How will the bulldozer be better then the i7?

It was mentioned several times in the threads that BD should work in AM3 mobo's with a BIOS update.

I expect that will make it a poor performer against the high end 2011 chips especially if AMD thinks it will use more cores to make up an IPC disadvantage.

Either way there most likely will be a quad channel bulldozer offering so quad channel chips from both companies can compete against each other.
 
Current Quads and Hexa core chips don't seem to be all that limited by dual channel RAM, at least for consumer desktop stuff anyways. For servers it will of course be a different story altogether.

I think BD is just going to be a more server oriented chip than previous designs. It should still work well enough for consumer apps, so long as they price it right and it isn't crap vs. Nehalem and Sandy Bridge it'll sell.
 
As long as they keep rocking the value market it won't matter.
Agreed. AMD will have a firm footing for a while if they keep coming out with great values. The Athlon II series is a great example of that. Even the second highest 1055T @ $200 is a great example of that as well.
 
I expect that will make it a poor performer against the high end 2011 chips especially if AMD thinks it will use more cores to make up an IPC disadvantage.

Either way there most likely will be a quad channel bulldozer offering so quad channel chips from both companies can compete against each other.


there will be 2 versions as far as i can tell when me and another guy got in an arguement/heated discussion in another thread about bulldozer.. from the looks of it the am3+(am3r2) will be a 4 core bulldozer(Zambezi) while the unannounced socket probably a consumer branded G34(or similar) socket will have the 8 core quad channel DDR3 bulldozer(Interlagos)..
 
I can't see why so many AMD antagonists think AMD's next generation bulldozer CPU will faulter behind even last gen Intel I7s.
 
I do not think bulldozer will be further behind versus i7 or even sandy bridge. 22nm Ivy Bridge is a different matter but that is too far off to speculate.

To me there is good and bad in the 2 threaded module approach. It allows AMD to put more cores on a die than it would if each core had its own instruction decoder, floating point unit and L2 cache. But then there are 4 integer pipelines per core instead of 2 or 3. Its hard to tell how this will perform especially because the diagrams do not explain the capabilities of each of the 4 pipelines. I mean can they all do complex operations, is one or more of them a load / store? Will AMD use SOI with their 32nm process? If they do not use SOI will their transistors be faster or slower than they are now?
 
Last edited:
Global Foundries is going to use SOI with High-K metal gating for the 32nm and 28nm nodes. This should help AMD keep temps and voltages down. Just look at what Lo-K did for Thuban.
 
I can't see why so many AMD antagonists think AMD's next generation bulldozer CPU will faulter behind even last gen Intel I7s.

Its because of the non-intuitive way they've brought up BD performance so far (ie. "33% more BD cores get you 50% more performance over current MC").
 
It was mentioned several times in the threads that BD should work in AM3 mobo's with a BIOS update. http://www.xtremesystems.org/forums/showpost.php?p=4501313&postcount=146 and here too: http://www.xtremesystems.org/forums/showpost.php?p=4502286&postcount=157

1st slide just says AM3, 2nd says AM3r2, but then the guy who posted it mentions that AM3r2 is just going to be called AM3+ and that AM3+ CPU's will be backwards compatible to AM3 with the BIOS update.

This is what is what Im most hoping for. Its the one thing that keeps me from going with an Intel platform. Every new proc line has a brand spanking new socket so if you want to upgrade, you have to buy a brand new motherboard and re-install everything again. Since I cant afford to build a top shelf gaming rig, I buy more budget parts and upgrade as prices come down so the ability to upgrade processors just by dropping them in is a big selling point.
 
Amd has the budget line on lock, if they can pick up the mid to top line they'd be in good shape
 
low end will be Llano so die shrinked Phenom II with gpu glued together.

Nothing to get excited.
 
a little to be excited over.....we are going to have the CPU power we have now, plus a more powerful GPU than Intel has integrated....enthusiasts may not go crazy for it, but OEMs will probably go ga ga over them.

(sorry if this point was made, been only briefly checkin the boards over the last three weeks and didnt read all three pages)
 
id like to see leaked numbers and lets compare Bulldozer to sandy bridge amd needs to fight intels next gen not the current gen to stay relevant
 
Last edited:
Perhaps ( 1 year from a Q3 2011 release date ) the chips clock so low that benchmarks would not look good.
 
Last edited:
Bulldozer will require a new socket, AM3+, to add features to enhance performance. It will be backwards compatible with AM3 processors, however. I would at worst expect parity with Nehalem parts and at best it should give Sandy Bridge a run for the money.
 
Amd has the budget line on lock, if they can pick up the mid to top line they'd be in good shape

I don't think they're targetting consumers at all with Bulldozer. It might do okay there, but I think trying to fight Intel on this ground is a war they can't afford right now. They're behind on process and the underdog in a big way. BD is targeted squarely at Xeon and the server market, and given the available information, I think it should perform very well for the kind of workloads those machines see. For that kind of work I expect the extra integer units can give close to the performance of a complete core for only 20% the die space (and power). That's a huge gain, with the big assumption that AMD has tuned the caches and front end well enough to keep the pipelines full.

Will it trump Sandy Bridge on single-thread performance? Probably not (though on FP workloads there's a possibility, BD apparently has a very strong FP core to try to make up for sharing it with two pipelines). However my feeling is it stands a good chance of beating it in overall performance/watt and performance/socket which are the metrics that matter in the high-margin server space. If they do well in the consumer market that'd be great, but they're really going after the higher margins on server CPUs. They might also be intentionally targeting smaller, higher-margin markets with these parts because they may not have the fab capacity, and will probably have low yields to start with. It does seem to make financial sense to stay out of the midrange, and stick to the value parts on previous-gen fab technologies where they have lots of capacity and the high-margin server stuff on the new, low yield fabs.
 
there will be 2 versions as far as i can tell when me and another guy got in an arguement/heated discussion in another thread about bulldozer.. from the looks of it the am3+(am3r2) will be a 4 core bulldozer(Zambezi) while the unannounced socket probably a consumer branded G34(or similar) socket will have the 8 core quad channel DDR3 bulldozer(Interlagos)..

If AMD brings out a consumer branded G34 (or similar) socket :D I'll be buying 4 of those! ;)
 
If AMD brings out a consumer branded G34 (or similar) socket :D I'll be buying 4 of those! ;)

Just buy a server board with only one socket? I'd miss the OC, though :(

EDIT: doh! epic fail on my part. This is [H]ardOCP... not [H]ardSHT. (Stock Hardware Technology!!!)
 
Just buy a server board with only one socket? I'd miss the OC, though :(

EDIT: doh! epic fail on my part. This is [H]ardOCP... not [H]ardSHT. (Stock Hardware Technology!!!)

Nah, Supermicro = crap!:mad:

I'll have no choice but to leave my next upgrade until next year.:( I'm only going to purchase what I need at the moment.:eek:
 
ASUS is in the server board arena, too...

still no OC features :(

That's why I'm not interested:(

One thing; AMD should do is come up with a socket that is 100% superior to the LGA socket so that the CPU can only be inserted in one direction only.
 
Last edited:
I disagree. SuperMicro, TYAN and Intel are the only 3 choices I would make if I needed a stable multi-socket server board. None of these will overclock however.

What you do not understand by what I said, is that I am after a workstation board that can overclock and that it can also in the future be upgradeable to 16-core BD CPU. AM3+ supports up to 8-core BD.

AMD needs to make their south bridges universal meaning workstation boards pared with 5690/5790 +SB850/950 rather than SP5100/5200.
 
I disagree. SuperMicro, TYAN and Intel are the only 3 choices I would make if I needed a stable multi-socket server board. None of these will overclock however.

You should drop Tyan from that list. They aren't remotely in the same league as the other two. Trust me. I've owned a few of them.
 
ASUS is in the server board arena, too...

still no OC features :(


asus has been in it for a while but they are a horrible choice.. while yes their boards look good.. they have absolutely garbage support.. they dont update bios at all even when a new processor comes out that it should easily support they still dont update the bios to support it.. so i dont expect much from the asus G34 board and it would be a miracle if they actually put a bios update out for the board to support BD's..

as far as overclocking goes on the G34 socket you need ES chips which have the unlocked multiplier..
 
Sorry guys, we will not support overclocking on Opteron. It is targeted at the server market. The 890-class chipsets and Phenom processors will allow overclocking, Opteron and 5600 series chipsets will not.
 
Sorry guys, we will not support overclocking on Opteron. It is targeted at the server market. The 890-class chipsets and Phenom processors will allow overclocking, Opteron and 5600 series chipsets will not.

I think we all know the drill. Those of us who work with server hardware know better than to overclock that kind of hardware anyway. If we are talking about an Opteron based workstation build? Maybe. However its up to ASUS or MSI to give us options in that arena. I don't think anyone expects anything at a chipset or processor level to help us do that.
 
What you do not understand by what I said, is that I am after a workstation board that can overclock

Actually I expected that was what you were getting at..

Server boards from Intel and AMD have not been good overclockers, I know I have had a dozen or so AMD dual socket boards and about 1/2 of that with intel dual socket boards btween home and work.

You should drop Tyan from that list. They aren't remotely in the same league as the other two.
They appear to be slipping. 5/6 years ago I would have put them 1st with dual opteron 940 boards like the 2882 and 2885 and 2895. I never got my hands on the 4 socket thunder board S4881 but it was on my wish list..
 
Sorry guys, we will not support overclocking on Opteron. It is targeted at the server market. The 890-class chipsets and Phenom processors will allow overclocking, Opteron and 5600 series chipsets will not.
Will you allow your board manufactures to, like Intel allowing (I think...?) eVGA to make the SR-2?
Actually I expected that was what you were getting at..

Server boards from Intel and AMD have not been good overclockers, I know I have had a dozen or so AMD dual socket boards and about 1/2 of that with intel dual socket boards btween home and work.


They appear to be slipping. 5/6 years ago I would have put them 1st with dual opteron 940 boards like the 2882 and 2885 and 2895. I never got my hands on the 4 socket thunder board S4881 but it was on my wish list..

I have a Socket F (quad socket F) nForce Thunder... :D (I think).


Uploaded with ImageShack.us

It's in the cabinet to my right, but I don't want to hook it up right now...:(
 
Sorry guys, we will not support overclocking on Opteron. It is targeted at the server market. The 890-class chipsets and Phenom processors will allow overclocking, Opteron and 5600 series chipsets will not.

Looks like your competitor might end up with my next upgrade then!:mad:

Your AM3 & AM3+ socket = crap!:mad:
8-core CPU limit.
16GB RAM limit.

Next year DDR4 will hit the market.:cool:
 
Looks like your competitor might end up with my next upgrade then!:mad:

that's why I asked the question. Since eVGA got the SR-2 out, does that mean AMD will allow it's vendors to follow suit... might even help increase sales of higher-margin server parts :cool:
 
that's why I asked the question. Since eVGA got the SR-2 out, does that mean AMD will allow it's vendors to follow suit... might even help increase sales of higher-margin server parts :cool:
The SR-2 board is big man and it won't fit in my case.:( The XL-ATX board does though. :cool:
AMD should get their ass into gear; like C32 to G34 socket would be fantastic!:cool:
How about BD powered Phenom FX CPU for use in a workstation/enthusiasts motherboard with 32 to 64 GB triple or quad channel DDR3/DDR4 RAM.;)
 
They appear to be slipping. 5/6 years ago I would have put them 1st with dual opteron 940 boards like the 2882 and 2885 and 2895. I never got my hands on the 4 socket thunder board S4881 but it was on my wish list..

They were slipping before that. The 2895 was a piece of shit. It was plagued with electrical problems and high failure rates.

Looks like your competitor might end up with my next upgrade then!:mad:

Your AM3 & AM3+ socket = crap!:mad:
8-core CPU limit.
16GB RAM limit.

Next year DDR4 will hit the market.:cool:

You do realize that Intel no more supports overclocking than AMD does? In fact you could say that Intel is less supportive of it. The only reason you can overclock Intel's server processors right now is because EVGA created the SR-2. If they created a similar board for Opteron's you could do the same thing. Neither of them has multiplier unlocked server processors. AMD however has many more multiplier unlocked processors in their lineup than Intel. Intel has only a handful of models outside the Extreme Editions and those only exist because of AMD's Black Edition CPUs.

that's why I asked the question. Since eVGA got the SR-2 out, does that mean AMD will allow it's vendors to follow suit... might even help increase sales of higher-margin server parts :cool:

I don't think it has anything to do with what they allow. At least not with AMD processor based chipsets. They do not have the control over board partners that Intel does. If EVGA wants to make a overclockable Opteron board all they need to do is design it and manufacture it. (Or contract the manufacturing out as they do now.)
 
Who said server cpus don't overclock well? I had a dual core 1.8 ghz opteron 64 back in the day overclocked to 2.8 ghz on air :)
 
Back
Top