What went wrong /g/
AMD will rise again
>>55429936
It was a small core CMT design made with servers and APUs in mind, something it did very well. It didn't do very well as a desktop gaymen processor, which lead to negative public perception.
>>55429936
MOAR COARS philosophy with lower IPC than K10
>>55429936
Nothing really, vishera was actually pretty decent for the price, and thanks to dx12 I won't need to upgrade for a while longer
Nothing really, the FX-8320 in my desktop is still running fine.
I can play muh vidya, and compiling stuff is decently fast.
i posted this in imbecile questions thread
:
i can't get my 8320 to 4.2 stable even with the temps under 50 degrees in prime95 and 1.448 vcore which is retarded.
one core always seems to fail during p95 no matter the voltage.
i had this thing running at 4.4 ghz at around 1.432 on a year ago .
Everything. I can't bear it.
>>55430139
Electromigration. If you keep pumping volts into it to maintain it as high as it can go you're just gonna make it worse.
>>55429986
>small core design
What?
>made with servers and APUs in mind, something it did very well
What
>>55429936
>>55429936
Meme design. They literally would've been better off with a K10 die-shrink and developing from there. Zen will probably go towards that.
>>55430292
A Bulldozer CMT module is small as fuck, area wise.
>>55429936
Stagnant architecture.
>>55430541
Nice, I got the one on the right and got 1ghz free out of the box. Stays cool, stable and runs at lower voltage than default. Also FPS in battlefield changes drastically when zooming/scoping. Just go intel.
>>55429936
incompetent management fired the Athlon64 team.
>>55429936
Nothing went wrong, paid 100$ for a brand new 8320 some years ago and its been providing wonderfully the whole time.
Gaming AAA games w a 7870 no issues
Emulators like dolphin run without a hitch despite preferring high IPC
3D scenes get rendered in times marginal to what an i7 3770 provides, for a fraction of the price.
Undervolted and runs at 1.4 ghz under light use and steps up to 4.1 ghz as needed, saves on power (muh 125w tdp tho)
Is ideal for heavily threaded apps, but does fine in limited enviornments too.. I aint paying 200 more dollars for slightly faster performance during the least demanding tasks my PC will encounter.
>>55430139
P95 is not stable on FX chips, dont use it to stability test, just use the built in stabilty testing in overdrive/wattman
>>55431075
but i lower my chip to 4 ghz at 1.4 and prime95 is stable .
i think my 8320 is dying .
RIP
>>55431075
P95 is definitely stable on Trinity & Richland, which share the same Bulldozer iteration, minus L3.
>>55431141
Heres how prime should be utilized for testing on an FX chip, and i dealt with this with both my 6300 and 8320.
Its basically good as the first preliminary check while trying to OC. If you can pass 10 minutes of prime on all cores with decent stress on the memory controller, move onto to a longer test in OCCT.
Just my personal experience with prime, running it for over 10 hours on stock, it would be very common to lose a worker even at stock settings, and i doubt both my chips were defective. No signs of stability issues outside of that whatsoever.
I shouldnt say prime95 doesnt work, rather it should not be the focus in testing your OC.
>>55431244
Edit.
If you pass 10 or 15 min of prime, go back and bump the OC more. Once it fails, go back to the most recent OC that worked, test prime again but then move to OCCT or the like after it passes for a longer test.
>>55430541
Considering it's an AMD-optimized title that doesn't really surprise anyone. The 8350 doesn't perform that well in other games and you have seen the same benchmarks on /g/.
>>55431349
Dunno, looks alright to me anon.
>>55431426
>NVidia GPUs
Now try that with AMD GPUs and their horrendous driver overhead issues. AMD cards need Intel CPUs to function properly.
>>55431537
That wasn't the claim. The claim was AMD cpus are terrible. Besides this is /g/ - Nvidia basically has a monopoly at this point so using that logic using an Nvidia card is representative of the most common gpu architecture(s).
However I do happen to have a picture using an AMD gpu.
>>55430292
It was made to take relatively little physical space so you could cram a lot of them on one die or get a decent number on one die while still shoving a very good IGP on it. These are two things it did well, it's a shame the single threaded performance was lacklustre and they never got turboboost to work.
>>55432071
For multi-threaded workloads vishera (lel bulldozer) is an absolute monster when you consider the Intel chips it competed against at the time.
>>55432132
>that 1100T
Thuban was pretty based.
>>55432132
Vishera is piledriver btw. But meh. Heavy multi threading was literally the entire focus, Intel actually stole from AMD`s book of advertising with the term "Megatasking." Not sure how people forgot about that, and gave intel credit for inventing that.
>>55432236
Clock it to 4ghz on all 6 cores and watch that shit fly. K10 chips in general were based as fuck desu.
>>55432294
>K10 chips in general were based as fuck
Damn right. Had a little Athlon II X4 620 in a (kinda crap) prebuilt that I got for free, managed to push that thing from 2.6GHz to 3.5GHz using the FSB. Took 1.4 volts but stayed surprisingly cool under a Hyper 212 Evo.
>>55431606
I like how these are based on in fps instead of average
>>55429936
Not enough Taiga
>>55429936
The architecture had too many bottlenecks and was designed for a clock speed that Global Foundries was never capable of delivering.
Props to AMD though for making it considerably more efficient on the way to excavator. Taking a pair of modules that used 90+W in bulldozer and cutting that down to less than a third in excavator while mostly maintaining clock speeds is no mean feat.
>>55429936
POO
>>55432853
Kill yourself.
>>55432899
What the fuck is wrong with you?
>>55432899
more like poo in the motherboard
>>55432905
there is no cure for pooposting
>>55429936
nothing the FX 83XX series are fucking fine
>>55432905
>Kill yourself.
Why?
>>55433110
>125w tdp
>Housefire
They where fine 3 years ago but now they are hopelessly out of date
ffs my i5 4690 has aged alot better.
Zen cant come soon enough
>>55430139
>i can't get my 8320 to 4.2
me either, just deal with it.
>>55432132
The problem is that's very limited. I'm encoding right now on 2 Intel quads and I'm just under full util because I'm single-thread limited by QTGMC. If I had two Visheras I'd have to split each encode up again or else half the cores would be idling. Having twice the single-threaded performance is just more useful and convenient.
I'd also be in a fucking swamp. I've disabled turbo on my Ivy and Haswell because their heat is bad enough. Having double the heat output for no gain would not be my idea of a good time.
>>55432789
i like how the AMD chips are overclocked to high hell too, as a 8320 owner il never reach those speeds
>>55433185
oh god anon why did you have to remind me. My 4690k is fucking 2. Its been chuggin along at 4.6ghz at 1.23 volts. Got it hot off the line at my sisters work. She tests the processors for intel, so she gets deals. I could get any processor 50% off.
Zen seems too good to be true.
The performance claims arent too outlandish, its just that AMD always fucks something up.
>>55433255
the ones off the line are cherry picked for employees, so the silicon lottery is always struck :D
>>55433255
>>55433269
kek i see the tech community is still as insular and muh sekrit club as ever
so glad i quit i just come here to troll and get vapedank new memes
>>55429936
>not Certified Shitwrecker™ BasedSiliconGod Xx_Keller_xX
This is the problem.
>>55430541
Intel on suikide clock LMAO
>>55433357
Thanks Based Jim Keller!
>>55433357
thanks based jim keller
>>55433357
You know Keller left AMD almost a year ago now.
>>55433331
dahell you getting at mate? You sound kind of mad that you cant get a 5820k for 20 more than a 8350.
>>55433397
>muh Jim Keller left AMD bla bla bla
This again...
>>55433357
Thanks Based Jim Keller!
>>55433397
but before he left do you know what he was working on?
>>55433397
he had finished Zen by then, and he never really sticks around long after the job is done.
>>55429936
Nothing if you're not a gaymur homo. My FX8350 chews through anything like it's nothing in Linux.
>>55433547
>it runs Firefox guis!
>>55430139
They call it silicon lottery for a reason. You didnt win.
>>55433632
>lives life vicariously through remote controlled cartoons
>mocking anyone
>ever
>>55433547
too bad 98% of /g/ is actually a mix of /v/ and /a/, i.e. the most windows'd demographic ever.
>>55433835
>using 125W to get half the shitposting performance
![FX-9590-73[1].jpg](https://i.imgur.com/XV96lwQ.jpg "FX-9590-73[1].jpg")
>>55432132
did anyone even buy a 9590 lol
thing gets so hot it has to watercooled
FX-6300 and FX-8350 were the best in the series
above that, you are burning a lot of wattage to get a small improvement (FX-9000 series)
and bellow that you might aswell stick to your grandma's computer
>mfw i dont intel and nvidia to have a monopoly but AMD only has 2 CPUs and 1 videocard worth buying atm
not sure about APUs tho, that looked quite promising, is the AMD A10 actually decent for gaming?
>>55434116
For older games and/or more compact rig builds, they're OK, but for anything bigger the lack of memory bandwidth to the onboard GPU will cause them to struggle a bit, necessitating a dedicated GPU of some kind.
>>55433357
Thanks Based Jim Keller!
>>55434116
>decent
If there's for sure to be any sort of gaming it's really hard to recommend an APU. You can get SO much more performance for only a small increase in price with a discrete GPU.
The only time I would recommend an APU is for a general el cheapo shitposting machine, where there's no reason to have a graphics card but the ability to throw a game on there might be nice.
>>55434165
DDR4 APUs are the things I'm most looking forward to right now, as the higher speed should help immensely with that bandwidth starvation.
>>55429936
I use my fx8350 oc to 4.4ghz for like 4 years no without problem, it's good enough for 60fps gaming and also goood enough for other stuff like converting (every core gets used), also for the average use like 90%of the time it is a overkill
it's pretty good for me honestly, don't know what you expect from a above budget cpu
I had a 8320 for the longest time, bought when it first came out. It's retired and running in my wifes computer but she plays overwatch and crap with me. She also has my retired 7950. But hey, she's fine with it. It honestly does pretty well (clocked at 4.5Ghz with 16GB of DDR3 2400).
She just is a tab whore, and typically has 10+ tabs open. 8320 manages that fine, and league of legends and whatnot.
I upgraded because I went to crossfire 7950(one of them died). I had a 4690K for a shortwhile till I delidded it and ran it at 4.9ghz for about a month and then it took a shit on me. Now I am running a 5820K @ 4.5ghz and it feels good man. Have a 390X now, bad idea IMO. Waiting for the next best AMD card to come out, or might jumpship and get a 1080 Ti or new titan. Gonna build my wife a new PC within the next 12 months and am eye balling a complete AMD build. I can't imagine Zen and a RX 480(490?) being more than $1000.
>>55433408
gee WHAT DO U THINK?
AMDs only rise to power was during the Athlon 64 days and Core2duo came out shortly after.
Apart from the obvious IPC differences and the architecture superiority another dimension of Intel is the fact that they actually do research themselves on the materials they use and so the final product is not only lightyears ahead of AMD, but also any other kind of transistor material tech.
I've had my FX-8350 for 2 years now and it's served me pretty well. I can't say I've ever been dissapointed with the processor until recently, but that's mostly due to an aged GPU that was already mid-low tier.
As expected of the memecores, the performance is really inconsistent across titles. Hell the i5-2500k still beats the 8350 in a good chunk of games (even threaded games) without OC.
>>55431057
Wow, we have the same computer. I too play AAA games
>>55429936
Everything. IPC worse than previous generation, power hungry, not a true 8 core, way too ambitious, etc.
They're trying to do something similar with the GPUs and I'm sort of worried about it. I get what Raja wants since it's better for all of us having 1080 performance at $400 instead of $700 but the problem is the same one the faced with Bulldozer: developers need to deliver. DX12 and Vulkan are already here, but it's up to the ones developing software to take advantage of the resources. I lost my mind seeing a Fury X paired with a 980Ti like they were the best pals. If it does deliver, we could expect a return of multi GPUs, but this time including the mainstream.
>>55432905
Butthurt shitskins or cuck detected.
FX-8300 here
feels good man
It games well, and renders audio like a dream. Runs very cool with the stock fan.
recently upgraded from an oced 8320 to a 2600k and haven't looked back, improved perf across the board even at stock speeds. no more shitty stuttering and low fps in >muh gaymes and no housefire required to get non-dogshit overclocks.
>>55436141
FX 4300. 4Chan at 120fps. All my girlfriends are impressed
>>55430766
They're not that small, which is why they failed as a concept.
A 32nm Bulldozer/Piledriver module with 2MB of L2 is 30.9mm2.
Two 32nm Husky cores in Llano with 2MB of L2 is 32mm2.
The module design only wins very slightly in die area because no uncore is needed for the two to communicate, however they are massively behind in performance per clock in every workload.
Looking at Llano vs Trinity where L3 was out of the equation was a complete embarrassment for AMD. Absolute regression in performance per clock, the only saving grace was achieving significantly higher clocks with the module. Despite that Trinity was still slower in some FPU heavy workloads.
>>55434195
Carrizo/Bristol Ridge only supports 2400mhz DDR4, it has the same memory controller as Kaveri.
Raven Ridge will likely have HBM on package.
DDR4 will never be relevant for integrated graphics performance.
>>55435359
Nothing in intel's Trigate process nodes is exotic compared to what the rest of the industry is doing, they're just bulk FinFETw with no insulating oxide on their top edge.
None of their planar process nodes were particularly unique either.
Having a good process does not mean they're using magic super advanced materials. They aren't.
>>55436141
>renders audio like a dream
Does it load webpages real gud too? lel
>all these 8320's
I am amazed. I didn't think it'd have users and that I was the only one.
I wish mine wouldn't overheat. I suspect it might be a cheap ass MicroATX mobo I'm using, since I can't think of any other reason.
>>55436346
>I can't think of any other reason.
i can think of one.
>AMD
>>55433357
Thanks Based Jim Keller!
Please. I want more cores, but I don't want to give Intel more than $1000 for an 8 core. Beat my 3570k and I'll buy your 8 core CPU at $600.
>>55436408
I'm still on a 3570K as well, OC'd to 4.0GHz and could probably get higher if I spend a few hours getting voltages stable. I haven't yet run into anything where the CPU is the bottleneck.
>>55436346
Mobo has nothing to do with it, unless the firmware isn't modulating your insufficient heatsink.
>>55436338
And only webpages
>Stock cooler
Goodness me.
Remember you guys, the Vishera temperature reporting is shot. Almost every single nub running one celebrates how cool it's running, throwing the logic aside that you can't cool below ambient with the stock cooler.
>>55433357
>Intel's high end offering is an 8 core
>Meanwhile Jim Keller the Intel killer is giving us 32 cores with Haswell IPC
Thank you Jim!
>>55436437
Sandy Bridge was too good for its own good. I get that Skylake is about 20% faster, but that took more than 4 years to get there. I'm perfectly happy with Ivy Bridge, but I'd love more cores for VMs, compiling, and ffmpeg.
My 8320 is dead .
I posted earlier that it couldn't stay overclocked at 4.2 GHz and now it can't hold stock speeds .
With fresh thermal paste and noctua dh 14
Rip .
>>55430139
It might be your board not supplying enough energy to the processor. Maybe try turning off that thing in the BIOS that makes it a quad core by linking 2 cores together as one. I don't remember what it's called.
>>55436461
>32 cores
>consumer
Nigga that's server enthusiast shit. We're getting 8 cores max on the >consumer side.
>>55436346
Only AMD Overdrive correctly displays the figure that your CPU senses.
The die does not have a real temperature sensor, and it does not output a figure in degrees. It has circuitry for measuring a metric they call thermal margin. Speccy and other programs never correctly display this value.
>>55436460
>Mobo has nothing to do with it,
Get educated. Insufficient VRM = LLC issues and throttling.
If it actually is overheating at all.
>>55436461
Summit Ridge is only 8 cores, socket AM4 won't support any MCMs.
>>55436485
The term you're looking for is enterprise. Not enthusiast.
They're Opterons, and will probably be around $4000~
>>55436480
;_;7
Shit, Zen is so close too. I'd say buy a cheap Athlon to hold you off, but I think it's using a different socket.
I had an FX-8120... it was average but dat 125W TDP was fucking ridiculous, it heated the entire fucking living room. The damn CPU wouldn't stay cool no matter what I did, I had a hyper 212 evo with a really loud fan, the fucking PC sounded like a jet engine... fuck AMD, never again.
>>55436482
Not possible , used 3 different PSU 2 high end mobos , 3 different GPUs , 2 sets of ram , and another cpu . My 8320 can't even hold stock anymore and there is massive amounts of glitching . It's fucking dead .
>>55436503
Hence the server enthusiast, but yeah, enterprise is more correct.
>$4000
Whew.
I wonder when we're moving to ARM on the desktop.
>>55429936
>Banking on multithreading
>Nobody uses it
Single core performance was shit and that made whole CPU kinda shit.
>>55436552
this and TDP
>>55436534
>mfw ihave been running a fx 6300 at 4.4 for like 3 years on a fucking 3+1 asus mobo and not a single problem with it
>>55436634
I had my 8320 at that same clock for about the same amount of time . I think that's why it died . Yours could be next
>>55436701
Voltage is what counts, not clocks.
>>55436721
It was at 1.408 @4.4 for close to 3 years.
>>55436733
welp, my 6300 is at 1.37@ 4.4
>>55436618
caring about tdp, neck yourself
>>55436860
You know neck means kiss right?
uguu :3
>>55429936
They took a gamble that multi core performance would be more important than single core performance because that was a cheap and easy option, that has only panned out for a few applications so far.
>>55434045
I have one ive got it on a triple 120mm radiator with 3 scythe fans never gets to hot
>>55437586
I bet that rarely ever touches 5ghz - the 9590 is a bitch in that regard (i.e it rarely ever turbos to 5ghz).
>>55433099
never buying again
AMD is so fucking shit. I used to think they were ok, but their latest products are so fucking bad.
- Bad power management
- Bugs and issues that can literally destroy your motherboard
- Slow
- Incompatibilities
Nvidia (and Intel) has no competition really anymore.
>>55429936
Nothing. I've used my 8150 for years to no ill effect.
