AMD fanboys.

[A6aKaH_Anastasiya]

12 hours ago, Dionyz said:

Running with an single threaded game is an pain with AMD.

 

https://www.cpubenchmark.net/singleThread.html

 

Just look at that. Try to find at least one AMD chip from the top. 

Exactly.

[Dabombinable]

7 hours ago, Prysin said:

-a snip-

I know all that, and K10 has better multithreaded scaling because it actually uses full processing cores, which aren't to be found in Bulldozer and its derivatives. And APU's are more for laptops-I know for a fact that a modern Excavator or Steamroller APU can outperform my Phenom II N970+MR HD5650/HD770 (100MHz OC, though I extend that to 150MHz when its a cool day) while consuming less power and being more suited to a thin laptop design from 2010.

[Prysin]

On 3/5/2016 at 1:25 AM, Dionyz said:

Running with an single threaded game is an pain with AMD.

 

https://www.cpubenchmark.net/singleThread.html

 

Just look at that. Try to find at least one AMD chip from the top. 

Unless you play games from pre 2007, most games support 2 cores from 2008 and up. Atleast from reputable companies. Indies is a different matter.

 

the most "single thread" reliant game currently beign actively played must be World of Tanks. Which is dual or tri core MAX. It has no clue how to use more then 2-3 threads.

 

Enter any game made past 2013ish and almost all of them uses 4 cores/threads AT LEAST. Some, like Crysis 3, The Witcher 3, Battlefield 4, GTA V, Fallout 4, Rainbow Six SIege, Far Cry 4, Far Cry Primal, Battlefield Hardline, Battlefront (2015edt.), Rise of the Tomb Raider and a few more, uses up to 8 threads.

 

Single core performance isnt alpha omega anymore. More and more software is multi-threaded. Perhaps not optimized for multiple threads, but it USES multiple threads. This means FX will suck much less then synthetic benchmarks suggest.

 

Again. Go find me any single game, not made by a indie dev, that was made after 2008 and uses a single fucking core.

 

Good luck. I have a feeling you'll need it.

 

17 hours ago, Dabombinable said:

I know all that, and K10 has better multithreaded scaling because it actually uses full processing cores, which aren't to be found in Bulldozer and its derivatives. And APU's are more for laptops-I know for a fact that a modern Excavator or Steamroller APU can outperform my Phenom II N970+MR HD5650/HD770 (100MHz OC, though I extend that to 150MHz when its a cool day) while consuming less power and being more suited to a thin laptop design from 2010.

then dont make the argument that PH2 is better. Because in 9/10 scenarios, it aint.

the remaining 1/10 is laptops. However a FX8800p  alone would compete with the CPU + dGPU + iGPU combo you got. If you get a laptop with dual channel memory that is

[Dabombinable]

4 minutes ago, Prysin said:

then dont make the argument that PH2 is better. Because in 9/10 scenarios, it aint.

the remaining 1/10 is laptops. However a FX8800p  alone would compete with the CPU + dGPU + iGPU combo you got. If you get a laptop with dual channel memory that is

Look, re-read what I said. When Phenom II where in production, they were good value for money. The FX series outside of a niche market, were'nt good value for money at launch and still aren't. There are a lot of APU that only have single channel memory controllers, which is why in most cases a Phenom II has significantly greater memory performance than an APU. Also, going off my Mobility Radeon HD5650/5770, which has a 128bit bus and 1GB dedicated DDR3 1600 (AKA, even with shared memory the better APU have more memory bandwidth as they can run DDR3 at higher speeds), I will never be getting an APU based laptop with a single channel memory controller-it gimps TF out of APU (especially the Trinity APU).

[Prysin]

14 minutes ago, Dabombinable said:

Look, re-read what I said. When Phenom II where in production, they were good value for money. The FX series outside of a niche market, were'nt good value for money at launch and still aren't. There are a lot of APU that only have single channel memory controllers, which is why in most cases a Phenom II has significantly greater memory performance than an APU. Also, going off my Mobility Radeon HD5650/5770, which has a 128bit bus and 1GB dedicated DDR3 1600 (AKA, even with shared memory the better APU have more memory bandwidth as they can run DDR3 at higher speeds), I will never be getting an APU based laptop with a single channel memory controller-it gimps TF out of APU (especially the Trinity APU).

when FX launched, they were way overpriced. Just like skylake was WAY overpriced vs Haswell for the first 2-3 months...

 

Also, atleast a lot of the Carrizo APUs have an additional RAM slot you can buy a SODIMM DDR3 stick and slap it into there.

 

BUT..... Bristol Ridge APUs, they should in theory use DDR4.... as Carrizo can use both DDR3 and DDR4

[Ryan_Ge]

I've been running an Vishera FX-8350 for about a half year now. I couldn't resist the 8 cores with a clock at 4.0 ghz. I'm pretty happy with my purchase.

[SImoHayha]

16 hours ago, k1ng_alex said:

 

 

Complete Shit sometimes. Some games work better than others.

 

The Division plays like shit, Dying Light does too. BO3 definitively.  Emulators choices are severely limited because lack of single core performance, which affects most games.

 

But most part I would've rather gotten an i5 over my FX 6300 any day. Probably will go to my nearby PC recycle center and find some old Intel parts

[harrynowl]

CMT would've been a decent design if AMD were capable of producing a high performance core in the first place.

[Citadelen]

2 minutes ago, harrynowl said:

CMT would've been a decent design if AMD were capable of producing a high performance core in the first place.

AMD were and are capable. Bulldozer was in many way ways a worst case scenario for AMD.

[Dabombinable]

1 minute ago, Citadelen said:

AMD were and are capable. Bulldozer was in many way ways a worst case scenario for AMD.

Its almost AMD's less shitty version netburst. High clock speeds, high power consumption, high heat output, low efficiency (and it is better, or lower, in all of those areas than Netburst).

[2dfx]

16 hours ago, aisle9 said:

I'm a fan of their APUs, particularly the A8's and A10's, if they're used properly. My wife's laptop has an A8, and while she does exactly zero gaming on it, I'm of the opinion that the A-series APUs are perfect for mid-tier laptops.

 

ALL OF THIS.

 

If I want ANY decent graphics capability in a laptop, my options are garbage integrated in the 300-700 range, meh dGPU's in the 700-1000 range, or spend out my ass for discrete laptop graphics.

 

Or, I can get an A8 or A10 in the 300-700 range and get capable performance without setting my wallet on fire.

[harrynowl]

1 minute ago, Citadelen said:

AMD were and are capable. Bulldozer was in many way ways a worst case scenario for AMD.

By the time bulldozer came out AMD hadn't been capable of producing a truly high end core for some time. They were pretty good at adding more of them and driving up frequency though.

[Citadelen]

4 minutes ago, Dabombinable said:

Its almost AMD's less shitty version netburst. High clock speeds, high power consumption, high heat output, low efficiency (and it is better, or lower, in all of those areas than Netburst).

Yes, cuts from management, Global Foundries incompetence and improper software support. This coupled with CMT, which is a good idea, doesn't like weak cores. Their cache however, no fucking idea what happens there, somebody got drunk.

[Dabombinable]

21 minutes ago, Citadelen said:

Yes, cuts from management, Global Foundries incompetence and improper software support. This coupled with CMT, which is a good idea, doesn't like weak cores. Their cache however, no fucking idea what happens there, somebody got drunk.

If they'd dedicated cache to each ALU instead of having both share it, we'd probably be seeing greater performance from their CMT implementation-considering there is a noticeable single threaded performance increase when you disable 1 ALU per module and the cache is no longer being shared. Pretty much the wrong CMT implementation at the wrong time.

[Citadelen]

1 minute ago, Dabombinable said:

If they'd dedicated cache to each ALU instead of having both share it, we'd probably be seeing greater performance from their CMT implementation-considering there is a noticeable single threaded performance increase when you disable 1 ALU per module and the cache is no longer being shared. Pretty much the wrong CMT implementation at the wrong time.

I wonder if we'll see CMT appear in the future, in whatever form it'll take.

[harrynowl]

14 hours ago, Dabombinable said:

If they'd dedicated cache to each ALU instead of having both share it, we'd probably be seeing greater performance from their CMT implementation-considering there is a noticeable single threaded performance increase when you disable 1 ALU per module and the cache is no longer being shared. Pretty much the wrong CMT implementation at the wrong time.

Well you have a single frontend doing fetch anyway

 

With a long pipeline like the one on FX you can make up for it by having good branch prediction but they don't have that either. The miss rate to miss penalty improvements from K10 didn't scale.

 

14 hours ago, Citadelen said:

I wonder if we'll see CMT appear in the future, in whatever form it'll take.

I hope so. The die space to multi-threaded gain is very good for HPC workloads. It's 50% die investment for ~80% of the multi-threaded performance.

[Yoinkerman]

I own a 4670k, 8320, 5600k, g3258, and a 4170, and amd performs a lot better than the internet would lead you to believe.  There's nothing wrong with going with an appropriately priced AMD like the 860k, 8320, 6300.  I wouldn't go 8350 or higher because the pricing is wack.

 

On the other side of the fence, there's also nothing wrong with getting an i3 with HT or a 4 core intel or better.

 

Just don't go saying "oh wow my 6700k is so much more amazing than my old 8350"  I HOPE SO IT'S 3x THE PRICE.

[Citadelen]

1 minute ago, harrynowl said:

I hope so. The die space to multi-threaded gain is very good for HPC workloads. It's 50% die investment for ~80% of the multi-threaded performance.

Yes, CMT would work well with Intel I think.

[Suika]

Glad I got rid of it.

[harrynowl]

1 minute ago, Citadelen said:

Yes, CMT would work well with Intel I think.

Well if you give it high performance cores and cache that's faster than a HDD it's a good implementation. I'd like to see other companies dabble with it but given the failure of AMDs implementation I think they'll just not bother. I see ARM designs using it to supplement the big.LITTLE idea.

[Emillionaire]

I have a friend, who bought a AMD FX-8350 and a R9 390 series. 
I bought a i7 6700 and GTX 970.

 

He often complains of his AMD buy, because my pc beats his.

 

[Dabombinable]

3 minutes ago, Citadelen said:

Yes, CMT would work well with Intel I think.

Considering how good the end branch predictor of Netburst was, definitely.

[Dionyz]

4 hours ago, Prysin said:

Unless you play games from pre 2007, most games support 2 cores from 2008 and up. Atleast from reputable companies. Indies is a different matter.

 

the most "single thread" reliant game currently beign actively played must be World of Tanks. Which is dual or tri core MAX. It has no clue how to use more then 2-3 threads.

 

Enter any game made past 2013ish and almost all of them uses 4 cores/threads AT LEAST. Some, like Crysis 3, The Witcher 3, Battlefield 4, GTA V, Fallout 4, Rainbow Six SIege, Far Cry 4, Far Cry Primal, Battlefield Hardline, Battlefront (2015edt.), Rise of the Tomb Raider and a few more, uses up to 8 threads.

 

Single core performance isnt alpha omega anymore. More and more software is multi-threaded. Perhaps not optimized for multiple threads, but it USES multiple threads. This means FX will suck much less then synthetic benchmarks suggest.

 

Again. Go find me any single game, not made by a indie dev, that was made after 2008 and uses a single fucking core.

 

Good luck. I have a feeling you'll need it.

What happens that the game I list is so fucking awesome that it defeats all games? EVERY SINGLE GAME has PISS POOR dodge mechanics. The only game that has an actual good dodge mechanics / combat is single threaded game. (Vindictus) This game you can SOLO raid bosses. If you are godly in hitting the spacebar.

 

Here is an example of gameplay. Incredibly hard boss

 

Tera, and Blade and soul cannot compare to this combat system. If I recall correctly Vindictus is using half life engine. 

[Prysin]

28 minutes ago, Dionyz said:

What happens that the game I list is so fucking awesome that it defeats all games? EVERY SINGLE GAME has PISS POOR dodge mechanics. The only game that has an actual good dodge mechanics / combat is single threaded game. (Vindictus) This game you can SOLO raid bosses. If you are godly in hitting the spacebar.

 

Here is an example of gameplay. Incredibly hard boss

 

Tera, and Blade and soul cannot compare to this combat system. If I recall correctly Vindictus is using half life engine. 

Vindictus or Mabinogi Heroes, released by Nexon in 2010, is NOT single threaded.

It runs off the Source engine indeed. Which as far back as 2005/2006 was updated to use AT LEAST FOUR CORES. The average game takes 3 years to develop. So the Source Engine would have been using multiple cores by the time development started. Your argument is thus Completely DEAD..

 

Dont believe me? Well, here is my source (pun intended)

http://techreport.com/review/11237/valve-source-engine-goes-multi-core

 

I told you you'd need luck.

Try again

 

EDIT:

fastest CPU by single thread in 2010 (When vindictus released) VS FX 8350

http://anandtech.com/bench/product/697?vs=444

 

my FX8320 running at 4.62 GHz scores around 110 points in Cinebench R15.... meaning FX is basically as good as THE BEST CPU you could possibly buy as a consumer back in 2010.

[Dabombinable]

5 minutes ago, Prysin said:

Vindictus or Mabinogi Heroes, released by Nexon in 2010, is NOT single threaded.

It runs off the Source engine indeed. Which as far back as 2005/2006 was updated to use AT LEAST FOUR CORES. The average game takes 3 years to develop. So the Source Engine would have been using multiple cores by the time development started. Your argument is thus Completely DEAD..

 

Dont believe me? Well, here is my source (pun intended)

http://techreport.com/review/11237/valve-source-engine-goes-multi-core

 

I told you you'd need luck.

Try again

It may be able to utilise quad cores, but Source games still run quite well on the least likely hardware combo. Eg. Celeron M380+Intel 910GML (later upgraded to a Pentium M770 as the Celeron didn't have speedstep).