[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology

View post   

[ Toggle deleted replies ]
>> No.73436086

Great, more snakeoil ""tweaks"" to ryzen 3000s shortcomings.

>> No.73436118

It was proven to be working and all of the info was sent to AMD for inclusion in the official stack soon, you dumb fuck.

>> No.73436144

3700x here, I got it back in August and no amount of tweaking could keep my volts below 1.45 on stock BIOS, all caused by iCUE software, but I need iCUE to set fan curves on my commander pro.

Went into BIOS and now get 4.3 all core at 1.367v and stable on AIDA64 at 78C.
Maybe I got a gold sample, but having 4.4 magic turbo on one core isn't worth it with those awful volts at stock

>> No.73436342

Read the info in links. Profiles for single-CCX CPUs are not out yet, but coming soon.

>> No.73436443

Well even when they launch I doubt I would revert back to a stock auto bios because of the way I watched volts idle at around 0.990v doing nothing, as soon as I ran a youtube video, 1.4^ upwards, only dropping to 1.38v on game load and max all clock of 3.6ghz, with maybe one core finding the chance to hit 4.4 on a loading segment.

Tried using offset of .5 and this would result in lower volts (still around 1.4), but again couldn't use icue, and cinebench was showing my CPU was starved and multithreaded lost 200 points.

>> No.73436692

Well my temps would be 45c on a AIO at idle, now with a fixed 1.367v I idle at 28 to 35c as the fans go up and down at 500 to 700rpm.
iCUE was the main cause of the voltage issues I'll admit, if I uninstalled the software it was 0.990v idle but, I still had 1.45v on cinebench r20 with 4800 and now my overclock with cooler temps and less volts and ALL core 4.3 nets me 5013, I will try any and all new chipset drivers and bios I can, but so far all of them can't give me a decent all core and "safe" volts

>> No.73438279

There hasn't been a single fucking poojet in AMD ever since Rajah was kicked out.

>> No.73438578 [DELETED] 
File: 18 KB, 267x297, 1497883390389.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.73438876

>not celebrating the part where intel will have to do a new arch and your next intel chip will be 2x faster than the one you have now

>> No.73438918
File: 1.33 MB, 1000x1333, Caesar.jpg [View same] [iqdb] [saucenao] [google] [report]

Thanks op. Please take this potentially amusing image in return.

>> No.73439191
File: 35 KB, 453x500, 401634073c3afc9f28a19e32648c0585.png [View same] [iqdb] [saucenao] [google] [report]

Here's what I noticed.
Instead of only one core hitting 4.6GHz on my 3900x I now have 3 cores hitting a bit above 4.6Ghz and the others are up too.
Change in benchmarks is within the margin of error, but temps seem to be roughly 2-5C degrees lower.

So the same or 1% better performance for lower temps and slightly higher clocks.

>> No.73441489


>> No.73441593

>but having 4.4 magic turbo on one core isn't worth it with those awful volts at stock

I think people have a fundementally flawed understanding of how ryzen regulates voltage ,
>1 the voltage you see in software will be the vrm output voltage
>2 ryzen 2nd and 3rd gen have on-die dLO regulators to control voltages across the die, and these are used so long as you arent overclocking in manual mode

What this means is that even if you see 1.5v in cpuz or other software, it is not reflective of the voltage the actual cores are receiving, , infact manual voltage mode at a lower voltage may cause more degradation than with precision boost enabled showing a high voltage.

>> No.73441929 [DELETED] 

>wait for sekrit sauce bro

>> No.73442295

It's all downhill from Navi. At least for Navi, Raja Koduri did fine work on it.

>> No.73442425


Linux driver commits seems to be mostly mix of Chinese and Indian doing the majority of the commits with some help from the Canadians where the ATi team used to be. But it is REALLY surprising they don't have a team in Santa Clara, where you can find a ton of talented people, and the software effort so far shows how badly AMD needs more resources and people on their drivers. All of their 2019 drivers past April have been complete shit.

Also the hardware being alright is not good when Nvidia has been outcompeting AMD since Maxwell and launching GPUs earlier than AMD can deliver. RDNA is promising but damn do they need to do something to catch up when that card is below 2070 performance and they are back to where they were with Polaris and waiting on Vega. Big Navi is already late by a year compared to the Nvidia 2xxx cards.

>> No.73442554


I'm not surprised, but they have to rebuild those teams now, they have more than adequate hardware but stuff like this power plan bug, RDRAND, and driver issues on Windows means you have to step up on the software front because it is actively hindering the success of your products.


True, but hopefully this is the last time AMD is behind on schedule with GPUs, they have been preempted by Nvidia doing launches like Super too many times. TSMC is reliable so it should just be a matter of whether they get first dibs on capacity allocated.

>> No.73444707

>So can you explain why when I used stock bios (ram xmp only) and I had icue installed my temps are 15c hotter at idle?
It's because AMD couldn't tune their auto voltages properly, essentially making zen 2 owners beta testers. If you find manual voltage working better then stick to that.

At least on Intel 9th gen, my auto voltage 5ghz OC sips 20% less power than my minimum stable manual vcore. Auto voltage should be superior provided the chip maker can provide a good voltage/cpu usage profile.

>I always read high volts (1.4+) and heat were CPU killers?
Nobody knows the long term effects but AMD, so it's up to you to trust their PR who claims 1.5v is normal.

>PBO in auto is better
PBO should be turned off on zen 2 as it pushes more power into the chip but barely increases performance at all.

>> No.73444735
File: 42 KB, 501x373, 1351534882781.jpg [View same] [iqdb] [saucenao] [google] [report]

Absolutely irrelevent you fucking autist.
The difference between idling at 30C and 45C is null.
The voltage is set by AMD and is not damaging at all because CPUs get damaged if the voltage is high and the current is high, high voltage with low current (aka 1.5V at idle) is perfectly fine and non damaging to the processor.

>> No.73444768

Are there any benchmarks done with this thing other than the cinebench one with 1% perf gain?

>> No.73444770

makes sense, thanks

>> No.73444844

So would you recommend a auto voltage but a manual clock?
I agree that I was essentially a beta tester and accept that, what software would show me amps? And what is considered high, if known at all.

>PBO should be turned off on zen 2 as it pushes more power into the chip but barely increases performance at all.

So many techtubers ran PBO, and AMD shilled for it.

15c idle is pretty relevant, especially when my fans have to ramp up to stop that temp climbing.

Yes the voltage is set by AMD, but AMD stated themselves that numerous monitoring softwares make draw calls and pull the CPU out of idle and ramp the voltage to 1.4/5. That was my issue currently

>> No.73444846

Probably only words for dual-CCD models (i.e. Ryzen 9). Didn't make any difference to which core was loaded during single threaded workloads for my R7 3700X, although the boost clock rose by 30 MHz.

>> No.73445317

>Probably only words for dual-CCD models (i.e. Ryzen 9).
looks like it, just checked the TPU forums and no R7 got improved results like the R9.

>> No.73445626

Servers don't matter

>> No.73445769

what motherboard are you using, i know that msi boards have bugged PBO and boost,

>> No.73445884

Aorus Master

>> No.73446442

Remember, AMDrones weren't even born then:


>> No.73446526


You won't need this patch if you are running the Ryzen power plan with PBO enabled tho. Kinda useless

>> No.73446544

>You won't need this patch if you are running the Ryzen power plan
Read the article, moron. The custom plan is better than AMD's own plan.

>> No.73446655

>indians work at the indian devision


>> No.73446718

that 1.5v is kinda.... well lets just say its not to spec

>> No.73447656

I haven't seen any benchmarks from this, not single-core, not multi-core, and not even gaymez. All I've seen is people posting pictures of their frequencies as if that means anything. Especially since there were some early AGESA versions that had really low clockspeeds but higher performance than

>> No.73449593
File: 239 KB, 1800x850, AMD-Marketshare-Mindfactory.De-October-2019-9XwZMcO.png [View same] [iqdb] [saucenao] [google] [report]


Weeeeew, laaads.

>> No.73450119 [DELETED] 


>> No.73450904
File: 305 KB, 1480x740, AMD-Marketshare-Mindfactory.De-October-2019-p0UjF0X-1480x740.png [View same] [iqdb] [saucenao] [google] [report]


>> No.73450924

TSMC say the process can handle 1.57v, and Zen2 only uses high voltage for low current loads.
Go slit your kike wrists FUD shill

>> No.73451407

Intel is still king of the laptop and mobile processors.

>> No.73451529

Renoir. Early next year.
Microsoft will feature more AMD chips in their Surface line. AMD is going to gain big market share with it.

>> No.73451672

You do understand that the i5 in the middle is the Ice Lake right? And its almost 50% faster than the AMD?

>> No.73452022

>Go slit your kike wrists FUD shill
just letting you know that i use ryzen myself, and dont go above 1.35v for daily/constant usage

>> No.73452162

Hey man just let the broke ass amd retards keep pretending. Not like it impa ts us actual humans

>> No.73452607

Ice Lake avg. of 5 ultrabooks = 17,812.4

Whiskey Lake avg. of 5 ultrabooks = 14,714.6

Exactly 18% difference like Intel claim, I see that simple math is difficult for you to understand but a company wouldn't claim something that is easily disproved. Again this is based on 5 ultrabooks(and one tablet for the Ice Lake). Since these CPUs are designed for Ultrabooks.

>> No.73452701

>lets compare ultrabook CPUs with desktop CPUs
>omg why is the ulrabook CPU so slow
>Intel is finished
Why am I wasting my time with you?

>> No.73452806

>And that's all Ice Lake is good for. You will not see it on desktop.
Do you even read what you typed?

>> No.73452950

Yes, I picked 5 ultrabooks because it would allow for the fairest spread. If you were to pick an OEM laptop v. say a Surface Pro, then chances are the OEM with full heat pipes and fans will be cooler thus throttle less. I am surprised that this is a hard concept for you to understand.

>> No.73452975

>And overall Ice Lake's results are boosted by using faster LPDDR4-3733 RAM.
Nobody argues with that. Still, that's a valid advantage for Intel. AMD can't into lpddr and they don't even plan to change it.

>> No.73453019
File: 84 KB, 640x780, 1542078202943.jpg [View same] [iqdb] [saucenao] [google] [report]

I think it would be generous to expect Ice Lake (Sunny Cove, 10 nm+) to perform like Broadwell. Shit clocks and worse ipc than the previous node, only exists to check off a promise made to investors. Sapphire Rapids (Willow Cove, 10 nm++) will probably have singlethreaded parity with zen 2 but of course shit yields mean less cores and hyperthreading performs worse than SMT. They're claiming Granite Rapids (Golden Cove, 7 nm) will be ahead of Zen 3 but there's no reason to believe they can stick to these extremely tight deadlines and release that by 2021.

>> No.73453059

You know >>73452282 was a benchmark on the Intel reference platform that was "full heat pipes" and 100% fan speed AT ALL TIMES, right? Go read the review if you don't believe me. It still either lost or was too close for comfort to the Whiskey Lake 14nm part, and most likely because of way faster RAM.
Face it, 10nm Ice Lake is a failure because despite having both a new architecture and a new process it's barely able to compete with the previous generation Whiskey/Comet Lake.
As for Geekbench either average way more scores with outliers removed or don't bother. Arbitrarily selecting results is stupid, especially since it's been proven that Ice Lake scores "better" when put in a freezer for the test duration.
The only saving grace of Ice Lake is the iGPU, which is pretty neat and finally a step forward from Gen9.5-forever.

>> No.73453068 [DELETED] 
File: 147 KB, 1714x1370, 1570827069481.png [View same] [iqdb] [saucenao] [google] [report]


>> No.73453314

The phoronix guy tested an icelake laptop at 4-10c ambient and it made little to no difference in performance, the node is shiet, its litterally uncoolable

>> No.73453454


lol intel fanboi getting btfu on the server category in all possible ways including power efficiency.
>Servers don't matter

>> No.73453463

As if having to change your mobo every upgrade you make is the same as "intel is your friend"

>> No.73453873
File: 707 KB, 584x1062, 72ff731ae22ba5353561ee20e932b35e843e2a8411b2692f2d1bac3ce5c1a8d0.png [View same] [iqdb] [saucenao] [google] [report]


>> No.73453926

>this dumbshit thinking any corporation is your friend

They're only your friend if they can extract money out of you

>> No.73454325

This is why I don't buy AMD.
To many magic updates and luck to get the dam chip to work right.

Want 4.6ghz boost?
We'll make sure you get the right Mobo and bios otherwise it's 4.4ghz.
Faster ram is better?
Well better hope your chip has good IF overclocking otherwise that 3800mhz kit will be slower than a 3200one.
Wondering why Zen 2 runs 1.5V and HOTtttt?
Oh, that's just bad reporting. Make sure to only use AMD software and nothing else. Everyone else is wrong. Nevermind thermal Probe numbers

>> No.73454435

>wew intel has a new 6/8 core on the same socket!
>oh i need a new motherboard
>wew intel has such great smt performance!
>oh the microcode makes it unusable with security
>wew intel has so many pcie lanes!
>oh 24 of them are chipset lanes
>wew intel has 18 cores !
>oh its using intel mayo inside and it vrm throttles
>wew intel is switching to solder!
>oh its not much better than the mayo, and now we can't delid this
>Wew a 28core @ 5Ghz??!!
>oh it draws 1200wats
>wew this nvme ssd is so fast!
>oh the new microcode nerfs it.

>> No.73455846

I literally heard that last line a week ago.

>> No.73456949

the eternal intel customer

also i forgot about the skylake BCLK fiasco ,bummer

>> No.73458676

AMD niggers fucked up real good this time.

>> No.73458788
File: 24 KB, 267x297, 1515020163518.jpg [View same] [iqdb] [saucenao] [google] [report]

>post a simple fix
>point out what windows is retarded af
>'muh amd lied!'
the absolute SOPA of this MACACO

>> No.73458840


>> No.73458866
File: 69 KB, 1024x809, 1568534663232.jpg [View same] [iqdb] [saucenao] [google] [report]

with better performance, better efficiency and higher clocks with one simple install?
yes i will, thanks

>> No.73459164 [DELETED] 



>> No.73459473 [DELETED] 

This list is not very kosher.

>> No.73459535

>7nm node higher clocks than 12nm
Only Ayymd has the talent to turn a good situation into a disaster with their gem of a marketing team.

>> No.73459650
File: 156 KB, 1300x900, AMD-Marketshare-Mindfactory.De-October-2019-0xXzUPs.png [View same] [iqdb] [saucenao] [google] [report]


>> No.73459667

>The jewing has begun
I already miss the old amd

>> No.73459684

>On Geekbench 4.3, the XPS 13 earned a score of 22,492, exceeding the premium laptop average of 15,576. It’s also higher than both the XPS 13 2-in-1 (with a 10th Gen Ice Lake Intel Core i7-1065G7, 19,225) and the Spectre (Intel Core i 7-8565U, 14,935).
Explain this: all ultrabooks, all Geekbench and 10 nm Ice Lake got destroyed by a 14nm++++++ part.

>> No.73460148 [DELETED] 

Just delid and apply chiller bro :)

>> No.73461292
File: 205 KB, 494x440, f31f6d5e536c497713a2045563e51f16b02d37112e7ddd1ead47ec64f08de534.png [View same] [iqdb] [saucenao] [google] [report]


>> No.73462915

>10 nm
ANON...https://twitter.com/aschilling/status/1189972358536290305 ~ https://www.techpowerup.com/260130/intel-scraps-10nm-for-desktop-brazen-it-out-with-14nm-skylake-till-2022

>> No.73463703

Fixed your mistake.

>> No.73465069 [DELETED] 
File: 1.31 MB, 1871x718, 1545681165013.png [View same] [iqdb] [saucenao] [google] [report]

> street shitting subhuman keeps shilling his own website
lmao. this subhuman again.

>> No.73465094 [DELETED] 
File: 320 KB, 395x385, 1556923787941.png [View same] [iqdb] [saucenao] [google] [report]

no thanks, pajeet.

>> No.73465172 [DELETED] 
File: 378 KB, 415x404, 1558534260558.png [View same] [iqdb] [saucenao] [google] [report]

> busted street shitting again
kill yourself, thanks.

>> No.73465212
File: 415 KB, 573x879, 49eer058g230knf.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.73465221 [DELETED] 
File: 409 KB, 898x849, 1549794718796.png [View same] [iqdb] [saucenao] [google] [report]

> he forced the admin to change his name on the techpowerup website because he's mad 4chan worked out he was a paid for prostitute
it's too bad we still know who you are. we can smell the shit from here. kill yourself, thx.

>> No.73465240

>Ryzen needs months of patches and updates to run at marketed speeds
Or you can, you know, just buy an Intel and have it work at marketed speeds right out of the box.
You also don't have to mess with motherboard manufacturer fuckery.

Name (leave empty)
Comment (leave empty)
Password [?]Password used for file deletion.