What the FUCK happened here?

What the FUCK happened here?

How was this so fucking bad, then they release Ryzen and it's actually pretty good.

What the fuck was this shit?
How did this even happen?

Other urls found in this thread:

docs.google.com/spreadsheets/d/1Dbt_7FiD8hTo2uuOIKBE3ATCDRqVRpAHFsKnieEncv0/edit#gid=87938175
overclock.net/t/1624139/official-ryzen-7-1800x-1700x-1700-owners-club-4ghz-club
semiaccurate.com/2011/10/17/why-did-bulldozer-underwhelm/
ark.intel.com/products/123613/Intel-Core-i9-7900X-Processor-13_75M-Cache-up-to-4_30-GHz
overclock3d.net/news/cpu_mainboard/intel_s_skylake-x_and_kaby_lake-x_cpus_will_not_be_soldered/1
gamersnexus.net/news-pc/2936-intel-i9-7900x-delidding-cpu-package-thermal-paste
gamespot.com/articles/intel-reveals-core-i9-and-x-series-cpus-includes-2/1100-6450414/
twitter.com/SFWRedditGifs

implying ryzen is good

Semiconductor design is very fucking hard.

Werks great for me :^)

AMD used up all their Jim Keller mana, so they brought him back to refill it.

No multithreading had a huge impact, it meant cheaper intel cpus with less cores could beat it in pretty much any benchmark.

also bulldozer cores were set up in a way that made them have shit ipc.

slow individual cores, plus shit multicore performance made them suffer horribly at anything besides very niche tasks. Overall a huge dumpster fire of an arch

They changed the chip design from how it was with phenom II (to technical to really go into atm) so you could go from phenom II x4 955 3.20 ghz to a FX 8300 3.30 (double core count and 100mhz speed increase) and the older phenom would give you better performance in games esp older single thread titles. Now FX would give better performance if you did only non gaming tasks but since I used my desktop for a bit of everything i refused to accept any performance hits. No compromising. So back to my trusty Phenom II. the FX is currently in service as my server cpu (a task it is very good for).

funny i did the 'upgrade' from x4 955 to a fx 8320. seemed better, but i don't really demand much from my hardware.
did the upgrade to ryzen and now using the phenom for a small homeserver that is probably drawing too much power to really justify.

They went from the traditional multi-core design to modules with each module split into two cores.

This ended up confusing programs which meant that the performance from the old Phenom chips to the FX chips dropped noticeably.

Programs would often time not fully utilize all the cores or utilize them poorly.

I remember tons of people on Sup Forums damage controlling the shit out of them, defending them as being worth it or calling anyone who said they were unoptimized, shills.
However, that was just the truth of it. I still recommended the FX 8350 to someone when the price went down to $150 and they had a $1000 budget.
But... Overall the CPU was architecture was shit.

In the end with Ryzen, they changed back to a much better system which rival's Intel and I can say with confidence is in many cases a lot more worth your money.

Callie and Marie can use up all my mana, if you know what I mean

Their multicore performance was good back when they first came out

It doesn't just rival Intels, Intel literally can't compete. The only reason they can even attempt to is their marketshare and how much money they have to throw at it, but Intel are still using 11 year old CPUs and can't shove more cores in there without yield issues. Ryzen introduces a new interconnect and CCX system that allows their yields to be fantastic and for them to create a lineup of chips purely based around binned versions of the 'flagship', whether it be clockspeed binning or cores not being functional.

8320 was piledriver. piledriver increased single threaded performance by 10%, which made it match phenom clock for clock. first release bulldozer, 81xx series, was at release, 20% slower than phenom clock for clock. it became 10% once microsoft patched windows scheduler for faildozer.

so:
>bulldozer (8150, extra)
10% SLOWER than phenom ii
>piledriver (8350, extra)
clock for clock equal

since you switched from a 955 to a 8320, the 8320 was technically faster since it was clocked an extra 300mhz and had a boost to 4ghz. it also had MOAR CORREESSS so extra threads made multi-tasking and overall system snappiness better.

by the time piledriver came out all it had over phenom ii was higher clocks and MOAR CORES.

Got my FX-6300 for ~$100. Well worth and currently on it.

Completely correct. Intel is now forced to actually change. They can't keep milking the same shit.
I wasn't kidding when I said I still recommended it to people. Also that CPU is still considered one of the best price to performance CPUs (last time I checked) in Passmark.

So I'll give credit where credit is due.

...

Considering how decent the current Ryzen chips are, is there any estimate as to how much or how little AMD might be able to improve over the next supposed 4 years of its' lifespan?

guess i fell for the moar coar meme. is it really a non-meme now with ryzen? you seem to know something and aren't just parroting

lmao dude Ryzen sucks donkey dick
only underage, the brown folk, and the poor eat that meme up simply because they don't know any bettet

...

nice try

Piledriver is slower than Phenom clock for clock, they didn't increase IPC to beat Phenom until Excavator, Steamroller came damn close though.

its my honest opinion that ryzen currently is actually process limited. i base this off of two things.

first:
>polaris
with the release of the rx 500 series, amd admitted they did nearly zero tweaks to polaris. the only difference was some tweaks to the DRIVERS, updating hdmi standard, and MORE MATURE 14nm process. that more mature process allowed them to increase clock speeds by up to 250mhz with the same voltage. the only reason why they increased the power limit was to allow for higher builds. but either way, prior with the orginal 14nm process on the 400 series, increased power limit still didn't allow the orginal to hit into the 1400 and 1500mhz range easily like it does now with the 500 series. 1300 range was average. i own a sapphire 580 limited edition myself and it hits 1450mhz at around 180 - 190 watts. compared to my buddies asus strix 480 that uses around 200 watts for 1305mhz.

this clearly shows that polaris was process limited on globalfoundries first 14nm batch. and that leads into.

two
>ryzen tops off at around 4.1-4.2ghz.
regardless with how much voltage you pump into it, its hard to hit 4.1ghz. 4.2ghz is pretty unheard of. i can count how many i've seen successfully run 4.2ghz with both my hands. this wall appears to be awfully similar to first release polaris and makes me feel like why amd waited until march 2017 to release ryzen rather than releasing it in late 2016 which nearly everyone suspected.

if you look at their first roadmap they released, as pic related, in 2015, you can see they at least where hinting at a 2016 release.

right now ryzen is in a really good spot. its single thread performance in virtually equivalent to broadwell from intel. there is some give, and take, but averages to around broadwell. gaming it hovers in the middle between haswell and broadwell for its average. again, some give, and take, but around middle due to its use of ccx design. which compared to the cmt design of faildozer, is 100% better.

>regardless with how much voltage you pump into it, its hard to hit 4.1ghz. 4.2ghz is pretty unheard of.
Well that isn't true, you just can't sufficiently cool the chip enough for the crazy voltages it desires to push past those clockspeeds. Just look at der8aur's records for the 1800X and 1600X, 5.8 and 5.9GHz respectively, better than Broadwell-E.

ryzen is non meme. i'm eh if you nit pick then you can say that. but really it came close. and piledriver ultimately came with much higher clocks over bulldozer that it became a mute point to go with phenom unless you could grab one that was cheaper. since piledriver hit such high clocks it always came out faster. there was, at least, maybe a 200mhz deficit between the two in the end.

anyways, cont'd

you act like thats common. 4.1ghz and above ISN'T common. that behavior makes people think they can go out there and pick up a 1800x and clock it to 4.1 or 4.2ghz and set it and forget it. thats not the case. thats not the case at all. i remember silicon lottery stating their own testing saw around 30% of 1800x able to run stable at 1.45v for 4ghz, with 20% for 4.1ghz on overclock.net. and when you look at overclocking results on overclock.net most people top off around 3.8 to 4ghz with ryzen. few will hit 4.1ghz or above.

ryzen does have a wall, very similar to polaris when it first launched. yes some people did get lucky with their polaris and hit 1.5ghz, but it was rare, and it was TYPICALLY under circumstances that most people wouldn't go as far to achieve.

like der8aur's 5ghz+ results. most people won't be running ln2.

>like der8aur's 5ghz+ results. most people won't be running ln2.
The point was that the chips are capable of it, just not with air/watercooling voltages.

Leaks show Intel's boost supposedly pushing the 7900X to 1.75V to hit the 4.5GHz boost they're advertising.

Out of curiosity, does overclock.net have some chart or anything with verified numbers for overclocks? Or would I have to search through results to find them?

well yes they are, if you cool it with ln2 and shove insanely high volts into it. just like polaris. but seeing polaris able to now run 1.4 to 1.5ghz with a simple more MATURE 14nm process, and seeing ryzen currently behaving the same as polaris did when it first launched on the eariler 14nm GF process, really makes me believe that at the moment, ryzen is HELD back by current GF 14nm process.

it will be interesting to see how high they can take ryzen with GF 7nm with zen2. as zen2 will just be a refinment of zen on 7nm with a few tweaks left out of inital zen.

The other thing is that at 1.45v you are degrading your chip if you're running for extended periods of time. So it's generally not worth it for most people to run that kind vcore.

docs.google.com/spreadsheets/d/1Dbt_7FiD8hTo2uuOIKBE3ATCDRqVRpAHFsKnieEncv0/edit#gid=87938175

and
overclock.net/t/1624139/official-ryzen-7-1800x-1700x-1700-owners-club-4ghz-club

1 shekel has been deposited to your Intel® Advertising© Center© Account. Keep up the good work Shlomo!

makes me wonder that early zen engineering samples where limited to 3ghz and below was because of early globalfoundries 14nm. we see intel "leaking" es samples of their processors at speeds very close to what they launch as, but ryzens early es samples where clocked very low. with most around the 2.8ghz mark.
its like at the last minute is when we saw 3ghz and above clocks leak. that be around the time global foundries was probably releasing a more mature 14nm. ryzen was the first to use it, with polaris coming a month later on it.
its actually refreshing to see ryzen being more so limited on process rather than architecture. as bulldozer was heavily limited because of its architecture.

they intentionally made bulldozer complete shit so that their next flop (ryzen) would look okay in comparison

Still better than Intel.

Caution. Image may cause הִתאַבְּדוּת

holy fuck that voltage hurts to just look at, please tell me it's fake

don't overclock!

HOLY FUCK I didn't see that before.

Several things.
The architecture was layed out mostly by hand as the tools for virtualizing a cpu design where in there infancy. They could only simulate components. Lots of misses that couldn't be counted for until it tapped out. Once a new CPU architecture tapes out you are stuck with it for several years at least.
The CPU was designed to be a speed demon but Global Foundries screwed up big time. 5ghz was the target clock speed. If AMD had Intel's fabs they could have reached this clock goal. They also betted heavily on heavily parallel dumb math becoming the norm in the server space but GPGPU completely took over that market.
Also severe mismanagement in general.

It was from computex.

unless you are buying a sub $100 processor you are retarded for buying an Intel chip right now. You would be better just buying a 1600x for like 240 for general purpose gaming and computing.

>Ryzen
>good
>needs expensive Samsung ram
>can't overclock ram
>motherboard issues
>still slower single core performance than intel cpus
>can't overclock for shit

None of that really has anything to do with Ryzen, Blame Mobo manufactureres.

As for the single core, its fast enough unless you do nothing but play games like a faggot all the time.

Intel can't compete for render hardware.

Don't forget that the Phenom II X6 Black Editions existed too and could turbo all the way up to 3.7ghz stock. If you have one of those you've basically had no need to upgrade for 7 whole years until just now where un-optimized games like RE7 are starting to give them of problems

As for Intel multi-core, it's fast enough for general purpose multitasking unless you do nothing but stream games like a PewDiePie faggot all day
: ^ )

FX failed because the silicon did not scale power down as much as they had hoped going to the smaller fab node

It was meant to run at 5+ ghz but with decent power consumption but leakage prevented this. Their design looked great on paper but physics prevented great execution.

4.1 at 1.44v for the 1800X/Crosshair VI Hero
4.2 at 1.39v for the 1998/Zenith Extreme
Better vrm will bring the voltage down and clock higher.

Actually, its better for a media server, which is exactly where intel is getting btfo.

Nice try though.

They made a gamble on future software design and lost due to a combination of faulty engineering and shitty software like microshit windaids.

It was still not even close a fuckup as netburst or epic, but without a stranglehold on OEMs to forcefeed the fuckups like intel did with P4/pentiumD, AMD just had to live with it.

AMD also had to eat a ton of shit due to offloading fabs and taking on ATI, but now it looks like that is starting to pay off. APU bought them through the dark times, a nice new GPU and CPU design. It may be soon enough we see the true payoff with unified core APUs which will leave intel and nvidia eating shit.

5-7% slower IPC ain't shit, and multicore performance more than makes up for it. And none of that other shit has anything to do with Ryzen itself, just process tech and motherboard manufacturers.

Speed demon bipelines have never succeeded and never will. Some kind of amnesia keeps pushing manufacturers into this corner until they get bitten, again.

No, Ryzen has higher IPC than intel's processors. IPC means instructions per clock.
Notice that intel's processors typically run at higher clockspeeds- the outperforming 7700k outperforms at 4.8Ghz compared to ryzen's 4GHz.

It has higher IPC than Broadwell e.g. 6900k, but not kabylake. Lisa Su said this herself when they were talking about gaming performance reviews not being too hot

>needing 16 cores to stream vlc
Heh, Intel cores are perfectly fine for home use and Photoshop. Try again

Yea, but I'm not paying a thousand dollars just to have 16 threads, go suck a dick shill.

I'm not paying 500 dollars just to have moar coarz. Go back to telemarketing pajeet

>More cores is bad
>That's why intel charges so much to have more cores
corelets btfo

semiaccurate.com/2011/10/17/why-did-bulldozer-underwhelm/

>why intel charges so much to have more cores
Intel charges more for moar coarz and features in Xeons and such because businesses can fall for memes too
Enjoy your overpriced Chinese cartoon watcher because you fell for the moar coarz meme I guess

>2017
>falling for the ringbus meme
lmao pleb

Enjoy your Infinity Fabric

Because it takes 4+ years to design a high perf cpu. AMD had to guess at the start of the design process what transistors would be like when the design would be ready for tape out.

They guessed wrong. Had bulldozer been able to clock at the speeds they originally designed for, they would have had a great cpu with bulldozer.

>He enjoys having toothpaste for TIM
>He pays a tax to overclock his chips
>He likes having a TDP over 150W
>I would hope he's at least getting paid to have these opinions

>day in the life of an Intelcuck.

If you own Intel stock, you would be being paid, in dividends and stock appreciation. They actually make money, unlike AMD.

wow ryzen's l1 and l2 cache kick intel's ass.

>He thinks it is acceptable to have literal RFID tags on his chips
>He's probably doing it for free

Jim "i'll bend her colon like uri geller" keller

>Intel shills are grasping literal fucking NANOseconds

I actually own intel stock. (But I do that because I like the company and product).

I owned AMD after the K7 came out, but ended up selling it too late (I lost money). I owned several Athlon MP systems. Still might buy a dual socket threadripper.

That being said, what is the worry with the RFID chip? I don't think it can be read through a case and with the CPU under the heat sink. Afraid they will start refusing a warranty exchange because the chip reports overclocking on a RFID check?

>I actually own intel stock
You are literally a fucking shill lmao Jesus Christ you are pathetic.

nopers. I own the stock because I kind of understand the market and products. I own more Apple than Intel, but don't own any Apple stuff. I don't buy a stock that I don't know anything about.

>i do own intel stock but you should still absolutely listen to what i say

Still better than Intel delidded corelet dead stutterfires.

this board needs ID's. You are confus.

>I own more Apple than Intel
>When the enemy gets so retarded they start throwing ammo at you

>but ended up selling it too late (I lost money)
OY VEY!

>so fucking bad

FX 8320 here. All my games run fine. Stop this meme.

Jokes on you. I purchased a shit ton of Apple just before the 1st iPhone came out.

If you have to straight up lie to defend your fanboy preferences then it's not even worth pointing out why your wrong
>somebody in the thread shit posted about ringbus, literally the person I replied to
>pointing out why it's a negligible difference versus I.F. makes me "grasping"
Sounds like you guys are the ones grasping for use cases that don't exist for the average desktop user

>what is the worry with the RFID chip?
See Those are not ransom resistors. The RFID is reading from the CPU. Storing your keys. So that they can be recovered wirelessly.

Hey man, the 1st rule of investing is don't loose money. Fail.

what it really needs is gold stars for intel shareholders

>> Storing your keys.

Storing which key from where?

>But this time it will work, guys! I'm sure!
>Just imagine this speed demon!
This is just a human nature.

>If you have to straight up lie to defend your fanboy preferences
Ahem

>Housefire TDP
ark.intel.com/products/123613/Intel-Core-i9-7900X-Processor-13_75M-Cache-up-to-4_30-GHz

>More Jizz TIM
overclock3d.net/news/cpu_mainboard/intel_s_skylake-x_and_kaby_lake-x_cpus_will_not_be_soldered/1

>Botnet INDSIDE™
gamersnexus.net/news-pc/2936-intel-i9-7900x-delidding-cpu-package-thermal-paste

>All starting at the low low price of a THOUSAND FUCKING DOLLARS
gamespot.com/articles/intel-reveals-core-i9-and-x-series-cpus-includes-2/1100-6450414/


Any questions?

Probably not the right place to ask this but anyway, Is there any difference between the ryzen die and epyc/threadripper?

I recall reading somewhere that epyc had some improvements over regular consumer grade ryzen but I can't seem to find anyone backing that up.

Encryption, from ring 3.

I'm pretty sure someone said that they are basically like a few R7 1800x's stuck together since they scale so well.

May be wrong but I think that's about right.

I was planning on getting an used FX-6300, ASRock 970 Pro 3, an Alpenföhn Cooler and 8GB of Noname DDR3 Ram, all for 45€.

Is it a good deal although Bulldozer is shit?

Just get ryzen dude, it's alot nicer.

Love my 1700 so far.

So regular ryzen but in multiples? No more cache or anything more "professional" such as that?

its a newer stepping and supports more i/o and ecc memory but it's otherwise architecturally identical

Well I'm sure it does, I think it's got a lot more PCI lanes and such, but the CPU cores themselves I think are basically the same or at least very similar.

that might be a respectable build if you are homeless or otherwise destitute

Why would you compare Ryzen to a CPU that's not even out and for which benchmarks don't even exist yet? Ryzen is a good budget server, we get it, but Intel desktop parts are still cheaper with better TDP and are adequate for general purpose use. Why not have an HTPC with a Pentium or I3k in every room instead of spending FIVE HUNDRED DOLLARS just for the CPU before you even add anything else? That's just silly for shit posting , VLC, and PhotoShop type stuff.

>>Housefire TDP
>ark.intel.com/products/123613/Intel-Core-i9-7900X-Processor-13_75M-Cache-up-to-4_30-GHz


That's a funny bit of specs to get upset about. Supporting 128GB of ram but no ECC?! Not sure how upset I am about the 128GB limit either.

140W pfff that's nothing. Threadripper should be way up there at least, and IIRC Intel's TDP is more of an over estimate than the number AMD uses. As long as both cpus can throttle way back, (which they can) it will just be a race between efficiency, cost, and joules/opp.

If anything I want threadrippers TDP to be higher. It has a huge package intended for 4 dies. It will be super easy to get lots of heat out. (big advantage AMD) Who cares how many watts you are using, as long as you are efficient in joules/opp. That better thermal overhead will let AMD clock things up a notch.

>FIVE HUNDRED DOLLARS
Where is this number coming from? the R7's start at like 300 bucks on Amazon right now.