Jump to content
Sign in to follow this  
Dasa

AMD Zen

Recommended Posts

I spotted a YT review on the 1600x but it was like 1 hour long so I didn't watch it ...

Share this post


Link to post
Share on other sites

Looks the Vega is the RX560, RX570 and RX580.

Some think these are like rehashed RX400 GPUS which isn't what we wanted.

Share this post


Link to post
Share on other sites

Very interesting Cybes!

 

I sub'd to him as soon as I saw someone else link him here last year. He's always interesting - even if I have to admit to having to listen to a couple of things twice to get past the accent. ;p

  • Like 1

Share this post


Link to post
Share on other sites

he has some good points but his very one eyed

completely side stepping the ipc of the cpu in programs that care about cache\memory performance

 

saying games are running 300fps at low res now sure some games others are under 60fps due to the cpu this will improve a bit as bios\windows updates come out

his example of rainbow six siege is one that i tested which is still gpu bootlicked even at low res cherry picked!

 

bf1 results amd vs 7700k being near max sure if the game can make use of over 8 threads otherwise ryzen may well be at its max already as well

also bf1 is known to be buggy with 100% cpu usage for some people and only ~40% for others with the same cpu

 

programing bottlneck no shit same can be said for every program in existence none are perfect some worse than others but if one cpu is powerful enough in some areas to make up for that and the other isnt...

 

indicating bulldozer was better as 4 years later the second revision is able to beat the 6 year old 2500k in some new games afer putting up with years of sub 60fps performance....

focusing on average percentages to hide the fps deficiency

 

and again showing ipc in synthetics that only care about cpu performance not the cache system performance he really is pushing his side sorry but he is full of shit

https://www.techpowerup.com/231268/amds-ryzen-cache-analyzed-improvements-improveable-ccx-compromises

aida64 may not be accurate for ryzen at this stage but there is memory performance problems there

 

 

on the final note true low res results are not a indication of future performance when improved multithreading is taken into account its only a indication of future performance for current games with a faster gpu or future games that continue to use old engines

but they are a better indicator than high res results as at least you can still get a rough idea of how the cpu will scale with the extra cores

with high res results you can see nothing but the gpu's performance in many games

 

 

conclusion most amd fans will now admit bulldozer was a failure

somebody that is still trying to justify it... epic fanboy

Edited by Dasa
  • Like 1

Share this post


Link to post
Share on other sites

Bulldozer was ill conceived and a false dawn. Ryzen learnt things from intel, that single core performance is still important.

Share this post


Link to post
Share on other sites

he has some good points but his very one eyed

...

 

somebody that is still trying to justify it... epic fanboy

 

I said "interesting" - not "correct". Personally, I wouldn't know one way or another, and would have to check in several places before drawing a conclusion.

  • Like 1

Share this post


Link to post
Share on other sites

truth is i dont think any of us will know for sure for some time yet

there is much work to be done before we will get a real look at how it performs in programs like games

 

edit

apparently he said bulldozer is crap in that review so i may have gone to far with my fanboy comments above

Edited by Dasa
  • Like 1

Share this post


Link to post
Share on other sites

DirectX and driver improvements will only go so far. Game patches the same. At the end of the day it's x86 code being run, sure the AMD and Intel offerings can have special capabilities or compilation requirements that can produce some optimisations.

 

The thing is, every chance the Dx and driver stuff will flow through. As for games, don't be so sure. There's been a fair history of games that favour one GPU type over the other, to say nothing of games that plain never worked properly on ATI/AMD hardware and next to nothing was done about it.

  • Like 1

Share this post


Link to post
Share on other sites

I'm guessing most games have intel optimised code due the compilers they use, and most gamers use some form of intel. No one is gonna bother optimising for bulldozer k10 but Ryzen might get that. Ryxen probably just killed of the k10 now

Edited by Jeruselem

Share this post


Link to post
Share on other sites

another review this one actually shows a improvement to min fps from smt in two games

http://www.techspot.com/review/1348-amd-ryzen-gaming-performance/

 

some overclocking info

Silicon Lottery:

Ryzen 7 1700
93% reach 3.8GHz @ 1.376V CPU VCORE (Or less)
70% reach 3.9GHz @ 1.408V CPU VCORE (Or less)
20% reach 4.0GHz @ 1.440V CPU VCORE (Or less)

Ryzen 7 1700X
100% reach 3.8GHz @ 1.360V CPU VCORE (Or less)
77% reach 3.9GHz @ 1.392V CPU VCORE (Or less)
33% reach 4.0GHz @ 1.424V CPU VCORE (Or less)

Ryzen 7 1800X
97% reach 3.9GHz @ 1.376V CPU VCORE (Or less)
67% reach 4.0GHz @ 1.408V CPU VCORE (Or less)
20% reach 4.1GHz @ 1.440V CPU VCORE (Or less)

 

Share this post


Link to post
Share on other sites

unless your game to see how your chip scales past 1.44v and have the cooling\vrm to cope

Share this post


Link to post
Share on other sites

Real-world test, not just numbers - and without needing your Scottish-filter! ;)

 

Share this post


Link to post
Share on other sites

has anyone seen the recommended voltage for 24/7? Cant seem to find anywhere saying "stay under X". Seems like after 1.4v it really heats up much more.

 

Also, at the end of the techspot report for 16 games, basically at the end, they avg everything but at 1080p Ryzen is 12% slower than the 6900k and 3% on 5960x. 1440 it is 4% slower than the 6900k and on par with 5960x. I mean for the price, I think that is reasonable vs the 6900k and 5960x in gaming performance. 1700 non X is probably the best value if you OC it just some. 7700k and 6700k are still kings of games, I dont see that shifting much but possibly for future games if they get optimized.

Edited by gamble

Share this post


Link to post
Share on other sites

From the looks of current testing, get 1700 because the 1800 isn't much better by much.

Share this post


Link to post
Share on other sites

plus your going to smash my 6700k firestrike physics score with ryzen :)

Share this post


Link to post
Share on other sites

Id be really happy with 4ghz for some bench smashing fun!

 

I found this in Kitguru, dont really know their site much but they had:

 

"There are suggestions that 4.2GHz should be achievable using 1.45V, though neither of the chips that we received were able to hit such levels. AMD also suggests keeping voltage below 1.45V (more towards 1.40V for long-term, 24/7 usage) in order to enhance processor longevity."

 

Seems pretty high to me... as long as the temps are reasonable Id be game

Share this post


Link to post
Share on other sites

Ryzen smt issues can be blamed om windows 10 scheding on the os. Linux kernel 4.10 has the same issue.

Edited by Jeruselem

Share this post


Link to post
Share on other sites

Shouldn't that be "can't" ?

You can't really hold the software responsible, it was here first by a long shot. SMT works better with good memory (random) bandwidth which it seems is a lacking area.

  • Like 1

Share this post


Link to post
Share on other sites

as is often the case for early adopters of new tech bugs abound

from here

Gigabyte Aorus AX370 Gaming 5 / 7 bug report thread

http://www.xtremesystems.org/forums/showthread.php?293279-Gigabyte-Aorus-AX370-Gaming-5-7-bug-report-thread

Gigabyte Aorus AX370 Gaming 5 / 7 User Requested Features

1. Change date format from MM/DD/YY to DD/MM/YY
2. Change current UEFI layout, it's clunky and all around poor
3. Put Q-Flash somewhere immediately visible in the UEFI
4. Temperature sensors could do with some re-calibration, appear to be inaccurate by 2.7c (measured with Infrared temperature gun, DT8380)
5. Add options to disable on-board devices, such as audio.

Gigabyte Aorus AX370 Gaming 5 / 7 Confirmed UEFI Bugs & Issues

Gaming 5:

1. Mouse input / responsiveness is laggy in UEFI. Experienced with hardware: Roccat Kova+
2. Resolution appears to be stretched at POST and with the Windows 10 logo while booting when Aorus splash screen is disabled. Experienced with hardware: BenQ GL2450 connected via HDMI, Gigabyte Windforce3 GTX 980
3. LED keyboards are unusable in the UEFI once they enter sleep mode (keyboard backlight goes out) until a reboot. Experienced with hardware: Ares K5, Sunt PK35
4. XMP profile must be loaded before memory multipliers will work. Experienced with hardware: G.Skill Ripjaws V 3200MHz (F4-3200C15D-16GVK)
5. Memory multiplier is limited to a maximum of loaded XMP profile. There must be a way to be able to set higher multipliers.
6. XMP and memory multiplier options are in both Advanced Frequency and Memory Settings menus
7. CPU Clock Ratio is in both MIT and Advanced CPU Core settings menus
8. Apparently redundant naming for some settings in the UEFI. Eg; CPU LLC: Normal, Standard, Medium. Three settings, all essentially with the same meaning.
9. Extremely bad latency, possibly Agesa related and Windows 10 related. See below image for this one. Removing the memory sticks then putting them back in seems to fix latency for the first boot but then latency problem returns on the next boot or reboot.

latency_zpsilujomy1.jpg

Real world impact of this latency problem:
abysmal_zpspy8qghat.jpg

Edited by Dasa

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×