• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

BD Observation A :

Status
Not open for further replies.
L

lightp2

Gawd
Joined
Oct 28, 2009
Messages
672
OK I declare this thread is to find positive. In Life there's always circumstances so you try to determine what kind of attitude you want to pursue (atleast with standard reasonable intention)

There is one particular observation in hardware.fr review

A: ATX12V
1. FX8150 : 3.6GHz 8-core 4.8/48/111.6
2. FX6100 : 3.3GHz 6-core 3.6/40.8/68.4
3. X6-1100T : 3.3GHz 6-core 14.4/52.8/93.6
4. 2500K : 3.3GHz 4-core 3.6/22.8/52.8

Notice FX-8150 8-core 4.8 = 4 x 1.2 = 4.8
Notice FX-6100 6-core 3.6 = 3 x 1.2 = 3.6
so the baseline seems fairly consistent, and it is not far from 2500K score. And it seems to be much improved from 1100T 14.4

Before somebody blast me, yeah I don't really know exact meaning of the graph, but I assume the score must be the lower the better based on 2500K/2600K scores.

For example, when FX6100 is running 100%, it only reaches 68.4.

Implication --> since Opteron parts currently max at 3.3GHz, that means power consumption looks to be in very good shape. This is the important part.

Why quote FX6100 and not FX8150? Because FX8150 seems like it will meet overclocking needs per generic review.

However, there is no overclocking on server setup, so FX6100 is more representative of general health, and it looks to be just fine.

There are improvements.

Cheers.
 
In the legionhardware review, handbrake

1. X6-1100T 6-core 3.3GHz score 99
2. FX-6100 6-core 3.3GHz score 96.2

So if the score is similar on 6-core, but likely power consumption reduction (see previous post), you can slot 1 more module nicely into the same processor with similar socket power-limit, you gain one module-2-int-core-1-shared-fp-core.

This is not relevant for desktop, but very relevant for servers.

There are improvements.

Cheers
 
1. If you already have a compatible AM3+ motherboard, especially those with single-core Sempron or dual-core Athlon 64 II,

2. FX-4100 seems like a reasonable upgrade.

3. FX-4100 2Module-4-int-core-2-256bit-shared-fp-core at 3.6GHz is frequently shown as extremely close to "half" of FX-8150 flagship chip in various benchmarks.

4. Since you only have 2Module, your heat/power buffer zone is much greater than FX-8150 when you intend to extract maximum clock performance under reasonable power consumption.

5. If you can get reasonable price, it maybe of consideration since under 32nm it already starts at 3.6GHz by default.

6. If overclock is not your preference than try undervolt. You also have access to all the new processor features like SSE4.1/4.2 AVX, AES, misc.

7. And you have option of turning it into either
7.1 Pentium Dual-Core MODE for single thread IPC demand
7.2 Core i3 Hybrid MODE : Default
7.3 Core i3 Semi-Hybrid MODE : Custom
1Module - 1-int-core-1-non-shared-fp-core : Old single thread Games Mode.
1Module - 2-int-core-1-shared-fp-core
ramp the clock but stay reasonable power consumption.

8. Finally, if you already running quad or six-core PhenomII or comparable Intel setup then maybe this is not so relevant.

9. Emphasis : for compatible AM3+ existing motherboard users running Sempron or Athlon 64 X2 and generic usage demand.

10. BEST of all, if you are running 990/970-level chipsets with SB950, maybe most of the BD features available to you, Virtualization IOMMU ala VT-D (need to check), you can clock high and you can get all of BD features. Yes, maybe it is not as fast as other suggestion, that's life, sometimes you choose the combination you feel comfortable. Note that it is the same as non-K get all features but no overclocking.

computerbase.de already hitting 5GHz at 2-core MODE. I think, need to check.

I generally suggest that current "generic workload", anything 3.6GHz is more than enough . Watch out for power consumption.
 
Last edited:
Someone with resource please kindly verify for endusers.

--this is speculation--this is not universal truth--high possibility it is wrong--please note --

example FX-4100

2M-4i-2fp 3.6GHz approx 200 (per core 50)

2M-2i-2fp 4.6GHz approx 179 (per core 90)

1M-2i-1fp 4.6GHz + 1M-1i-1fp 4.6GHz approx 215 (per core 71)
 
Last edited:
Of the frequently referred hardware.fr 4M8T | 4M4T | 2M4T graph which also get posted here elsewhere,

1. 18-listed items, most game items 4M8T is slightly worst than 4M4T
2. Other items Bibble is most beneficial with 180% 4M8T against 100% 4M4T. Other items top around 158%, which translate to 6-core-plus result for favorable multi-threading.

Thus under "generic circumstances where you are willing to give up a bit here and there but overall reasonable combination"

1. For FX-8150 3.6GHz run 4M4T at 4.6GHz
2. For FX-4100 3.6GHz run 2M2T at 4.6GHz
3. For FX-8120 3.1GHz run 4M4T at 3.9GHz

Under such circumstances, with tight-volting adjustment, one-int-core-off-per-module, your performance, with cooperating BIOS/firmware, maybe OK, power pending review...

1. You can also try deactivating 2Module, 4M8T--> 2M4T and only re-activate remaining 2M where needed. This should save power. But performance is less than 4M/4T so you may need to ramp higher clock to compensate, which needs higher volts so your module-power-saving is cancelled out.

2. Another reason for this series is 4M-8i-4-fp presents a lot of variation in actual usage scenario so there are many flexibility. There are indeed certain software situation where 4M8T lower clock lower-power-consumption is better than 4M4T high clock high power-consumption

3. In the standard server environment, the usage scenario is quite fixed so the optimization consideration is more set. For example, x-servers/rack with limited wattage per rack so no matter how creative you are, you have to work with supplied power and heat because there could be hundreds' racks inside data center with limited space.

4. For home desktop, from 800MHz to 5GHz in between exists entire spectrum of IT industry, so the CPU core here can meet more challenging situation. Hence maybe there exists a space where more user-space-controlled configuration can improve certain setup.
 
Last edited:
Here in this post, I am going to look at another aspect of BD FX models.

1. From Anandtech review, BD FX die size is 315mm^2 with 2-billion transistors.

2. From history, AMD/ATI both faced similar type of situations in the past for CPU/GPU projects.

3. Thus likely AMD is fully aware but they have to make a choice, simple as that.

4. FX models are primarily server parts adapted for desktop use. I think now this is fairly obvious.

5. Cache is critical for server workloads everywhere. For common desktop usage, lesser is just fine. This is applicable for both Intel/AMD. for generic desktop workloads, notice there's little daily usage difference between C2Q-9xxx and C2Q-8xxx. Similar for Phenom II and Athlon II, So yes, Llano and subsequent APU are the actual focus of common desktop.

6. Which brings me back to the point 4, yes you are actually getting "BARGAIN" server parts! IF you know how to tweak the server parts to your advantage, and if you have a genuine need for server parts, I guess.

7. This is the reason why FX models are as they are, and this is the reason why various scenario are discussed in this series.
 
Last edited:
lightp2 said:
In the legionhardware review, handbrake

1. X6-1100T 6-core 3.3GHz score 99
2. FX-6100 6-core 3.3GHz score 96.2

So if the score is similar on 6-core, but likely power consumption reduction (see previous post), you can slot 1 more module nicely into the same processor with similar socket power-limit, you gain one module-2-int-core-1-shared-fp-core.

This is not relevant for desktop, but very relevant for servers.

There are improvements.

Cheers
Click to expand...

However had AMD just die shrunk Thuban they would have had a much more efficient and cheaper to produce design. So yeah bulldozer sucks.
 
Big let down on this one for me. Even the 4 core FX struggled to keep up with a 4 core Athlon II in one review I read.

The 6 core FX looks less attractive than the older Phenom II 6 core processors and it costs more too. And don't even get me started on the top of the range 8 core FX which at times is beaten by the older 6 core Phenom II. Heck AMD could have juiced up the PhII shoved some more lv 2 cache on there and they would have had a better solution. I just don't see the 8 core as a realistic price either £200 for a CPU that's only competing with the 2500k and I had hoped to see the 6 core be near that level.

I'd wait on this one AM3+ users wait for the price crash and we hope better tuned processors maybe next year..Ivy bridge will cause some problems too mostly the ones that make AMD slash prices to try to shift CPU's

I don't care much I'm hardly desperate for an updated CPU but it's a big let down esp considering the time we waited it seems all for nothing.
 
Dear Sirmo, Facebook, Mr Spocko,

In the beginning of this thread, it is written "OK I declare this thread is to find positive."

1. By consciously participating, you have completed the transition. Even though there might be blurry understanding, soon you will see why...

2. First to address the question of Thuban. According to Anandtech, Thuban die size is 346mm^2. BD FX is 315mm^2 so it is already smaller.
2.1 BD-FX in the slightly smaller dimension, has AES, AVX, SSE4.1/SSE4.2 and two additional cores, plus probably other things I do not know.
2.2 There's only so much you can do within reasonable RnD with respect to Revenue.

3. Remember this, with similar performance many will still choose SB. Where SB/Thuban shrink with no new features, many will still choose SB. Since in majority of situation, SB will be just fine. A 5GHz Thuban Shrink is wonderful to overclockers but very little reason daily use. Simple as that.

3.1 And you must remember if you insist (Thuban shrink with no new features but only high clock-rate) It is not enough as some want new features (AES/AVX/misc added to server, so you cannot use it to supplement server, and thus you still need separate server parts developments. Refer 2.2

3.2 Thus BD-FX situation is actually as it is. Server parts adapted for desktop use.

As for Facebook, yes
1. Bobcat appears to be doing fine
2. Llano appears to be OK, Newer parts incoming.
3. Some users are able to clock Llano desktop to really high-clock rate
note: for daily generic usage, 3.0GHz Quad is more than enough with reasonable power consumption.

Per Mr Spocko, yep I guess you scenario is as described initially in post 3. If you already running a Quad Phenom II 45nm, with overclocking to near 3.6GHz to 4.0GHz, it will be difficult to address the upgrade unless you have very clear software usage scenario.

Bibble photo software seems to be an exceptionally multi-threading friendly piece of code. In hardware.fr review, 4M4T-100 and 4M8T-180. However, it is true many common games are unlikely to reach such extensive improvement when presented with 8-core.

So yes, just like life, sometimes you have to admit there are circumstances really requiring conscious and honest decision. I give an example, Virtualization users maybe more inclined because they want reasonable server parts in their home computer to satisfy their curiosity. However, likely most home users are not into this kind of determination. Yes, in all honesty, here SB is a very worthy consideration.

Thank you for your kind contribution.

Quote

"OK I declare this thread is to find positive."

/End Quote

Cheers



p/s : Do not believe everything written on the Internet. This could be a spin series. Apply analytical skills when reading unsubstantiated materials.
 
Last edited:
In this post I am going to look at more closely per software situation

Again refer to hardware.fr comparison o Deneb Thuban 2M4C 4M4C 4M8C Yorkfield SB4C4T and SB4C8T

the most obvious right from the start

1. Between Deneb 4-core and Thuban 6-core, on same clock-rate basis

Crysis 2, Arma II, ROF, F12010,StarCraftII,Anno1404

1.1 The score is little difference

1.2 What this tell us is : Many games see little difference with respect to Deneb and Thuban.

1.3 Why spend time elaborating on this part ?

1.3.1 To answer some screaming for Thuban shrink with no new feature. The only thing you can rely on is to ramp the clock, which then leads to higher volt, rising heat, rapidly increasing consumed wattage

1.4 Some will then come to claim that software are making use of more threads,

1.4.1 Obviously it will take time, apparently many are not, so a Thuban shrink are still facing 2-4 thread demand mostly for short-to-mid-term. Intel recognises this situation so even enthusiasts from Intel tell you honestly if you only care for generic gaming, 2500K is generally reasonable choice.

1.5 This part is here because this forum is overclocker and gamer domains


2. Based on the same logic, 99% will immediately argue than BD should have better IPC than Deneb/Thuban, NO?

2.1 You are not wrong!

2.2 If you are willing to turn to 4M4I4FP, with a bit of tuning everywhere (takes time as new generation product requires understanding and adjustment everywhere), indeed it is OK, with new software features AES/AVX/SSE4.1/SSE4.2. (this part where software is now able to take advanatage. Previously due to lack of SSE4.1/4.2, some multimedia-based software switch to slower iteration of algorithm. (maybe, need to check)

3.3 Note here 6 major factors : die size, wattage, clock-rate, IPC, new features, and core-counts. Finally hidden factors are resource/time. Choose your combination!

3.4 An example will be helpful here : Take the example of Handbrake Thuban and FX-6100 in post-2, the result is close. How does that relates to 3.3?? First remember although FX-6100 is 6-core, the actual physical die (also smaller than thuban) probably has 8-core insides. Thus here core-count, new features, wattage (under lower-clock, consumption seems fine) are the chosen path.

3.5 the rest you figure out on your own.



Focus of this post : everything in this world is there to teach you, even Intel and screaming users. This post is a collection from the web and many forum postings.

I wrote earlier, eventually you will realize, everywhere everything eventually is there radiates in its own light, you just need to make sure stay calm and correct mindset to assess the situation.



Because I need to cut short this series, so I summarize the following quickly

Eventually Bulldozer leads to movement to new software features AES/AVX/SSE4.1/4.2/XOP/FMA whatever, scheduler improvement and other multi-threading support now higher attention and involvement. These ACTUALLY benefit Intel and ARM as well!

Intel/AMD/ARM/nVIDIA/TI/QUALCOMM/Apple/Samsung/GF/TSMC/Other foundries, whatever will live happily ever after BECAUSE --> ??? "I am not going to elaborate here, ..."

some I can write though, if multi-thread ok, you no need to ramp clock to insane level, so you no need to ramp volt to insane level, so power consumption is down.

1. circular reference so there is an incentive to write multi-thread
2. which will need more core
3. which will suit Intel/AMD just fine for server parts
4. which will suit ARM just fine as processing cycle now in datacenter
5. which will suit Apple/Google just fine where iPad/Android tablets incoming,
6. which will suit all the 3rd party tablets manufacturer just fine, new revenue,
7. Which will suit software house just fine, new development software products
8. which will suit Microsoft just fine, ...
9. The rest you figure out on your own....

Bulldozer --> Lifting the entire IT industry, only one reasonable step at a time where it is sensible, else just move along.

Do not believe everything written on Internet. Apply analytical skill on unsubstantiated materials.
 
I have no idea wtf you're trying to say but thank you for keeping this dribble in your own thread.
 
Quoting myself from AT

BD is like a graduate school research project. You need to run simulations to figure out which combo works well for which load and what not.

To make their engineering effort worth it, AMD should publish the internal architecture specifications in the form an architecture level simulator with knobs for architectural variations (using 2 cores versus 4 cores, increasing the size of the cache, not sharing the cache for high priority threads etc.)

They should also publish an open source compiler for BD architecture.

I think after thousands of graduate students have simulated BD for their course projects, AMD engineers will FINALLY to figure out what the optimal configuration should be.

I am sure all MB manufacturers must be thanking their stars that AMD kept BD pin compatible and all it needed was a BIOS change. Imagine the hit they would have taken if AMD had asked them to support a new socket, with new boards and what not. Perhaps that by itself was a give-away of what was coming.
 
Status
Not open for further replies.
Back
Top