AMD Ryzen 1700X CPU Review @ [H]

[grtitan]

On a more serious note, given that my gaming pc is attached to my TV @ 1080P, Ryzen doesnt do much more for me, compared to my i5-3570K@4.2.

But for others, it still a good buy.

 

[Burticus]

Am I the only one excited about Ryzen 3? Right now you can't really get anything good in the $100-$150 price range; a locked 3.7GHz Core i3 isn't really acceptable in this day and age. R3 will be unlocked, and even with the poor voltage scaling we're seeing an R3 at 4GHz should be sub-100W for day to day use, and sub-50W at 3GHz. Pair that with a 1060 or a RX 480, 8GB of memory, and a 450W PSU and you've got solid HTPC/console replacement for $500, with excellent encoding performance to boot.
R3's niche is also something Intel can't immediately enroach on; AMD can sell cheap R3's harvested from otherwise-worthless 8c dies, but Intel doesn't have such a route - their 4c die is actually quite large because of the iGPU and it is unclear whether they can harvest i5's with defective GPU's. They could drop 7350K prices or add HT to the i5 to put pressure on AMD, but in any case us consumers win in the end.

Yeah I'm waiting for the lower binned chips too. Worst case it's a de-activated defective cores.... but given AMD's history with core unlocking... I'm very curious.

 

[Deleted member 134608]

Meh. AMD giving me the opportunity to spend more than I ever have to get a CPU thats half the price of a CPU I would never consider purchasing is not a value proposition to me. Especially since all I do is office apps and gaming. I'll stick with the Skylakes that are being cleared out at reasonable prices in Canada.

 

[SighTurtle]

Quick question for those of you who already went through the reviews, Win 10 vs Win 7, any type of interesting differences or is it something pretty much the same performance?

 

[Priller]

yes, you realize that prime95 is likely worse on Ryzen, so even in the worst case scenario, it's using less power than the 6900k.

Kyle posted Prime 95 power draw numbers. Did you think I was taking a shot at Ryzen or 6900K? I was doing neither.

Just surprised that Prime 95 would pull that much more power.

 

[Lastan010]

Ryzen's single core IPC performance at 4.2 ghz is better than Skylake's at same clock speed according to the review from Guru3d.

I think if AMD improves this CPU with future revisions to achive higher clock speeds I think Ryzen will also beat Intel in low res gaming.

 

[SixFootDuo]

If you guys consider everything going in and becoming a part of this experience from the very start when we first heard about Ryzen then one can very quickly understand that from the very start the 8 core / 16 treads Ryzen laid the ground work for that being it's most compelling feature. I am positive that AMD knew going in to the Ryzen's design it could not compete with Intel in game performance. People are easily nearly at least at first persuaded with glitz and flash and 8 cores / 16 threads is certainly that.

I spoke out against the lower single core threaded numbers and urge caution. Go re-visit all my past posts. I got a lot .. A LOT of hate over off WCCFtech for voicing this.

Still excited for AMD and all the new Ryzen owners. As a gamer however, I'm going take care of myself /hiphop slap of the chest two or three times then /throw up duces as I walk into the kitchen acting gangsta only to grab me a juicebox

 

[cyclone3d]

Wow they compared 3600 RAM against a bunch of 2133 RAM... and the 3600 RAM is faster?!? What madness is this?

Lack of reading comprehension FTL.

Look at the 7700k 3866 score and then the Ryzen 3600 score.
Ryzen wins by 4GB/s on Reads, and by 6GB/s on Writes even with the lower speed.

And even the 2133 score for Ryzen is about 4GB/s faster than 7700k at 2133.

The other 2133 setups there... are QUAD channel. 5820, 5930, 5960, 6950.

AMD has been way behind in RAM throughput in the past. This much of an increase.. and them even besting Intel is HUGE!

 

[NKD]

I read the entire thing, what AMD stated is half BS, the other half isn't even telling us where the problem really is, You have Kyle, D_Dan, and a few others, that have all stated what I and others that have more experience with this type of issue, there is an underlying problem that is not going to be easy to fix. Yet you bring in an article, where AMD will tell us what? Nothing but smoke so people will get nothing useful out of it?

Things like this just don't get fixed with setting changes, bios fixes, microcode fixes.

True and I do know and accept the fact you all have more knowledge about this. He did post that new new asus bios increased performance around 10%. So there is something that can be addressed. Also he said there were other people telling him that they are getting better performance with gigabyte board which he will get around to testing. Plus also he said this doesn't happen in all games. I wasn't talking about AMD, I was talking about editor stating those facts. Also stating windows even timer or something giving him 5% boost. Some of these issue can certainly be new platform teething issues. Not all but some of them.

You did seem to leave those parts out. My link was for readers to read the article in entirety and ofcourse AMDs excuse was lame but that wasn't the entire context of the page.

 

[sirmonkey1985]

I don't think it's bios related. I think AMD has just developed a more efficient power control that enables them to max things out on their chips. 1milliamp regulation etc.

ahh ok, just seemed really odd usually there's at least a couple 100 mhz variance between reviews at the least.

 

[sirmonkey1985]

Laughing at all the people making excuses for Ryzen's relatively poor gaming performance. Get over it.

i mean if you're buying a 400+ dollar cpu to game at a low ass resolution where the difference is noticeable you might have other problems, lol. i'm just glad there are more options out there now.

YES, I am quoting my own reply...

Take a look at the Guru3d review. They were able to bench the RAM at 3600.

The throughput absolutely smokes Intel in dual channel mode.

http://www.guru3d.com/articles_pages/amd_ryzen_7_1800x_processor_review,13.html

didn't notice guru3d tested it at 3600 thanks for pointing that out, looking at the review now.

 

[razor1]

True and I do know and accept the fact you all have more knowledge about this. He did post that new new asus bios increased performance around 10%. So there is something that can be addressed. Also he said there were other people telling him that they are getting better performance with gigabyte board which he will get around to testing. Plus also he said this doesn't happen in all games. I wasn't talking about AMD, I was talking about editor stating those facts. Also stating windows even timer or something giving him 5% boost. Some of these issue can certainly be new platform teething issues. Not all but some of them.

You did seem to leave those parts out. My link was for readers to read the article in entirety and ofcourse AMDs excuse was lame but that wasn't the entire context of the page.

Its a minimal increase to what we are seeing the deficit is. And have and a few others have stated, there can be tricks or ways to fix the issue to some degree. But how much err that is up in the air. Even that 10% i don't think is in all games :/.

I seriously did not expect to see Sandy Bridge performance in games and sometimes under, there are reviews that show this with 3200+ mhz memory as well, I was expected Broadwell level.

 

[Romeomium]

Doing more reading, it doesn't look like there'd be a benefit to disabling 7 cores, to try for a max OC on a single one due to the CCX structure. BUT, what if 4 cores were disabled? Any theories on if this would increase headroom? Obviously, people don't want to gimp an 8core chip, but it might give insight into how the R3's will perform. The problem with my theory is that I have no idea on what the R5 6 core chips will look like, unless 2 of the cores are physically disabled in some other way.

 

[SimGuy]

Doing more reading, it doesn't look like there'd be a benefit to disabling 7 cores, to try for a max OC on a single one due to the CCX structure. BUT, what if 4 cores were disabled? Any theories on if this would increase headroom? Obviously, people don't want to gimp an 8core chip, but it might give insight into how the R3's will perform. The problem with my theory is that I have no idea on what the R5 6 core chips will look like, unless 2 of the cores are physically disabled in some other way.

CCX being a cluster of 4 cores and 8MB L3, the question would be is the 6C R5 2x CCX's with 3/4 operational cores sharing L3 (ie: failed R7 QC on 1 of the members of the each CCX).

More importantly, is the CCX on the 4C R3 a set of 2x CCX's with 2/4 operational cores sharing L3 cache or a single CCX that passed validation 4/4...

 

[Romeomium]

Exactly.

 

[NKD]

I won't have any problem choosing ryzen. I mean there is absolutely no reason not to pick that one above everything. I understand here in this forum you might have hardcore gamers. I would like to see 4k and 1440p results for all games. I don't game much and even if I did I think it would be fine at 1440p. But I have no reason to upgrade since I have 6850k, but for those people that are streaming heavy, and multimedia heavy, or workstation envionment. This processor delivers hands down and lives up to hype.

This is a good article from TPU They have put together. There title is kinda messed up lol but the article is pretty fair. I keep saying at low resolutions it might be due to intel being the top dog for so long that all games are mostly designed and built on those systems. Is it hard to believe that no developer probably gives two shits about AMD systems due to the lack of innovation there? I am sure we will get some performance gains in games as not all games display the same behavior. I don't understand why it is so hard to believe. AMD has fully admitted that they have a lot of microcode updates to do on the memory performance side. We all knew how lock down memory settings and performance were. We also do know how memory performance can effect games at lower resolutions. If we see in difference by second half of the year then clearly is not as fast as intel on lower resolutions, but I believe they will make some improvements

check the last line out as well. Overall a smoother experience on AMD chips, so to each his own I guess.

 

[twzTechman]

Not seeing much of an incentive to upgrade from my 4770k. I do some occasional video encoding, but mostly do gaming and reading stuff like [H]ardOCP. I am tempted to jump at a Nividia 1080 with the recent price drop. That would give me more of an improvement for $500 than a new CPU/Motherboard/Memory/M2 SSD which is gonna run a minimum of $1000.

Also - i think there is some potential for the AMD systems to mature and likely improve in their gaming prowess.

 

[lostin3d]

Thank you for the review.

I refuse to subscribe to the obvious battles and really like how you pointed out the more pertinent points of 'what are you going to use it for?'

I use mine for both but I know people who will specialize one way or another since they can afford it.

Thanks again.

 

[Lastan010]

from the conclusion inside the Ryzen review: "For our single threaded and multi-threaded benchmarks, Ryzen turned in lackluster scores at best."

I fail to see where Ryzen did so lockluster in multi-threaded benchmarks.

according to guru3d review, Ryzen at 4.3ghz beat Skylake at 4.3ghz in single threaded and multithreaded benchmarks.

 

[Singularity_Survivor]

Ryzen when prices drop.

Lol, when prices drop, they are already roughly a third the cost of fellow 8 core Intel offerings.

 

[JustReason]

Sorry but I have to do this, no choice...
But AMD uses less watts at idle than Intel so You must buy it if you have to pay a lot for your electricity bill.

 

[DooKey]

I smell viral marketers.

 

[Simplyfun]

Sorry but I have to do this, no choice...
But AMD uses less watts at idle than Intel so You must buy it if you have to pay a lot for your electricity bill.

But this is how much it's going to be heard.

 

[Singularity_Survivor]

The R1700X is slower than the 7700K in *every* gaming benchmark. And the 7700K is $50 less.

I mean...I'm sorry that AMD didn't do better? If you need 8 cores then sure, go for Ryzen. For gaming I see no reason whatsoever to choose the AMD product over Intel's offering.

You do know a 4/8 Ryzen will cost about half what a 7700K sells for, right?

 

[Trimlock]

Well, the decision to buy one isn't easy for me, I don't need the performance in games and the cores would help but that platform is under whelming still.

So I can build a 1700 with a cheap mobo now and wait for Skylake X which is supposed to kick the chipset lanes up to 24x PCIe, or go all out on RyZen now and hope for some boards with Alpine Ridge?

Doesn't matter right now, with a lack of boards on the market right now I'm forced to wait. As it looks I may skip my want for a u.2 board and get Asus Pro + 1700x.

 

[Romeomium]

Limited solely by the architecture/process? Intellectual dishonesty is the same as lying........just saying.

What? Dude, GloFo is known for poor clock scaling. We have no idea how many chips runs they've taped out to perfect the process. I'm sure it will improve in the future, but yes - it could very well be limited by this. Also this is a new architecture - there are certainly tweaks to be made.

 

[WhoBeDaPlaya]

Limited solely by the architecture/process? Intellectual dishonesty is the same as lying........just saying.

Kyle ain't bullshittin'. Look at my post from a few pages back (or this thread on AT).

AMD pushed Ryzen outside the optimal bounds of GloFo's 14nm process :
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

Look at that big a*s voltage jump beyond ~3.9GHz.

 

[Ieldra]

Is it not possible that enabling SMT causes performance regression due to the increased potential for the windows scheduler to ignore CCX locality and assign data to threads on another CCX?

I'm guessing there's a major penalty for cache reads from the other complex, wasn't it limited to 18GB/s or something?

This would be easy to test if you could be certain that disabling 4 cores disabled one CCX entirely, if the performance regression from using SMT disappears then...

 

[Singularity_Survivor]

What? Dude, GloFo is known for poor clock scaling. We have no idea how many chips runs they've taped out to perfect the process. I'm sure it will improve in the future, but yes - it could very well be limited by this. Also this is a new architecture - there are certainly tweaks to be made.

I get the loyalty but don't become a Negan (a Kyle in this case).

 

[Singularity_Survivor]

Seems like I have zero incentive to upgrade from a 5960X. Oh well. Better luck next time, AMD!

How much did you pay for your 5960X?

 

[razor1]

I get the loyalty but don't become a Negan (a Kyle in this case).

Dude AMD is pushing the chip beyond thermal and voltage ideal ranges or near the ceiling of them so when overclocking it goes beyond, you will not see the exponential increase in power usage if they aren't. We have seen this from Fiji, Polaris and now Ryzen all at different frequencies, different processes too, but the attributing factors are node and architecture.

 

[Trimlock]

Is it not possible that enabling SMT causes performance regression due to the increased potential for the windows scheduler to ignore CCX locality and assign data to threads on another CCX?

I'm guessing there's a major penalty for cache reads from the other complex, wasn't it limited to 18GB/s or something?

This would be easy to test if you could be certain that disabling 4 cores disabled one CCX entirely, if the performance regression from using SMT disappears then...

I'm sure it's coming from conflicts that deal with how SMT prioritizes cache access. Does the two CCX's have the ability to cross share L2?

 

[Presbytier]

You do know a 4/8 Ryzen will cost about half what a 7700K sells for, right?

There is no reason to assume the 4/8 is going to due any better

 

[Trimlock]

Dude AMD is pushing the chip beyond thermal and voltage ideal ranges or near the ceiling of them so when overclocking it goes beyond, you will not see the exponential increase in power usage if they aren't. We have seen this from Fiji, Polaris and now Ryzen all at different frequencies, different processes too, but the attributing factors are node and architecture.

Some people just want to insult others because they don't only say nice things about their favorite company. I'm still trying to figure out how Kyle is lying, he's done a pretty good job so far with this review.

 

[razor1]

There is no reason to assume the 4/8 is going to due any better

Yep can't assume it, and them being speced out at lower frequencies is telling too.

 

[Trimlock]

There is no reason to assume the 4/8 is going to due any better

There's plenty of reason to assume it will do a lot worse, mainly because I think we have seen the max speed for this silicon.

 

[razor1]

Some people just want to insult others because they don't only say nice things about their favorite company. I'm still trying to figure out how Kyle is lying, he's done a pretty good job so far with this review.

Not just Kyle by his account all reviewers are lol. My my.

 

[Simplyfun]

Yep can't assume it, and them being speced out at lower frequencies is telling too.

You don't have to look far, the major jump to 65 W on the 1700 alone is saying it all. Having said that, I'm seriously happy with things the way they are.

Maybe they should term their new 1800X "pre-overclocked for your convenience" and make it a marketing strategy.

 

[Presbytier]

The funny thing is this does not seem like a bad chip, and further revisions may make it really shine (Zen+). But, there is no reason to get mad at people who are just a little disappointed in both its single threaded performance and gaming performance.

 

[razor1]

You don't have to look far, the major jump to 65 W on the 1700 alone is saying it all. Having said that, I'm seriously happy with things the way they are.

Maybe they should term their new 1800X "pre-overclocked for your convenience" and make it a marketing strategy.

actually yeah the x1700 is the best chip out of Ryzen's line up and probably will stay that way.