Fudzilla: Bulldozer performance figures are in

Page 87 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

drizek

Golden Member
Jul 7, 2005
1,410
0
71
hey guys...foud this:

from: https://www.reddit.com/r/hardware/comments/l2g82/im_just_gonna_say_it/

i also read that charlie from S\A said that bulldozer have something special for gamming

and than we have those amd comics...with FX team and Ruby
and that strange empy space on the die photo


i might beeing paranoid Oo

I have heard about 3 or 4 different people under NDA say this exact same thing. And I think all of them have reputations/ aren't 13 year old trolls.
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
I have heard about 3 or 4 different people under NDA say this exact same thing. And I think all of them have reputations/ aren't 13 year old trolls.

It is hard to imagine what that special gaming sauce might be, though. AFAIK, video drivers have never been a bottleneck, have they? Unless AMD has worked out some way to offload some GPU work to the CPU. Which would be... strange. But if effective, I wouldn't complain...
 

Olikan

Platinum Member
Sep 23, 2011
2,023
275
126
It is hard to imagine what that special gaming sauce might be, though. AFAIK, video drivers have never been a bottleneck, have they? Unless AMD has worked out some way to offload some GPU work to the CPU. Which would be... strange. But if effective, I wouldn't complain...

for amd, yes they are...
the radeons don't have a hardware scheduler (like fermi and the upcoming GCN), so all the schedule have to be made via drivers (reads, cpu)...imagine putting 1000+ shaders to work, at the same time! it's near imposible! that is called "vliw bubble"

and that expling why, amd don't have dx11 MT-support, the old 100% cpu usage bug and any hight performance drivers results in several bugs,
vliw-4 vs vliw-5 and GCN improving gaming while beeing compute focused
 

Olikan

Platinum Member
Sep 23, 2011
2,023
275
126
sorry for second post...but....:eek:
3. Pair AMD FX with a high-end AMD Radeon™ graphics card: To unleash the power of the AMD FX CPU when it launches, you’ll of course want an equally powerful graphics cards. AMD recommends matching the AMD FX CPU with an AMD Radeon™ HD 6850 graphics card or higher. As always, make sure you have the latest drivers – includingAMD OverDrive™ software – installed to get the most of your system.*
AMD Blog

it really seems that we have a special gamming thing here
(well, can be just marketing too)
 

podspi

Golden Member
Jan 11, 2011
1,982
102
106
for amd, yes they are...
the radeons don't have a hardware scheduler (like fermi and the upcoming GCN), so all the schedule have to be made via drivers (reads, cpu)...imagine putting 1000+ shaders to work, at the same time! it's near imposible! that is called "vliw bubble"

and that expling why, amd don't have dx11 MT-support, the old 100% cpu usage bug and any hight performance drivers results in several bugs,
vliw-4 vs vliw-5 and GCN improving gaming while beeing compute focused

Hrm, then it seems like BD's advantage performance might be very short lived, as GCN is coming soon, and Fermi is already here :awe:

But I didn't know that, thanks.
 

inf64

Diamond Member
Mar 11, 2011
3,884
4,691
136
I doubt any magical increases in performance will happen. But let's wait and see.
 

lOl_lol_lOl

Member
Oct 7, 2011
150
0
0
There are no odd number "cores" for Bulldozer. The minimum is 2 "cores" (or appropriately "clusters"). :D

1 - Based on Die size and manufacturing process, what is the maximum number of cores/modules possible on the BD arch?

2 - Also, If the min. is 2 cores, could a higher module design include an extra core (i.e., 9 or 11?).

3 - Is the L1 cache exclusive to each core (1MB) or 2MB for each module?

4 - Can we 'estimate' the scalability of BD from 4 - 6, 6 - 8 or maybe even 8 - 10?

Thanx in Advance. :thumbsup:
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,809
1,289
136
1 - Based on Die size and manufacturing process, what is the maximum number of cores/modules possible on the BD arch?

2 - Also, If the min. is 2 cores, could a higher module design include an extra core (i.e., 9 or 11?).

3 - Is the L1 cache exclusive to each core (1MB) or 2MB for each module?

4 - Can we 'estimate' the scalability of BD from 4 - 6, 6 - 8 or maybe even 8 - 10?

Thanx in Advance. :thumbsup:


1.-
8-12 cores

2.-
No it's either the module works or it doesn't and is then disabled

3.-
The L1d Cache is exclusive to a core and is 16KB
The L2/L1i Cache is inclusive to both cores and L2 is 2MB and L1i 64KB

4.-
Well Zambezi and Vishera will only have 8-cores and Steamroller will have 12-cores
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
AMD Opteron Socket G34 16 core Interlagos has dual (2) 8 core BD dies in the same LGA (Land Grid Array), same as MC(dual 6 cores).

Red lines are 2x dual 128-bit memory channels

magnycoursmcmanddiagram.jpg


amdmaranelloplatform630.png
 

Despoiler

Golden Member
Nov 10, 2007
1,968
773
136
From live event http://www.twitch.tv/amd

This is a PR event. They have a bunch of SC2 players playing SC2 @ over 7ghz with LN2.

AMD just confirmed in the video and chat that NDA is gone next week.

Amd: I am personally on Twitter at @Thracks.

Amd: No benchmarks. Please read reviews next week.

Amd: It is a belt buckle. Amd: The chip was engineered in Texas, after all

Amd: 4) It overclocks over 4.5GHz on air, and over 5GHz on water. These are just averages. You may get much higher.

Amd: All I can say is that we're giving you 8 cores that overclock like crazy for less than $250.
 
Last edited:

Crap Daddy

Senior member
May 6, 2011
610
0
0
All they do is advertise the OC capability. From what we found out from different leaks, since AMD will not give away any benchmarks of this piece of technical marvel, IPC is crap, clock for clock comparison with Sandy Bridge is abysmal, some AMD haters say that it's worse than Thuban but hey, you can overclock like crazy for just 250$. 4.5GHz on air? My 2500K is doing the same with a 20$ HSF and it costs 220$ so what's the big deal?
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
All they do is advertise the OC capability. From what we found out from different leaks, since AMD will not give away any benchmarks of this piece of technical marvel, IPC is crap, clock for clock comparison with Sandy Bridge is abysmal, some AMD haters say that it's worse than Thuban but hey, you can overclock like crazy for just 250$. 4.5GHz on air? My 2500K is doing the same with a 20$ HSF and it costs 220$ so what's the big deal?

I don't think it'll be worse than Thuban. For that, the FX Six-Cores would need to be slower than the Phenom II X6, and John said there is an IPC increase in comparison to K10.5. The loss of performance mentioned since last year by AMD is in comparison to having everything dedicated on a theoretical Bulldozer, not a Phenom II.

The integer cores lose an ALU, but according to AMD that shouldn't be anything significant because the architecture itself is faster. Given the clock speeds, number of cores and pricing, I'd expect this to only be a max 10% IPC increase, with something like 5% being more probable. What's more worrying is the FPUs, since those are at a much bigger disadvantage than the integer cores when it comes to sharing resources. Things like POV-Ray and Cinebench 11.5 care a lot more about FP performance.

It will be good for multi-threaded workloads, especially given its overclocking potential, but not much else. Going CMT isn't a very good decision for most desktop workloads, but a great one for servers, virtualization, and render farms.
 

Crap Daddy

Senior member
May 6, 2011
610
0
0
I agree but the sad/fun part is that AMD is marketing these chips as "gaming" CPUs. I highly doubt that these "gaming" FXs will even come close to a mere desktop processor as the 2500K or the 2600K is gaming scenarios.
 

inf64

Diamond Member
Mar 11, 2011
3,884
4,691
136
I don't think it'll be worse than Thuban.
Sounds crazy but just wait for benchmarks. It's true unfortunately. Prety dissapointing that they enter 2012 with a Bobcat level of x86 performance instead of K10++ on steroids that could fight at least Nehalem/Westmere. This way they will barely beat Magny Cours in server space. Crazy stuff.
 
Status
Not open for further replies.