What went right?

What went right?
The amdmen they actually did it they killed Intel
7nm and zen 3xxx is gonna be a bloodbath next year

Attached: amd_ryzen_2_specs.jpg (1170x647, 87K)

Other urls found in this thread:

wccftech.com/rumor-amd-bringing-12-16-core-ryzen-cpus-to-am4-in-2019-roadmap-into-2020-detailed-zen-2-zen-3-zen-5/
twitter.com/SFWRedditGifs

you can't fight normals and mindshare

>What went right?
-Dr. Lisa Su management of AMD
-Global Foundries got IBM knowledge/Foundries for free
-Jim Keller, the J*w Killer, inventing Infinity Fabric
-Papermaster and Team got their shit together and made a very good uArch

Attached: lisa_su_pepe.png (1000x1000, 386K)

How comes none of them beats the 8700k in gaming? The 8 cores coffee lake release will be a bloodbath for ayymd.

that's if they run enough to get a valid result before catching fire

Attached: 1b70c8cd-73d1-4cab-bc44-903e46d0351a.png (255x345, 105K)

because the world has more important things to do than wasting time on gayms like fa/g/gots and /v/anillasoyboys

>gaming

Clock speed. The IPC of Zen is superior to coffee lake, but Intel's process is extremely refined so their chips can clock like mad. A 5GHz 6-core Zen would be on par, or slightly faster than an 8700k at 5GHz. Problem is, AMD just can't get there yet.

AMD likely has a 2800X chip in reserve for to compete with the 8-core Coffee lake, I'm assuming it's a 10-core chip. Even though it'd still lose in the clock speed war, AMD could still brag about superior core count and productivity numbers.

I suppose the real test will be 7nm, and if they are able to pack on Vega graphics on a 3800X. Then it could take advantage of the onboard GPU encoding that Adobe recently pushed out.

>What went right?
XFR2/PB2 and the lower memory latency.

That's it, really. Though they're big deals and huge improvements for a "refresh".

Does tho, just depends on games. It's so close it doesn't matter on every game, though. Can get either one, while 2700X gives much better non-gaing performance on top of it.

Attached: BCLK 102.png (500x613, 81K)

I have reservations on what AMD calls "7nm"
As TSMC doesn't have true 7nm, just parts of the chip are 7nm.

Don't get me wrong, I use a 1700x but there is still a lot left to be desired.
The 2000 series I'm going to skip as they are pretty much pushed to their max anyways unless you go with voltages that I'm uncomfortable with.

Local stores got retired E5-2697 V2 12C/24T 2.7~3.5Ghz with brand new x79 mobos for $570 USD
E5-2660 V2 10C/20T 2.2~3.0Ghz + new x79 mobo for $300 USD
E5-2680 V2 10C/20T 2.8~3.6Ghz + new x79 mobo for $350 USD
and they throw in a giant tower cooler to boot.
3 year warranty with 1st year fast exchange.

1866mhz Samsung DDR3 ECC ram is dirt cheap (16Gx4 for $300 USD) compared to standard DDR4

2700x is $370 USD by itself here.
Bottom of the barrel DDR4 2666 Kingston Value Ram 16Gx4 is $670 USD here.
DDR4 3200 G.Skill Ripjaws V 16Gx4 is $850 USD here.

Intel 8700K is $358 USD

I live in the Asia Pacific where AMD doesn't give a fuck about the market and prices are high to make up for the low prices in the US.

Dude are you really going to dig up this tired shit?
The entirety of Intel CPUs aren't 14nm, either.

>used parts are cheaper than new parts with better perf/watt and absolute performance
WOW REALLY

>The IPC of Zen is superior to coffee lake
no, it isn't

Delid

Yes, it is.

Like posting about it on Jow Forums?

Eat shit while you watch anime.

wccftech.com/rumor-amd-bringing-12-16-core-ryzen-cpus-to-am4-in-2019-roadmap-into-2020-detailed-zen-2-zen-3-zen-5/

Attached: 1525734916567.png (3000x3000, 945K)

It's roughly the same but coffeelake comes out on top overall because of much better avx performance. Technically ipc doesn't really make any sense. What matters is the actual performance/hz, which will vary between tasks

10 core isn't happening. The ccx are limited to 4 cores and amd only makes two kind of die. The 2x ccx for 8 cores and the apu with igpu. The only other option is threadripper and epyc which are multi-die on package and are massive as a result.

>Clock speed
No it isn't. Why does the i5-8400 with a max clock of 3.8GHz still beat the 2600 with a max clock of 3.9GHz and 6 more threads

uops/cycle is lower on intel though

why do you even need a fast cpu these days? video card is more important for both gaming and deep learning

Keep on dreaming, there's no way Intel can keep those high clocks on 8 cores without melting the socket. And that's without mentioning that they are starting to enter the territory where bingbus scales like absolute shit.

I won't buy amd until they stop getting their ass handed to them in games by locked i5s
>inb4 3400memehz

Attached: Average.png (1336x1998, 91K)

>asses handed to them
>10 frame difference
nice try rabbi

>hurrrrr

Technically Ryzen has a superior architecture for IPC, but it is very difficult to tell since compilers are not able to optimize extremely well for Ryzen to this date. Ryzen is wider than Coffee Lake's design being more balanced in load/store pipeline to ALU pipeline, a larger uop cache, and can fetch 6 uops (as opposed to Intel's 4; and on top of that AMD offers fused uops) at a time at a peak code cache bandwidth of up to 32 bytes/clock. Theoretically, because of the increased width (ALU/AGU) and the lower bottleneck from instruction prefetch and dispatch there should be higher IPC. This is a moot point though, because rarely can you compare things apples to apples if you need to compile real world applications, and we can't really forget about real data cache latencies. Previously, Ryzen 1 was on average around 7 or 8% behind clock for clock against Skylake-X, really not so bad, considering there was a small uplift in cache latency in the Zen+ architecture giving on average a 3-4% advantage.

I would also say that AMD's architecture will still be more scalable using IF even after Intel switches over the mesh since there will be compounded penalties in latency for each increased core count whereas AMD's design will only increase latency dependent on 4 different core-core communication methods, which can easily be optimized.

Perhaps SMT isn't leveraged well in games. Also reducing memory latency by jacking up the RAM frequency is extremely impactful in games. Zen+ memory latency is still 40% higher than Coffee Lake's even after aforementioned SDF latency improvements.

Unfortunately, OpenCL and Vulkan kernel development is way harder than writing C/C++, and not all applications can leverage vectorizability; most C compilers have a hard time autovectorizing SIMD operations. Fortunately for ML, it hits both areas where GPUs are good, very high memory bandwidth and massively parallel.

WaitTMing for the 3800x desu

Attached: 57h.png (518x456, 68K)

>8700k in gaming
The CPU doesn't matter for gaming, you stupid kid.
An AMD FX 8350 is enough for every game on the market.
It's just about the GPU!

Battlefield 5 CPU requirements:
>CPU: Core i5 6600K / AMD FX-6350

>jim keller
Nigga he's a bloodthirsty mercenary who's literally helping jewtel tear a new one on AMD now.

Attached: reaction.jpg (621x742, 172K)

>helping jewtel tear a new one on AMD now.
last thing i heard was that Jim is doing something with ARM at Intlel

>muh games
It slaughtered it in every way and has more consistency especially when it comes to client side streaming and general productivity

Muh dik

Attached: 1527340930879.jpg (480x602, 27K)

>hating on the guy that is carrying the whole industry on his own back juts for the sake of technological advancement and the consumers

>the 7nm promise
everyone falling for it, really makes you think.

Attached: 16195261_835633586578799_5781974250411418825_n.jpg (480x360, 19K)

7nm with lots of cores is starting to get me hyped, but it will be the last Ryzen gen on this dead socket. If you could wait, Zen 5 (or whatever the next series is) seems like the better option with a refined 7nm process and the possibility to upgrade your cpu later. Possibly ddr5 ram also at that point.

Better amd under promises and over delivers unlike the collosal fuck up that was bulldozer polaris and Vega.
Intel is going to clock their 10nm shit to 5-6ghz+ like in the p4 days just to keep up while having temps through the fucking roof while shipping the same shit cooler since 2002
Can't believe poozen makes you intards so angry

Attached: 1527139987312.jpg (405x720, 53K)

exhibit A

Attached: 1359917014843.jpg (451x389, 19K)

Am4 is pretty naf anyway I hope they make something better with am5
I like my as rock x470 k4 but there's shit that needs improvement

It's OK young fag we know the i7 9xx was your first Intel CPU
Meanwhile amd is going to blow Intel a new asshole for over a decade.
I haven't seen Intel this mad since pentium 3 days

Attached: 1527263506605.jpg (1024x678, 82K)

such as

>it's real

Attached: Screenshot from 2018-05-28 14-04-44.png (1116x859, 115K)

Ddr5 support

you guys are a great mixture of funny and pathetic I like it.

Attached: 1364060507391.jpg (1000x800, 115K)

Welcome to fourchane

Attached: 1527138268075.jpg (3000x1996, 771K)

fuck off
you have no idea what you're talking about

Attached: 1527147467951-03.jpg (101x101, 4K)

>As TSMC doesn't have true 7nm, just parts of the chip are 7nm.
That's one of the advantages of MCM and why jewtel is trying to develop their own MCM system.

Can someone inform me (the uninformed) did AMD actually beat Intel on performance and price?

Attached: 1252015180219forgetsheisntontheinternetfunnycomic.jpg (460x396, 26K)

LOL when Vega runs cooler then their house fire shit u know Intel are in literally hot water

i7 and i9 are poor value
The 6core i5 is decent and everything else is shit overheating thermal jizz anyway
2700 and 2600 are the new midrange and high end prosumer kings

The tl;dr is that AMD pulled another Athlon style coup over intel by jumping an entire generation over and forcing intel to scramble to catch up.
Is intel still currently competitive?
In some tasks yes, but in others they're getting shitwrecked.
Can intel catch up in the next couple of years?
No, just recently intel admitted their 10nm process wouldn't be ready for full production until 2020 at the earliest.

The real chip war won't start until AMD wins in single core. Until then it's always a use case compromise between raw speed and cores.

in the enthusiast, HEDT and server, yes. mid-range is a little contentious and depends on your use case. enthusiast processors come with a decent cooler for mild overclocks and only holds XFR2 back over 4GHz, are about the same price with higher clocked RAM costs added in, have better thermals, better well-rounded performance, longer lasting platform, etc. HEDT/server checks these marks but no cooler included and has 50% more PCIe lanes. power efficiency is around the same now.

if you're only gaming then you'll save 40-50 bucks total buying i5 8400 as opposed to the 2600 since you won't need to buy faster RAM to compare, but you lose out on unlocked multiplier and you will lose out multithreaded performance for computational tasks and streaming/multitasking. there's also no reason to buy i3s now that ryzen 3s are about the same price and have a way better igpu. only untapped market is pentium/celerons which will be shortly countered by the athlon series for oem market, and pentium+1030 is around the same performance as the 2400g but you get 2 extra cores on the 2400g.

basically amd is gaining ground fast and more than likely 7nm will put amd in the advantage for the first time ever.

Octa-core Covfefe Lake will clock lower and be worse than 8700K for games in 99.9% of cases.

Amd always had some sort of advantage before bullshit dozer and Vega royally fucked them
I hope 2020 we get a good gpu with mcm infinity fabric out the ass and absolutely kills nvidia

>I hope 2020 we get a good gpu with mcm infinity fabric out the ass and absolutely kills nvidia
That's supposedly what Navi will be.

I can't wait for the thermal throttling and damage control and all the bricked failed delids from idorts

DELID

Attached: 1494886290711.png (633x900, 162K)

Nah I'm talking after that
Hopefully hbm2 scales up as gddr6 seems kinda meh especially compared to 5x

he unironically got hired by Intel this year.

>charges jewtel millions so he can destroy them from the inside

Attached: shit wrecker with certification.png (816x1020, 400K)

> same or very close TDP to coffee lake
> amazing Ryzen core count
> up to 4.3GHz
Nothing beats the 2700X on multicore performance right now (on normal desktops obviously). Also, the 2600 is simply unbeatable at it's price tag.

>destroy them from the inside
He's a dickstabbing engineer who gives literal no fucks to who he works for as long as the cash is flowing and he manages to make something which beats whatever he made before.

yeah the problem is fabs are retarded, samsung can't get their shit together. a few nodes down and i think hbm2 will have comparable prices.

he's like a badass mercenary motherfucker.

> muh extra 7fps on gaymen
> literally the same gaming performance as the i5 8600K
meanwhile, on multicore:
> 2700X has a 22% advantage on multithreaded applications
> "b-but what if I delid and overclock it to 5GHz? Check mate!"
That would still make the stock 2700X 15% faster

If you're just a gaming kid, get the i5 8600K, delid it and OC to 5GHz to get those extra 7fps.
On the other way, if you're a programmer, game developer, video/audio editor or an actual adult, the 2700X is the fastest, hands down.

Attached: asian_boobs.jpg (681x1024, 235K)

He’s like the worlds best assassin who’s been hired to kill himself.

>no TSX support
garbage

Attached: pic.png (985x295, 23K)

> users with a compatible CPUs have reported upwards of 40% higher performance in some cases
show me a single case

Also, show me a 8700K beating a 2700X on any multicore workload.

yeah with it's huge 256kb of l2 cache amirite

:^)

go on their discord and ask

what even uses TSX besides one unoptimized emulator

>show me a 8700K beating a 2700X
I don't know, looks at pissmark or geekbench scores, you'll probably find what you're looking for

TSX is a tradeoff and the compiler can't guess critical sections very well in most cases, that being said, TSX is more important from a practicality standpoint because it will make memory concurrency way better in the future. Software solutions are currently too garbage but AMD has been working on their implementation, ASF, which will likely be implemented in further iterations of Zen. Hardware transactional memory is complicated though.

>which will likely be implemented in further iterations of Zen
Not as far as I'm aware
It's more likely that a future iteration of Zen will get TSX

even on shitbench the 8700K gets beaten.

Attached: 1E4CC351-E2B2-432B-B4E5-FC5384DE3BB2.jpg (2048x192, 75K)

>both at 3.7GHz
>barely beats intel with 2c4t advantage
Now clock 8700k to its default turbo of 4.7Ghz

that's the base clock, brainlet. Both chips boosts

Attached: 1466749502756.jpg (640x640, 65K)

those are base clocks

you have to remember that this is a benchmark in which a 1700 has half the ST of a 8700K

wait, how?
I don't see how a 1700 could be that bad

>The tl;dr is that AMD pulled another Athlon style coup over intel by jumping an entire generation over and forcing intel to scramble to catch up.
Intel is still using the fucking ancient ringbus uarch, can they even design a new uarch?

>can they even design a new uarch
Probably not considering they had to hire Saint Jim Keller.
Intel as a whole seem to be in disarray with investing in trash like internet-of-things where x86 has no place, also they waste money on diversity and other political bullshit, and not to mention that their fabs can't keep up with the competition.

Attached: 1463713054949.jpg (1748x1210, 437K)

that's goybench for you, it just puts the cpu under load for a few minutes, then give it a completely arbitrary score

1700 is haswell tier IPC, but obviously not that bad.
ZEN+ is Coffee-Lake tier tho.