AMD Ryzen 7 1700X Review

[PaulAlcorn]

grammar police. the last use of the word "of" in the article should be "or"

Thanks

 

[falchard]

As a game developer, it makes sense to expand out to a reasonable amount of threads. You are developing for a wide range of systems, and building out the number of threads allows you to target a much wider amount of systems. For systems with less cores and more frequency, they can calculate the individual threads faster. For systems with more cores and less frequency, they can spread the workload. The only real issue is the design process is more tricky.

 

[mikhayahu]

Great review! Could you possibly do a piece on the effect of overclocked RAM on Ryzen performance? I've run into a few reviews around the web which claim significant performance improvements on Ryzen with faster, lower latency RAM. They say the speed of the "infinity fabric" interconnect is directly related to RAM speed. Anyway, I would love to see Tom's take on that.

 

[Kuriente]

What I would like to see are some benchmarks that apply to multitasking gamers. Most gamers I know don't JUST game. Many will live stream, record game footage, listen to music, have browsers open on a separate display, have downloads going in the background, etc... I would be curious if Intel's 7700K loses some of its edge when a few extra tasks get tossed its way while gaming. Personally, I think this, plus future-proofing ought to be considered when looking at Ryzen, not just how it performs when 100% dedicated to playing a few specific 2016 game titles.

 

[Jakko_]

Why was Dota 2 not tested? The only game patched for Ryzen so far, and why not use faster memory on the Ryzen system?

 

[chalabam]

I will no more stand the sloooooooow image loading. Sometimes images don't even load ever.

 

[Olle P]

Why did you only test 1080p performance? ...

You answered your own question: because a lower resolution shows the true power of the CPU. ...

Only if paired with the CPU usage numbers.
In for example BF1 on low resolution a 7700K at stock speed will run eight threads at >90% load most of the time. With the same settings a Ryzen 1800X at stock speed will spread the load over its 16 threads running at <50% load.
So yeah, the test show that the true power of Ryzen is almost twice that of the Kaby Lake.

AMD has confirmed that all R5 will be either 3+3 or 2+2. Sounds like a huge waste of silicon for the quad-core version but that's how it is.

One can assume that these CPUs will be the parts of the Ryzen 7 production that failed in production, so rather than just discarding the imperfect chips (which would be a waste) they're disabling some cores and re-brand them as Ryzen 5.
I wouldn't be surprised if new versions manufactured to be fewer cores will emerge later on when production is running smooth.

 

[dalauder]

You can add me to the "AMD blew it with marketing" pile. These first chips should have been released strictly as workstation chips that blow Intel out of the water on price/performance.

AMD should publicly be stating that in today's software ecosystem, the R3's or R5's will be the gaming sweetspot. There's no point in disappointing people as they'll find out what the real performance is the day the chip gets released.

 

[dalauder]

In for example BF1 on low resolution a 7700K at stock speed will run eight threads at >90% load most of the time. With the same settings a Ryzen 1800X at stock speed will spread the load over its 16 threads running at <50% load.
So yeah, the test show that the true power of Ryzen is almost twice that of the Kaby Lake.

Which brings us to how AMD kind of blew it on the marketing. Since we were going to find out anyways, they probably should've stated that R5's or R3's will be the gaming CPUs while the R7's will compete with the socket 2011 chipset and Xeons.

Then we'd see these benchmarks and go, "That's pretty good gaming for a workstation CPU. I wonder what their $160 R5 can do?" Instead we're left wondering why this CPU costs this much for power we can't use.

 

[pecul1ar]

@add other i5
Some of the people looking at Ryzen are looking for a new build from their old i7 or i5 so yeah, I hope Toms can indulge their longtime readers/lurkers.

@memory clocks
Kinda hoping Tom's would do a more in depth writeup about this and its effects on Ryzen. But as it stands, you guys have to change mobos (been following different forums and seeing Gigabyte ones correctly getting 3200 using specific memory) just to address this.

 

[mikhayahu]

Here is one of the reviews I mentioned above where OC'd RAM had a significant effect on Ryzen game performance. I am considering a Ryzen 5 for my next build, so I would love to see this independently verified (or shot down).

https://www.youtube.com/watch?v=RZS2XHcQdqA

 

[InvalidError]

One can assume that these CPUs will be the parts of the Ryzen 7 production that failed in production, so rather than just discarding the imperfect chips (which would be a waste) they're disabling some cores and re-brand them as Ryzen 5.

The L3 cache is bigger than the CPU cores and there is only one SKU with reduced cache. If the CPU core defect rate is so high that AMD expects up to four bad cores per die in sufficient volume to justify a quad-core CPU to sink all defects into, it seems unlikely that the defect rate in the L3 cache would be so low that AMD only needs one 8MB L3 SKU.

My guess is that a majority of quad/hex-core R5 is fully working but crippled as needed to meet demand, just like the Phenoms were. Many people are eager to find out if there will be a core re-enable hack like there was back then and see what is the likelihood of getting CPUs with genuinely defective cores.

 

[shknawe]

As far as gaming goes, a game is considered running at peak performance at 60 fps. Ryzen doesnt even breath hard at this level. For $399 you get $ 1000 6900k processing power. It has no problem meeting and exceeding 60fps in any game. So you are getting super bang for your buck, and future proofing. Don't get caught up in silly fps number game, 60 fps is the sweet spot anything over that is silly bragging rights. Imho Ryzen is a winner if silly gamers, which I am one, get over their my gaming rig is faster than your gaming rig mentality.

 

[dusty13]

One can assume that these CPUs will be the parts of the Ryzen 7 production that failed in production, so rather than just discarding the imperfect chips (which would be a waste) they're disabling some cores and re-brand them as Ryzen 5.

The L3 cache is bigger than the CPU cores and there is only one SKU with reduced cache. If the CPU core defect rate is so high that AMD expects up to four bad cores per die in sufficient volume to justify a quad-core CPU to sink all defects into, it seems unlikely that the defect rate in the L3 cache would be so low that AMD only needs one 8MB L3 SKU.

My guess is that a majority of quad/hex-core R5 is fully working but crippled as needed to meet demand, just like the Phenoms were. Many people are eager to find out if there will be a core re-enable hack like there was back then and see what is the likelihood of getting CPUs with genuinely defective cores.

i seriously doubt that.

hex cores will probably be crippled 8cores, either deactivated or because they hat 1-2 faulty cores.

for the quad cores however, amd would be just silly not to produce simple quad cores, especially since ryzen is basically two quad cores put together on a single die with a high bandwith bridge, so they already have a fully functional quad core unit at hand.

so i expect that R3 will be a single quad-core unit of ryzen. they have this anyway and need it too as their apu's will also work with that and so will their mobile chips. i highly doubt we will see 8 core / 16 thread mobile chips anytime soon due to the power restrictions unless maybe in some crazy uber gamer/workstation laptop.

crippling 4 out of 8 cores would not be a feasible business case. saving more than 50% die size at the cost of having a secont production line running is the only way to go economically, especially at the price amd intends to sell those chips and considering that those will be the higher volume parts too. entry-level always sells more units than highend. no sense in producing highend and selling it for entry.

 

[TJ Hooker]

@dusty13 you can doubt all you want but AMD's already confirmed that quad core Ryzen 5s use the same 8 core dies as Ryzen 7, but with two cores disabled on each CCX.

 

[10tacle]

Why did you only test 1080p performance? ...

You answered your own question: because a lower resolution shows the true power of the CPU. ...

Only if paired with the CPU usage numbers.
In for example BF1 on low resolution a 7700K at stock speed will run eight threads at >90% load most of the time. With the same settings a Ryzen 1800X at stock speed will spread the load over its 16 threads running at <50% load. So yeah, the test show that the true power of Ryzen is almost twice that of the Kaby Lake.

^^False logic. You can't compare those %CPU usage numbers directly and claim that one performs better than the other based on that. Apples and oranges. It's like directly comparing GHz speeds between AMD and Intel and saying that because the Intel CPU has a higher GHz speed it means the AMD chip must be slower. Completely different architectures cannot be compared directly. The only thing you can do is compare their performances directly. Which is what hardware review sites like Tom's does.

 

[10tacle]

Don't get caught up in silly fps number game, 60 fps is the sweet spot anything over that is silly bragging rights. Imho Ryzen is a winner if silly gamers, which I am one, get over their my gaming rig is faster than your gaming rig mentality.

^^People who have never gamed on a 120Hz-144Hz FreeSync or Gsync 1080p monitor always say what others "don't need." But yes, I do agree otherwise that this is one long overdue CPU from AMD and overall a knock out of the park. Not only on performance, but price. Once again AMD takes the lead overall on Intel IMO in this market. Next I'd like to see them take on the 4-core i5s and budget 2-core i3s at lower price points as well.

 

[InvalidError]

crippling 4 out of 8 cores would not be a feasible business case.

AMD has confirmed that the R5-1400 is 2+2 from the same dies as the R7, so there is no point in arguing about feasibility, this is a fact, it is done.

And no, removing a CCX does not reduce the die size by "more than 50%" as each CCX accounts for only ~25% of the die area. Nearly half of the R7's die size comes from the PCIe, DRAM, other integrated interfaces and miscellaneous support functions that have nothing to do with CCX/core count.

 

[dudmont]

Here's a question I've seen no one discuss(if it has, sorry for missing the discussion). Can you disable some of the ryzen's cores and get a much better overclock(say into the 4.4ghz range)? if so, then the gaming performance would close markedly, would it not? So little software is optimized for more than 4 threads anyways.
In the end, you'd not have to wait for developers to optimize for Ryzen, and once they do, you'd still have all 8 cores and 16 thread capability.

 

[-Fran-]

Here's a question I've seen no one discuss(if it has, sorry for missing the discussion). Can you disable some of the ryzen's cores and get a much better overclock(say into the 4.4ghz range)? if so, then the gaming performance would close markedly, would it not? So little software is optimized for more than 4 threads anyways.
In the end, you'd not have to wait for developers to optimize for Ryzen, and once they do, you'd still have all 8 cores and 16 thread capability.

Read the post above you.

Long story short: no.

Cheers!

EDIT: Missing letter.

 

[Avro Arrow]

All that this article has really proven is that CPU gaming benchmarks have become completely irrelevant. Even in Ashes of the Singularity, the title in which Ryzen had its "worst" showing, the R7-1700X's minimum frame rate was 39fps. Hell, even my old FX-8350 maintained a minimum fps of 31 which means that it would be either impossible or next to impossible to tell one game from the other based on CPU alone. This article has told me that I don't need to upgrade my FX-8350 yet, and probably won't have to for quite awhile.

People have to realise (ESPECIALLY THE AUTHORS OF THESE ARTICLES!) that this is NOT A GPU! This is a CPU which means that gaming performance, by definition, is its ability to keep out of the GPU's way, nothing more, nothing less! The ONLY job that a CPU has in gaming is to be fast enough to maintain a playable minimum frame rate. Everything else, from managing frame rate variance to minimising artifacts is the job of the GPU. The CPU doesn't cause frame rate variance, it just handles physics and AI. If it's fast enough not to slow the game to below 30fps, the CPU is successful at playing that game, end of story.

These tests are also irrelevant for future prediction because nobody knows what the future gaming situation will be. If we go by Tom's Hardware's own "Harbinger of Doom" reviews of my old FX-8350 (If you thought Ryzen got slammed, oh boy.. read the review of the FX-8350!), one would think that it would be only capable of 5fps average by now and yet, here we are with its worst showing being a minimum of 31fps in Ashes of the Singularity which means that my old FX-8350 would be 100% successful at properly running any of these games. Playing these games using any of these other CPUs wouldn't look or feel any different than gaming on my old FX-8350 provided that the amount of RAM and GPU are the same.

Now, the turn time on Civilization VI is 14s more per turn, which can be annoying but unless you're some Civilization freak, it's not enough to spend hundreds on a new CPU. For me, if the extra 14s were a problem, I'd just OC my CPU when playing that specific game to minimise it. For all the modern CPUs, the R7-1700X technically finished last but was only 3.74s slower per turn than the i7-7700K. Now, I've played several iterations of Civilization as have pretty much everyone else here since Sid Meier is easily the most successful Canadian video game programmer in history. We all know about waiting for the AI at the end of a turn. If less than 4s more per turn is enough to annoy you, then you definitely have a problem but it has absolutely nothing to do with your computer.

These days, focusing on gaming benchmarks for a CPU is irresponsible. Gaming should be looked at but not as a major feature of the CPU because it isn't, not anymore. Look at the sheer stupidity of those numbers! It's not like back in the day when hardware hadn't outpaced software to this degree and a CPU could actually drag a game down into unplayable frame rates. Reviews like these offend me because they put so much importance on something that is LITERALLY MEANINGLESS and completely ignores that which makes a CPU great - Its ability to enhance the overall computing experience of the user and a CPU can't make gaming better than another CPU that already games well. Based on everything I've seen, Ryzen does that better than EVERY Intel CPU in this test because of its incredible versatility.

This is what the conclusion of this article SHOULD BE:

Windows will install and load noticeably faster, programs will install faster, Windows will update faster and anti-virus programs won't slow things down so much. In addition to that, it can run ALL games perfectly!

 

[teekaypl]

What about 1440p results??? I heard AMD Ryzen beats Intel....

 

[InvalidError]

This is a CPU which means that gaming performance, by definition, is its ability to keep out of the GPU's way, nothing more, nothing less! The ONLY job that a CPU has in gaming is to be fast enough to maintain a playable minimum frame rate. Everything else, from managing frame rate variance to minimising artifacts is the job of the GPU.

Everyone has different thresholds for what the "minimum playable frame rate" is. For me, anything below a steady 40+ fps gives me headaches so I want to keep frame rates at 50+ as much as possible. As for frame time variance, the CPU does have an effect on that as you see the frame time variance go down with higher CPU and system RAM overclocks, along with tighter system RAM timings. If the CPU had no effect on variance, none of these would make any difference. The faster the CPU can be done with its part of the job for a given frame, the quicker it can hand it off to the GPU and the less total frame variance you get.

 

[dalauder]

As far as gaming goes, a game is considered running at peak performance at 60 fps. Ryzen doesnt even breath hard at this level. For $399 you get $ 1000 6900k processing power. It has no problem meeting and exceeding 60fps in any game. So you are getting super bang for your buck, and future proofing. Don't get caught up in silly fps number game, 60 fps is the sweet spot anything over that is silly bragging rights. Imho Ryzen is a winner if silly gamers, which I am one, get over their my gaming rig is faster than your gaming rig mentality.

You are forgetting about VR gaming. It would be nice if AMD had a viable chip for 90fps or so like many VR headsets need. As and game optimization stands today, there are some titles where Ryzen can't hit this target. I'm VERY happy with Ryzen and these hiccups are expected.

But there is room for relevant improvement.

 

[SomeTechDude]

Who cares if i7-7700K is faster. Do you really need more than 60fps and a nsync / freesync monitor to have a great experience?