• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Quick BD comment/question

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

JKessell

Member
Joined
Apr 6, 2011
Location
Long Island, NY
Just a quick comment/question about BD.

I understand that the full details of BD haven't been divulged yet, but could someone please explain to me how BD could potentially not be on par with the SandyBridge processors?

I mean AMD and Intel are the two big dogs in the game, but is Intel's R&D really that far out of the range of AMD's?

I just don't get how BD could potentially be under-performing in relation to the SB, when Intel's already creating the Ivy Bridge; which could surely be the death-knell for the upcoming BD.

I just don't get it. :\


EDIT: Just some background information. I was holding out for BD because the new 990 boards will offer SLI and I'd like to try out SLI as well as BD, but if BD isn't going to be performing nearly as good as a SB, I guess I'm just gonna trash the AMD cpu/mobo and head to Intel...
 
As for knowing whether or not Bulldozer will be able to compete with Sandy Bridge, or even 1366 for that matter, we will not know until Bulldozer is released. I'm hoping it is up to par with SB, as that would make for interesting benchmarking results and good results for those gamers out there.

I'd say wait for it to come out and see what the results say before making your decision.
 
The situation isn't so much that Amd's current research is behind that of intel's. The performance deficit comes from a transition years ago. During the original the hay-day of the original FX processors, Amd was able to charge what intel currently charges for their processors because Amd was way ahead in terms of performance. However, in the mid 2000's, intel released the Core2 series. Amd, at the time, was releasing the Athlon x2 refresh as well coming up on the original Phenom processors. The R&D at intel took what they learned from the P4 failures and applied new techniques for manufacturing to make a much more power and performance capable processor. Amd's current generation at the time was focused on "true" multi-core processors, and intel began to pull away.

To put it into perspective, Amd caught up to the 65 and 45nm core2 chips (released in 2006/2007) with the release of the Phenom II processor line (in 2009/2010). Ever since, Amd has been fighting an up-hill battle against one of today's biggest computer companies. Thank god for government regulations against monopolies. :rolleyes:
 
So Chance, are you saying AMD made a tactical blunder in focusing R&D on "true" multi-core processors as opposed to virtual multi-core processors?
 
So Chance, are you saying AMD made a tactical blunder in focusing R&D on "true" multi-core processors as opposed to virtual multi-core processors?

I purposely didn't delve too deeply into that because I don't quite understand it all. However, I will mention that while Amd was making "native" quad cores, intel was simply slapping two duals together and still out-paced Amd for 3 years after. Maybe, just maybe, the native quad cores shouldn't have been the focus during that time. :shrug:
 
The problem is amds imc has barely improved since the socket 939.

Whilst bearing in mind different processor models will give different results,
if you look at the read and write bandwidths for intel dual channel setups
from an intel core duo (775) to i3/i5 (1156) to the latest Sandybridge, you can see gains every time.

read 16485 i7 2600k
14000 -/+ i7 965 (tri channel)
9000 -/+ i5/i3

8194 phenom 975
8601 thuban 1100t


write
18581 17 2600k
12432 i7 965 tri channel
7-9 k i5/13

6826 phenom 975
6900 thuban 1100t

If the benchmark results are true then the sandybridge memory controller is an absolute beast and even dominates intels tri channel (socket 1366).
 
Hmm, I'm going to direct this just to OPs post.

AMD used to be on top, and than think to poor funding for the R&D for the Phenom Series, what should have been the come back for AMD created a large gap between Intel and AMD. The original Phenom was a poorly designed multicore, single threaded aspect CPU. AMD tried to attack Intel at their own game, Single threaded applications. Intel is really good at making processors do one task really fast. The Pentium 4 showed that. After taking that architecture, moving some things around and adding really good branch predictors, the Pentium 4 became the Crore series. From there Intel made good choices in working with Nahalem and Yorkfield. SB isn't a real big push in a new direction. It took Nahalem, and made it better. Mostly with the front end, and widening some of the address lanes and allowing a couple more instructions in the mix.

AMD is making the right choice now, as they did in the past: attack intel on a field they don't have. IMC, AMD did it first, and it took Intel up to the 755 to get an IMC to work. Multi-core design, Intel didn't do true multicore for some time. They thought MCM was a good fit.

Now AMD is going in the direction of complete Multi-threading design. Even with Single-threaded applications will be split among one of the modules. This is something Intel has stayed away from. They want to keep making Single-threaded orientated CPUs, because thats what CPUs are good for. CPUs are not suppose to be doing a bunch of loads at once. Its suppose to get stuff in and out quickly.

In the future AMD will grow further away from CPUs and try to implement GPU architecture more and more into their CPUs. The Shared FP will soon become a GPU Array like processing unit.
 
There may have been something about IMC architecture that prevented that. You'll notice Intel didn't go to an IMC until they had true quads. Hard for me to believe it took them five years to copy AMD if there wasn't a hitch in there somewhere ...

That's a great point. Amd had the right idea by pulling the memory controller off the board, but it might have been the wrong time to do so. The performance increase was nice because it no longer required a beast of a board to clock well, but the total performance was still very lack-luster.
 
I am also holding out for the BD. Hopefully the results will be worth the wait. I'm more of a pricing person myself. I used to buy Intel, then recently I bought an AthlonII x3, all of my comps before this Athlon have been Intel. However the single reason I went with AMD versus Intel was because of pricing. It wasn't until very recently that I have become interested in the performance side of things. Before, if it could play what games I wanted on whatever settings I was a happy camper. Now I am becoming interested in OCing and what not, but it is hard for me to part with the type of money that Intel is wanting for their processors. So it would be nice to see AMD be able to compete with them, and for a fraction of the price
 
Hmm, I'm going to direct this just to OPs post.

AMD used to be on top, and than think to poor funding for the R&D for the Phenom Series, what should have been the come back for AMD created a large gap between Intel and AMD. The original Phenom was a poorly designed multicore, single threaded aspect CPU. AMD tried to attack Intel at their own game, Single threaded applications. Intel is really good at making processors do one task really fast. The Pentium 4 showed that. After taking that architecture, moving some things around and adding really good branch predictors, the Pentium 4 became the Crore series. From there Intel made good choices in working with Nahalem and Yorkfield. SB isn't a real big push in a new direction. It took Nahalem, and made it better. Mostly with the front end, and widening some of the address lanes and allowing a couple more instructions in the mix.

Not trying to offend or anything but this isn't exactly accurate. The original Phenom was never poorly designed, AMD lacked the manufacturing abilities to release the architecture properly.
Core takes very little from the P4. The Core architecture is based on the Pentium M which itself is based on the Pentium 3. Intel shifted back to focusing on IPC, Netburst was built for raw MHz.
This is why AMD was able to overtake Intel with the A64. Intel moved away from IPC with the P4, in the pursuit of pure speed. AMD released the A64 at about the same time that the P4 slammed head first into a power wall and for the first time in the companies history they beat Intel by a large margin.
Intel abandoned Netburst and refocused on IPC and single threaded performance. AMD has realized that they cannot beat Intel on single thread performance and have shifted strategies.
I wouldn't expect BD to beat SB in single threaded applications, although it should be close, but BD should completely stomp SB when it comes to multi threaded performance.
 
??? Not sure what you mean here. The first IMC's were in the K8's, which were excellent CPUs for their time.

Alright then. Since I keep fumbling with my letters, and apparently can't get this right, I'll put it as simply as I can:

The original Phenom Series sucked. Amd had a decent lead and then pissed it down their leg.
 
In case you're unaware the 1100T is a Phenom II just like the 975BE. Same generation, same IMC with maybe a few small tweaks. No redesign, no socket changes. Personally, I'm surprised there's that much of a difference between them.

As for there being no improvement since s939 my benchmarks don't agree with that statement.

Most likely. :)

I didn't say no improvment, I said have barely made much improvement,
In 2004 using my socket 939 and corsair xms4000
I was hitting reads of around 8100, and writes of around 7400.

I'm well aware of the tweaks from c2 - c3 to e0 steppings.
my point although not very well written was basically trying to state that AMD have had their imc since 2003 (754 sing chann). skt 939 Dual channel DDR modules and k8 made sense at the time and could even outperfrom ddr2 systems.

However once am2 took over, the benefits were no longer seen sometimes performing worse than ddr, and with tweaks along the way to am3.
Its been a slight increase here and there, with support for stable lower timings and increased speeds with more populated dimms.

One thing Amd deserve more praise for though is their support for both ddr2 and ddr3 memory controllers on their chips. And long socket lifes it's helped me out big time with my am2+ board.

Correct me if i'm wrong, I've lost track with the intel side of things,
But is Intel not on their second gen Memory controller with Sandbybridge.
Which after 3 years of their first in Nehalem, they have gone back to a dual channel that offers over double the bandwidth of what Amd have ever achieved in 7 years.

Theres a few things I'm interested in, Bulldozers memory performance and Ivy Bridge with it's supposed Quad Channel.
 
Back