messages from your friends and a billion other things.

Now, all this stuff is lightweight and occasional, and MOST of the time a CPU with 4 fast threads would be able to do them in the background without affecting your framerate in-game. But that's not to say that there aren't occasions where more than 4 cores (or more than 4 threads on CPUs with Hyperthreading) wouldn't come in handy. Say you're installling 26GB of Rage, you know it's gonna take awhile, wouldn't it be nice to launch some other game in the meantime? Or extract an archive? Or transcode a video? I'm sure even those of us with i5s have experienced slowdown when doing more than 1 CPU-intensive thing at a time. You say "why would you? who uses their PC like that?" Well, it's not NECESSARY to be able to use it like that, but it's not necessary to play Rage on HD resolutions either is it? I can still play Quake 2 on my dad's ancient Pentium II. But if you CAN do all this stuff, why wouldn't you?

If you're talking about "need", then yes, only a tiny niche market "NEEDS" 8-threaded integer processing. But for a lot of people it will be nice to have, and it will make a mild but noticeable improvement in their computing experience. I'm not just talking about Bulldozer here either - if this weren't the case, a lot fewer people would be buying 2600Ks - why would you when the 2500K is identical in every way apart from Hyperthreading?

 

[Gashman]

insane thing is, we are all most of the time upgrading just for the sake of upgrading. had a think last night about it, wanted to originally put a Bulldozer is this system until it turned out to be no faster (in some cases slower) than the X6 I have in here already, thought to myself what does my X6 struggle to handle in the first place? and the answer to this question was nothing. so I ask myself why do I need to change the processor? and the realistic answer was that I don't.

further more would be absolutely shocked that someone using a Bloomfield processor could justify to themselves why switching to Sandy Bridge processor is really worth the trouble, does Bloomfield struggle with anything, again the answer is no so why are so many people so eager to upgrade? could understand if the processor in question was a Cedar Mill or something along those lines, then fair enough Cedar Mill > Sandy Bridge is a rather impressive increase in performance, but Bloomfield > Sandy Bridge or Phenom II > Bulldozer, the differences in performance just aren't worth it not when you consider that a lot of the time your Bloomfield or Thuban is sitting around doing nothing, why upgrade to a processor that is going to face the exact same problem?

in the desktop market its not the processors that need to improve, its the programs using them more than anything, make better use of the hardware we have rather than continually improving hardware to brute force programs into running smoother. also rather than use incredible amounts of resources and cash to design and eventually fabricate Bulldozer, why on Earth didn't AMD improve K10.5 since there are improvements to be made, speed up the L3 cache, improve the north-bridge, make it more power efficient, then chuck two more cores on and fabricate it using 32NM process, the result a processor that is better overall than Bulldozer and you can call it K11 or something!

look at Llano as an example, it already has an IPC increase on Phenom II, combine that with a snappy L3 cache and high speed north-bridge and it would be a good chip, since you can already get ~10 - 15% increases in some programs with Phenom II by simply overclocking the north-bridge, and Llano has an increase off the bat, so could be talking about getting another ~25% out of Phenom II by simply making some architectural improvements, all the effort that went into Bulldozer and it couldn't have a chance at competing against an eight-core Phenom II with a 25% increase in current IPC and possibly 25% increase in clock speeds, that would crush Bulldozer.

 

[Final8y]

Everyone here knows that, and of course only 2-3 of those threads are actually using any significant CPU time at any given time, but I thought that in a forum full of techies and enthusiasts I wouldn't need to point out the obvious:

Do you know how Anandtech, Tom's Hardware etc. run benchmarks? Install Windows, install drivers, install benchmarks (whether that's a synthetic benchmark or a game), run benchmark. So if they're testing Crysis, all that's running is Windows and Crysis.

Now that's as it should be of course, as it gives us an unadulterated view of a system's capabilities, but is it real world use? We all have 4+GB RAM these days, do you exit Chrome when you launch Crysis? Do you exit MSN? Dropbox? Evernote? uTorrrent? The little Gmail notifier thingy? The billion little other things you've got running there (many of which you don't even know are running, like Adobe's virulently annoying autoupdater)?

No of course you don't - our systems aren't the pure clean installs that people bench on, they're a torrid mess of crapware all checking things online occasionally, generating random threads, downloading stuff, indexing your hard drive so advertisers can spy on you, updating themselves, popping up messages from your friends and a billion other things.

Now, all this stuff is lightweight and occasional, and MOST of the time a CPU with 4 fast threads would be able to do them in the background without affecting your framerate in-game. But that's not to say that there aren't occasions where more than 4 cores (or more than 4 threads on CPUs with Hyperthreading) wouldn't come in handy. Say you're installling 26GB of Rage, you know it's gonna take awhile, wouldn't it be nice to launch some other game in the meantime? Or extract an archive? Or transcode a video? I'm sure even those of us with i5s have experienced slowdown when doing more than 1 CPU-intensive thing at a time. You say "why would you? who uses their PC like that?" Well, it's not NECESSARY to be able to use it like that, but it's not necessary to play Rage on HD resolutions either is it? I can still play Quake 2 on my dad's ancient Pentium II. But if you CAN do all this stuff, why wouldn't you?

If you're talking about "need", then yes, only a tiny niche market "NEEDS" 8-threaded integer processing. But for a lot of people it will be nice to have, and it will make a mild but noticeable improvement in their computing experience. I'm not just talking about Bulldozer here either - if this weren't the case, a lot fewer people would be buying 2600Ks - why would you when the 2500K is identical in every way apart from Hyperthreading?

Bulldozer/ 8 cores aside you make some valid points and that's why i really like my hex because 4 cores was slowing me down even when my PC boots up the hex is faster, its self being on 100% usage on boot as there is a lot loading at that time.

I don't close nothing down for gaming and sometimes ALT TAB to get other things going then ALT TAB back into the game, have even run 2 games at the same TMS and LFS, don't ask me to explain that one.

When i use DVD for games or software i would install simultaneity if the installers didn't block each other and keep playing while installing more.

Having more than one screen does tend make you want to do more at once.

 

[Final8y]

insane thing is, we are all most of the time upgrading just for the sake of upgrading. had a think last night about it, wanted to originally put a Bulldozer is this system until it turned out to be no faster (in some cases slower) than the X6 I have in here already, thought to myself what does my X6 struggle to handle in the first place? and the answer to this question was nothing. so I ask myself why do I need to change the processor? and the realistic answer was that I don't.

further more would be absolutely shocked that someone using a Bloomfield processor could justify to themselves why switching to Sandy Bridge processor is really worth the trouble, does Bloomfield struggle with anything, again the answer is no so why are so many people so eager to upgrade? could understand if the processor in question was a Cedar Mill or something along those lines, then fair enough Cedar Mill > Sandy Bridge is a rather impressive increase in performance, but Bloomfield > Sandy Bridge or Phenom II > Bulldozer, the differences in performance just aren't worth it not when you consider that a lot of the time your Bloomfield or Thuban is sitting around doing nothing, why upgrade to a processor that is going to face the exact same problem?

in the desktop market its not the processors that need to improve, its the programs using them more than anything, make better use of the hardware we have rather than continually improving hardware to brute force programs into running smoother. also rather than use incredible amounts of resources and cash to design and eventually fabricate Bulldozer, why on Earth didn't AMD improve K10.5 since there are improvements to be made, speed up the L3 cache, improve the north-bridge, make it more power efficient, then chuck two more cores on and fabricate it using 32NM process, the result a processor that is better overall than Bulldozer and you can call it K11 or something!

look at Llano as an example, it already has an IPC increase on Phenom II, combine that with a snappy L3 cache and high speed north-bridge and it would be a good chip, since you can already get ~10 - 15% increases in some programs with Phenom II by simply overclocking the north-bridge, and Llano has an increase off the bat, so could be talking about getting another ~25% out of Phenom II by simply making some architectural improvements, all the effort that went into Bulldozer and it couldn't have a chance at competing against an eight-core Phenom II with a 25% increase in current IPC and possibly 25% increase in clock speeds, that would crush Bulldozer.

+1

 

[manveruppd]

in the desktop market its not the processors that need to improve, its the programs using them more than anything, make better use of the hardware we have rather than continually improving hardware to brute force programs into running smoother. also rather than use incredible amounts of resources and cash to design and eventually fabricate Bulldozer, why on Earth didn't AMD improve K10.5 since there are improvements to be made, speed up the L3 cache, improve the north-bridge, make it more power efficient, then chuck two more cores on and fabricate it using 32NM process, the result a processor that is better overall than Bulldozer and you can call it K11 or something!

You're totally right, a souped-up die-shrink of Phenom II would've made a better desktop processor, but most of AMD's business is in servers and laptops these days, and the BD design is just more power efficient so more suitable for that. Or it should be at least - currently at load it sucks down electricity like Megatron on a bender in a wind farm, but I assume that's partly due to Global Foundries not having gotten the manufacturing process just right, so should improve. At idle it's already lower than Sandybridge despite having twice the transistor count, which is pretty impressive, so they're halfway there. Trinity should be promising for subnotebook devices.

Bottom line is they're a small company (comparatively speaking), they probably didn't have 2 design teams able to one work on a server CPU and one on a desktop one. At the price they pitched it it's still competitive though, even though Sandybridge is superior in most things.

 

[manveruppd]

When i use DVD for games or software i would install simultaneity if the installers didn't block each other and keep playing while installing more.

Hehe I do that even though I'm only on a C2D. In fact I do it a lot less than I used to cause installing things is reasonably fast on this, but I used to do it a lot back when I had a single core (Athlon XP 1700+, then moved to a 2500+ XP-M @2.2). Since everything was going on a single hard disk too it probably slowed things down MASSIVELY, but I still did it, even when it led to everything crashing and burning - I was just too impatient!

 

[Final8y]

Hehe I do that even though I'm only on a C2D. In fact I do it a lot less than I used to cause installing things is reasonably fast on this, but I used to do it a lot back when I had a single core (Athlon XP 1700+, then moved to a 2500+ XP-M @2.2). Since everything was going on a single hard disk too it probably slowed things down MASSIVELY, but I still did it, even when it led to everything crashing and burning - I was just too impatient!

LOL, I'm a patient man but if i think i can get away with more at once i will.

 

[Martini1991]

Lets see if the fx8150 actually turns up today. My vote's on no.

 

[CAT-THE-FIFTH]

Bulldozer is actually smaller than Thuban and IIRC they are only buying usable chips from Global Foundries,meaning the latter is probably taking the hit.

The FX4100 uses the same die as the FX6100 and FX8100 series and is around £90 to £95. The FX8120 is around £160 to £170. I suspect they could make a profit even selling the FX8150 at £140.

Bulldozer FX 8150 is two billion transistors...

SB 2500k is less than half of that.
The Phenom II 1100T is over 900 million transistors.

Thuban is 346MM2 and Bulldozer is 315MM2.

 

[Martini1991]

Thuban is 346MM2 and Bulldozer is 315MM2.

I was talking transistor count, but which do you think is harder to make, Thuban or Bulldozer? That'll be Bulldozer, which is running hotter and consuming more power with horrendous yields? Bulldozer, which would make I assume the die size frugal in this situation.

 

[CAT-THE-FIFTH]

I was talking transistor count, but which do you think is harder to make, Thuban or Bulldozer? That'll be Bulldozer, which is running hotter and consuming more power with horrendous yields? Bulldozer, which would make I assume the die size frugal in this situation.

Again they are paying for usable chips as been mentioned in multiple news articles. So it is Global Foundries which is taking the main hit here,not AMD.

On top of this it really seems they are diverting a lot of Bulldozer production towards commercial customers. They are supplying 38400 16 core Bulldozer CPUs to Cray for Oak Ridge National Laboratory. That basically means nearly 77000 8 core Bulldozer dies just for one customer and that was known before launch.