Overclock.net › Forums › AMD › AMD CPUs › AMD 6350 12 Threads?
New Posts  All Forums:Forum Nav:

AMD 6350 12 Threads? - Page 3

post #21 of 33
Quote:
Originally Posted by Faithh View Post

Rather than trying to prove me wrong or being constructive, you're insulting. That's called trolling. Sure, anything you have to say about bulldozer being a true x-core is trolling by default. It just is. No matter what, you are wrong. Literally it's a crime saying the 8350 is a quad core with CMT. A fact or not, you're wrong regardless.

Well, since the most widely used OS platform in the world treats it like a HT quad core, for all intensive purposes it is a HT quad core. That is the real world situation. I honestly don't understand why you care so much on how it's classified. That really does not matter, at all. What does matter is performance. And no matter how you spin things the FX chips have been a let down since launch. The 8350 was slower then the 1100T it was suppose to replace. Hence the reason they sell for a lot more used then a new 8350.

How about we just forget about this largely failed line of chips and keep our fingers crossed Zen is actually good. Because this five year old argument has been killed, resurrected, killed again, resurrected, molested, then killed some more.
 
FreeNAS Box
(10 items)
 
 
CPUMotherboardGraphicsGraphics
Intel 4930K Rampage IV Extreme EVGA Titan X SC EVGA Titan X SC 
RAMHard DriveHard DriveCooling
Dominator Platinum 2133 840 Pro 256gb 840 evo 1tb EK Supremacy EVO 
CoolingCoolingCoolingCooling
Swiftech MPC 655 EK reservoir 250 EK Titan X nickel block EK Titan X nickel block 
OSOSKeyboardPower
Arch Linux Win10 Pro Corsair K95 RGB ax1200i 
CaseMouse
900d Logitech g502 
  hide details  
Reply
 
FreeNAS Box
(10 items)
 
 
CPUMotherboardGraphicsGraphics
Intel 4930K Rampage IV Extreme EVGA Titan X SC EVGA Titan X SC 
RAMHard DriveHard DriveCooling
Dominator Platinum 2133 840 Pro 256gb 840 evo 1tb EK Supremacy EVO 
CoolingCoolingCoolingCooling
Swiftech MPC 655 EK reservoir 250 EK Titan X nickel block EK Titan X nickel block 
OSOSKeyboardPower
Arch Linux Win10 Pro Corsair K95 RGB ax1200i 
CaseMouse
900d Logitech g502 
  hide details  
Reply
post #22 of 33
Quote:
The "core" you moved out, can it function? Not really no. The Intel core you moved out, can it function? Yes. A core is supposed to be able to function on her own, which the cluster can't do.

Seeing as how there are motherboards that allow you to disable one core per module and still have the whole chip totally functional, as per what I alluded to earlier, yes it clearly is still functional.
post #23 of 33
Quote:
Originally Posted by NaroonGTX View Post

Seeing as how there are motherboards that allow you to disable one core per module and still have the whole chip totally functional, as per what I alluded to earlier, yes it clearly is still functional.

A cluster isn't a core. I don't understand why people can't get over it. You're trying to say a module alone is a triple-core because there are in total 3 clusters (SIMD cluster incl). That's wrong.
post #24 of 33
Quote:
Originally Posted by Prophet4NO1 View Post

The 8350 was slower then the 1100T it was suppose to replace. Hence the reason they sell for a lot more used then a new 8350.





Both of the above statements are patently false. For the past year and probably longer 8320/8350's have been more expensive on the second hand market than 1090T/1100T. An 8350 will whup an 1100T in any benchmark you throw at them. I have both and can tell you in every day usage the 8350 is miles ahead. Throw in an overclock and the gap just widens. Many here will say the same thing.
post #25 of 33
Quote:
Originally Posted by chrisjames61 View Post

An 8350 will whup an 1100T in any benchmark you throw at them.
except outdated versions of cinebench amiright biggrin.gif
and the 860k is beat by the 955 too laugher.gif
Edited by 7850K - 2/9/16 at 3:42pm
Summit Ridge
(16 items)
 
ASUS R510DP
(8 items)
 
 
CPUMotherboardGraphicsRAM
AMD Ryzen 1600X ASRock Fatal1ty AB350 Gaming K4 HIS Radeon HD7870 IceQ 2x8GB G.Skill TridentZ 3200 CL16 
Hard DriveHard DriveHard DriveHard Drive
Plextor M6S Plus 256GB SSD Toshiba X300 6TB Toshiba X300 6TB Toshiba 2TB 
CoolingOSMonitorMonitor
Deepcool BETA 400 ST Windows 8.1 Pro x64 HP S2031 20" Samsung SyncMaster 932BW 19" 
PowerCaseMouseMouse Pad
Seasonic S12G 750w Lian Li full tower Logitech MX310 SteelSeries 4HD 
CPUMotherboardGraphicsRAM
A10-5750m 3.5Ghz ASUStek A75M FCH HD 8650G + HD 8670M dual graphics 2x4GB Samsung 1600Mhz 
Hard DriveOptical DriveOSMonitor
Toshiba Q300 SSD 120GB Panasonic CD/DVD Windows 8.1 Pro 15.6" 1920x1080 
  hide details  
Reply
Summit Ridge
(16 items)
 
ASUS R510DP
(8 items)
 
 
CPUMotherboardGraphicsRAM
AMD Ryzen 1600X ASRock Fatal1ty AB350 Gaming K4 HIS Radeon HD7870 IceQ 2x8GB G.Skill TridentZ 3200 CL16 
Hard DriveHard DriveHard DriveHard Drive
Plextor M6S Plus 256GB SSD Toshiba X300 6TB Toshiba X300 6TB Toshiba 2TB 
CoolingOSMonitorMonitor
Deepcool BETA 400 ST Windows 8.1 Pro x64 HP S2031 20" Samsung SyncMaster 932BW 19" 
PowerCaseMouseMouse Pad
Seasonic S12G 750w Lian Li full tower Logitech MX310 SteelSeries 4HD 
CPUMotherboardGraphicsRAM
A10-5750m 3.5Ghz ASUStek A75M FCH HD 8650G + HD 8670M dual graphics 2x4GB Samsung 1600Mhz 
Hard DriveOptical DriveOSMonitor
Toshiba Q300 SSD 120GB Panasonic CD/DVD Windows 8.1 Pro 15.6" 1920x1080 
  hide details  
Reply
post #26 of 33
Quote:
Originally Posted by 7850K View Post

except outdated versions of cinebench amiright biggrin.gif
and the 860k is beat by the 955 too laugher.gif
Yes, Cinebench 11.5 will show Stars CPU's to be faster than 15h because it doesn't recognize 15h and therefore doesn't use a proper codepath for it. It also makes no use of any instruction sets that showed up on AMD after its release.

If you want to make a 15h CPU look better than it actually is, you use Passmark, which maxes out every resource. If you want to make it look worse than it is, you use deprecated old benchmarks like Cine 11.5.

An 860K can match an i5-3570 in Passmark. In Cine 11.5, it can't even match an Phenom II 965.
     
CPUMotherboardGraphicsRAM
AMD Athlon X4 870K 4700mhz 1.63v ASUS A88X-PRO Radeon HD 6970 G.SKILL Ripjaws X Series 16GB DDR2133 
Hard DriveHard DriveHard DriveHard Drive
Kingston V300 Toshiba 2.5" laptop HDD, 1TB Micron C300 SSD Generic 2TB HDD WL2000GSA1672 (external) 
Hard DriveOptical DriveCoolingOS
Seagate Momentus XT 500GB ASUS DVD-RW Prolimatech Black Series Megahalems Linux Mint 18 Cinnamon "Sarah" 
OSMonitorKeyboardPower
Windows 10 ViewSonic VG2030wm IBM Model M Fractal Design Newton R3 600W 
CaseMouseAudioOther
Phanteks Enthoo Pro Logitech Marble Mouse Behringer UCA222 Upgraded Realistic Minimus-7 speakers, Lepai 20... 
CPUMotherboardGraphicsRAM
Intel Core 2 Duo Mobile T9900 Dell 0G848F Intel Mobile 4 series 4GB Crucial DDR2-6400 
Hard DriveOptical DriveOSMonitor
OWC Mercury Electra 3G 44GB SSD stock DVD-RW Linux Mint Cinnamon 17.1 "Rebecca" 1366x768 WXGA 
  hide details  
Reply
     
CPUMotherboardGraphicsRAM
AMD Athlon X4 870K 4700mhz 1.63v ASUS A88X-PRO Radeon HD 6970 G.SKILL Ripjaws X Series 16GB DDR2133 
Hard DriveHard DriveHard DriveHard Drive
Kingston V300 Toshiba 2.5" laptop HDD, 1TB Micron C300 SSD Generic 2TB HDD WL2000GSA1672 (external) 
Hard DriveOptical DriveCoolingOS
Seagate Momentus XT 500GB ASUS DVD-RW Prolimatech Black Series Megahalems Linux Mint 18 Cinnamon "Sarah" 
OSMonitorKeyboardPower
Windows 10 ViewSonic VG2030wm IBM Model M Fractal Design Newton R3 600W 
CaseMouseAudioOther
Phanteks Enthoo Pro Logitech Marble Mouse Behringer UCA222 Upgraded Realistic Minimus-7 speakers, Lepai 20... 
CPUMotherboardGraphicsRAM
Intel Core 2 Duo Mobile T9900 Dell 0G848F Intel Mobile 4 series 4GB Crucial DDR2-6400 
Hard DriveOptical DriveOSMonitor
OWC Mercury Electra 3G 44GB SSD stock DVD-RW Linux Mint Cinnamon 17.1 "Rebecca" 1366x768 WXGA 
  hide details  
Reply
post #27 of 33
Quote:
Originally Posted by Faithh View Post

A cluster isn't a core. I don't understand why people can't get over it. You're trying to say a module alone is a triple-core because there are in total 3 clusters (SIMD cluster incl). That's wrong.

Nobody said a cluster was a core, or at least I didn't. In fact I was arguing against that from the jump. The point of contention is people who like to declare that a module/cluster = one core when that is false. It's just people trying to buy into the notion that the term "core" can even have a steady definition. What were the CPU's that didn't have FPU's back in the day? I guess they had ZERO CORES?

The only things that the integer units in K15 shares are the L2$, L1i $, the FPU (somewhat), and the decoder (except for Steamroller and Excavator.) Each integer unit has its own L1 data cache, instruction schedulers, ALU pipes, etc. The integer units aren't fighting for resources at all. There are physically two integer units to a module on the die, whereas with Intel's SMT (HyperThreading marketing name), the virtual threads are just using left-over resources from the actual core. That's not what's happening in CMT at all in K15 (Bulldozer.) K15 cores have their own instruction bus, data bus, control unit, and datapaths.

I'm not sure what you're referring to as "three separate clusters" inside a module but it has nothing to do with what I'm saying. AMD doesn't market the 15h products with "x amount of clusters, y amount of threads", they either just say 8 cores or x modules/y threads/cores. Either way you try to spin it there are indeed 8 PHYSICAL cores on a die with the Vishera die being used as an example. That's something that people need to "get over."
post #28 of 33
Lol. another one of "these" threads.

If you have opened up cpu-z and it says you have 8 cores and 8 threads, that's what you have. How do people mistake this>?
Quote:
An 8350 will whup an 1100T in any benchmark you throw at them. I have both and can tell you in every day usage the 8350 is miles ahead. Throw in an overclock and the gap just widens. Many here will say the same thing.

Couldn't agree more. Because the FX-8 series has 8 physical cores. It's not a hyper threading or SMT.

The cpu simply has only 4 floating point units which do not compose a physical X86 core but a added to feature.

Also ARM processors don't have a floating point unit and is still considered a core. However FP has been added to ARM processor cores.
post #29 of 33
Quote:
Originally Posted by ShrimpBrime View Post

Lol. another one of "these" threads.

If you have opened up cpu-z and it says you have 8 cores and 8 threads, that's what you have. How do people mistake this>?

Because Task Manager says four cores and eight threads. tongue.gif

There's a very good reason for that too. Bulldozer's design means it takes a fairly significant performance hit if both cores in a module are used simultaneously, depending on the load. So, if it were treated like a "standard" 8-core, then tasks would be assigned to the eight threads more or less at random. If we take four tasks to assign to four threads, then sometimes it acts like a "true" quad-core (one thread per module), and sometimes it acts more like two Hyperthreaded cores (with two threads each in two modules).

So, what's the solution? Treat it like it's Hyperthreaded! A good scheduler will look at an i7 and assign tasks to Core0, Core1, Core2, and Core3 before moving on to the additional threads. Likewise, by treating an FX-8000 chip the same way, tasks are assigned to Module0, Module1, Module2, and Module3, and only after that to the second cores in each.


So is it an 8-core? Yes. Why? Because there is no rigorous definition of what a core actually is. All this arguing about clusters and whatnot is completely pointless, because it isn't actually relevant to the definition of a core. Is a core a complete processor? Let's say it is. Well... Bulldozer still fits. Integer unit and backend are completely independent, and while the decoder and rest of the front-end are shared, there are multiples. Four decoders in fact. So... If you cut a Bulldozer module in half, you still end up with two complete processors. The FPU isn't necessary, otherwise the early CPUs wouldn't even be CPUs by this definition. I don't think it was until the 386 that Intel even integrated FPUs onto the die.
Triumvirate
(20 items)
 
Osmium
(8 items)
 
For Sale: [FS] Z97 system: Xeon and RAM
$160 (USD) or best offer
CPUMotherboardRAMHard Drive
AMD A10-7870K ASRock FM2A88X-ITX+ G.Skill Ripjaws X - 2x4GB @ 2400MT/s Samsung 840 120GB 
CoolingOSPowerCase
Evercool EC-HPS-810CP Windows 7 Pro 64-bit PicoPSU-120-WI-25 Mini-Box M350 
  hide details  
Reply
Triumvirate
(20 items)
 
Osmium
(8 items)
 
For Sale: [FS] Z97 system: Xeon and RAM
$160 (USD) or best offer
CPUMotherboardRAMHard Drive
AMD A10-7870K ASRock FM2A88X-ITX+ G.Skill Ripjaws X - 2x4GB @ 2400MT/s Samsung 840 120GB 
CoolingOSPowerCase
Evercool EC-HPS-810CP Windows 7 Pro 64-bit PicoPSU-120-WI-25 Mini-Box M350 
  hide details  
Reply
post #30 of 33
Quote:
Originally Posted by CynicalUnicorn View Post

So, what's the solution? Treat it like it's Hyperthreaded! A good scheduler will look at an i7 and assign tasks to Core0, Core1, Core2, and Core3 before moving on to the additional threads. Likewise, by treating an FX-8000 chip the same way, tasks are assigned to Module0, Module1, Module2, and Module3, and only after that to the second cores in each.

That's how it always has been done and how it should be done, that isn't some sort of a special treatment for SMT.
Quote:
Originally Posted by CynicalUnicorn View Post

So is it an 8-core? Yes. Why? Because there is no rigorous definition of what a core actually is.

Strawmanning, much? There's a definition for CMP and we've always called eg a 2 or 4-way CMP chip a dual core or quad core.
Quote:
Originally Posted by CynicalUnicorn View Post

All this arguing about clusters and whatnot is completely pointless, because it isn't actually relevant to the definition of a core.

Factually they're clusters, it's illogical giving them a terminology that doesn't have a definition.
Quote:
Originally Posted by CynicalUnicorn View Post

Integer unit and backend are completely independent, and while the decoder and rest of the front-end are shared, there are multiples. Four decoders in fact.

Yeah, just stratch the front-end completely and your integer units or your back-end will function so fine. No, the back-end is never independent. The back-end needs the front-end, vice versa. Using multiple decoders is pretty common, there's still a single decoder unit (except for Steamroller/Exca).
Quote:
Originally Posted by CynicalUnicorn View Post

Is a core a complete processor? Let's say it is. Well... Bulldozer still fits.

I'd like to see your "independent integer unit" functioning without her front-end. You need them both to form "a complete processor".
Quote:
Originally Posted by CynicalUnicorn View Post

The FPU isn't necessary, otherwise the early CPUs wouldn't even be CPUs by this definition. I don't think it was until the 386 that Intel even integrated FPUs onto the die.

What about no ALU's at all?

ALU's aren't needed; if literally everything what we have now is built on eg AVX2 (even the OS from the ground up) there's totally no reason to keep those useless ALU's there. We've got our SIMD units because of a reason; ALU's aren't cutting it. They're just soooooooo slow compared to SIMD integer units, there's really no reason to run any integer calculations on them if you can use much faster hardware.

Explain me; how much of a complete processor those alu clusters are when they're literally AFK during a FP workload and completely have nothing to do with the fact that Bulldozer still has 2 FP hardware threads available? That is achieved by using two identical FP units (aka a FlexFPU) which is unusual, those can work independent of each other or work together.

Intel only has a single cluster packing 3x 256 bit SIMD FP's/3x 256 bit SIMD units and 4x ALU's. So imagine when you're only doing FP workloads, your ALU's are afk.

Let's face it, a cluster isn't a complete processor.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › AMD 6350 12 Threads?