• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

Page 28 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I was hoping R5 was going to released earlier though. I thought I read latter part of April. Can't remember though.
 
CentroX said:
I changed my motherboard from an Asus Prime pro to a ASUS CROSSHAIR VI HERO

I mean, why not. I want the extra VRM cooling, plus I want to use my 3200mhz DDR4 sticks.
Click to expand...

It has higher RAM official support than the Taichi but lower power phases (12 vs 16) and lower rated capacitors (10k vs 12k)... so I'm torn on that.

I guess we'll see on NDA lift which mobos are better or if it even matters.
 
ultima_trev said:
Someone posted benches on userbenchmark with the 1700X (albeit it claims turbo was disabled). Interestingly, the floating point and integer scores favor Intel (15-20ish %). Here's some comparisons to Ivy Bridge E and Haswell E:

http://cpu.userbenchmark.com/Compare/In ... 2580vs3915

http://cpu.userbenchmark.com/Compare/In ... 1976vs3915

Is this due to the Ryzen having turbo disabled or could it be a design oversight that could cause Ryzen to underperform?
Click to expand...

Probably because the results are scewed due to overclocks on the intel side.

I you look at the graph below, you can see where the results are including the deviation. They also report themselves that the Ryzen was overall faster. (taking into account that it is comparing an average result from intel with a single result from Ryzen)
 
The Stilt said:
- That QS SKU doesn't belong to a 1700 nor 1700X (1800X).
- There is a typo in the QS SKU
- There is a typo in the CPU-Z info (K17X)
- The clocks don't match with 1800X
- The clocks don't behave like that on Zeppelin

No idea if the leak is real, thou 🙂
Click to expand...
I'm going to take a "wild stab in the dark" and say it's faker than my cat's regret over destroying one of my laptops with water.
 
Both GTA 5 videos seems like fake. 1700X(stock) will be on par with 7700k(stock) in most of the games. Highly single threaded games can create advantage for 7700k but that kind of games are almost inexisting nowadays... So overall 1700x(stock) gaming benches won't be 10% less than even 7700k(5 Ghz).
 
Any thoughts on the highlighted block?

ioF3P3u.jpg
 
tipoo said:
Any thoughts on the highlighted block?

ioF3P3u.jpg
Click to expand...
Seronx suggest that what you mark as WTF is actually the memory controller. 2x72bit (9 big blocks on each side) one at that. I have done some digging and he is right, Haswell-E has 2 of these WTF things. So yes, it is probably 2 channel memory controller.
What you think are memory controllers are probably GMI links.
 
tipoo said:
Any thoughts on the highlighted block?

ioF3P3u.jpg
Click to expand...

The "memory controllers" look like DDR4 PHY. Seems odd that they are located on opposite sides of the die given the modules are located next to each other and AMD usually stacks their PHY as follows:

OEmIyQb.jpg


The "WTF?" area looks more like the PHY arrangement AMD would use - and supposedly they are using RAMBUS PHY, so we shouldn't strictly expect them to look the same.

That would then suggest that the "memory controllers" are GMI links. Their split arrangement makes more sense as you may want to connect another die or what-not in any location rather freely.
 
THE CCX are mirrored(you can see it quite easily in the high rez die shot) so going north/ south is probably easier.

And yes i agree this GMI links are the long interfaces that are together, they look nothing like any other DDR interface i have seen in any dieshot.
 
The AMD Ryzen 7 1700 3,0 GHz 8-Core Processor at a Glance:
ryzen_content2.png

  • 8 "Zen" cores with a low 65 Watt Thermal Design Power (TDP)
  • 16 Threads thanks to Simultaneous Multithreading (SMT)
  • Very high single- and multi-thread performance
  • 3,0 GHz Base- & 3,7 GHz Turbo Clock with Precision Boost
  • With automatic XFR overclocking feature
  • Unlocked multiplier for manual overclocking (unlocked CPU)
  • Large memory cache: 4 MB L2 & 16 MB L3
  • Highly efficient, non-planar 14 Nanometre FinFET transistors
  • AVX2 instruction set extensions for complex vector calculations
  • Two AES units for fast and reliable data encryption
  • Modern AM4 platform with extensive features
  • Supports energy efficient DDR4-RAM (Dual-Channel)
  • Includes AMD Wraith Spire CPU cooler with RGB LED ring!
https://www.caseking.de/en/amd-ryzen-7-1700-3-0-ghz-summit-ridge-socket-am4-boxed-hpam-118.html
 
Andrei. said:
Fantastic article. Demolishes the claim that single-threaded performance still trumps gaming. Ryzen will clean the floor with Intel in all segments, gaming included.
Click to expand...
Final reviews will reveal the truth, be it good or bad. Interestingly, Canard PC Hardware included Arma III and Anno 2070 in their gaming result. This way the 3.15GHz base and 3.4GHz boost clocked ES likely left some performance points on the table.
 
itsmydamnation said:
THE CCX are mirrored(you can see it quite easily in the high rez die shot) so going north/ south is probably easier.

And yes i agree this GMI links are the long interfaces that are together, they look nothing like any other DDR interface i have seen in any dieshot.
Click to expand...

They're definitely horizontally mirrored and they have uneven connections to the outside world.

And it seems my long-standing hyper-focus on PC CPUs has allowed me to ignore that AMD's two SoCs for consoles splits the memory controllers as well:

AMD-Xbox-One-PS4-APU-die-shot.jpg


Clearly they know how to avoid the issues with that (or even make it into an advantage...).

There are 172 segments in the GMI bus, if anyone is interested 😛

The strangest thing is that AMD didn't mirror the adjacent cores in the CCX nor the halves of the L3.
 
lolfail9001 said:
Seronx suggest that what you mark as WTF is actually the memory controller. 2x72bit (9 big blocks on each side) one at that. I have done some digging and he is right, Haswell-E has 2 of these WTF things. So yes, it is probably 2 channel memory controller.
What you think are memory controllers are probably GMI links.
Click to expand...
i dont see what you are seeing in the haswell-E die shot, and then go look at skylake or broadwell die shots, nothing looks like that "GMI" interface.
http://www.extremetech.com/wp-content/uploads/2014/08/haswell-e-die-shot-high-res.jpg

the top looks like pic-e and i/o
the bottom looks like DDR
 
itsmydamnation said:
the bottom looks like DDR
Click to expand...
Yes, i thought the bottom DDR too. For whatever reason, i did find it more resembling of the WTF thing, than what we think is actual mem controller.

Besides, if the bot/left top/right are memory controllers, then where is any sign of ECC support? Would not that be absurd to have a building block for server MCM without ECC support in memory controller?
 
Not really seeing any reason why I'd need to replace my i7-6700K so soon after getting it last year. It's nice to think that AMD and Intel will have some interesting options when it is time for an upgrade. Thanks AMD!

Sent from my SM-G930F using Tapatalk
 
Agent-47 said:
i feel sorry for AMD, they seemed to have bought quite a few 6900x for all these demos at $1050 only to see their prices drop by upto 50% in a few days times. Sad!
Click to expand...

They've barely dropped at all. They're still selling for over $900 on fleaBay (on the open auctions).
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top