Official AMD "Ryzen" CPU Discussion

Status
Not open for further replies.
TSO said:
Why do people keep saying ryzen 5 and 3 will be better for gaming when they will have half the cache?
Click to expand...

...because cache is not a good measure. An i5-7500 has a lot less cache then a i7-6950X and the i5 is better at single thread applications, and thus gaming.
 
m52nickerson said:
The Ryzen R5 and R3 might be better for gaming, or more specifically have better single core performance. Often CPUs of same architecture with less core will have better performance per the individual core. For the Intel Skylake some of the i3's have and now can be overclocked to higher frequencies then the i5's. The Ryzen 1800X hit a world record clock speed on a single core, by disabling all the other cores.

So we could very well see higher frequencies in the Ryzen R5 and R3 CPUs and thus better single core performance.

Even now if you have the $330 to spend on the Ryzen 1700 your not going to notice any issue with gaming performance, the point that has been getting lost is that you are not going to see any benefits unless you are running a very high end GPU and running 1080p resolution, which very few people do. So an i5 is a better option right now in terms of cost for a pure gaming rig. However paired with something more likely, such as a RX 480 or even a GTX 1070 you are not going have any problems gaming with a Ryzen CPU and you get plenty of power to do more then gaming. If you have the budget, no reason not to go Ryzen.
Click to expand...
Yeah, I'm expecting a higher clock on the quad core processors, but considering the surprisingly pitiful numbers we've seen on air for Ryzen, how optimistic are you that we'll see a genuinely aggressive stock clock like 4.2GHz+?

I can't say I'm optimistic. You saw what happened when he tried to push that 1800X to a 4.1GHz OC on Noctua coolers. I have a feeling that the 1700X/1800X won't be upended by the R5 release in the same way that the 7700K towers over the 6900K for gaming.
Yari said:
When we are talking about R5 we're just talking about price/performance right, it's not like those are going to be better for gaming than if I were to buy a 1700/1800, or is there a chance they would? (I've been curious about that)

I'll probably jump on the Ryzen as soon as they put out some fixes for all the issues out there right now. I think it is going to do what I require of it, as far as gaming goes. And the power it has for anything outside of gaming is just going to be a big bonus.
Click to expand...
I don't think everybody else is seeing what I'm seeing (if Amazon sales positioning is correct: 1700 = #6 ; 1800X = #8; 1700X = #12. I espy AMD repeating one strategy of bullshit marketing. They're selling the exact same processor under two different labels, and they don't appear to even have any sort of practical "binning" distance between them.

Enthusiast gamers are buying the WRONG Ryzen processors! The 1700 and 1800X. These are NOT the best purchases. The best purchase is the 1700X.

I've now seen three separate reviews where both 1700X and 1800X topped out at 4.0GHz on air. Maybe the 1800X can separate on liquid, but in terms of architecture, it's identical to the 1700X. So at 4.0GHz, there will be effectively zero difference between the two. That's not how binning is supposed to work! The higher-binned chips always are believed to have less theoretical headroom, but in practical testing, they always hold as least a portion of their lead from that natively higher clock.

This is the FX 9xxx series all over again! Except that they launched them as part of the same line on the same day. Stop buying the 1800X! You're getting the same damn chip and paying $100 more for it! The 1700X actually opens up the throttle over the 1700 (a higher TDP is typically a surefire indication a chip can sustain higher voltages, and that appears to be the case from everything I've seen posted so far).

This is how I would rank purchase strength based on the numbers we've seen:
  1. 1700X
  2. 1700
  3. don't do it
  4. 1800X
 
Madmick said:
Yeah, I'm expecting a higher clock on the quad core processors, but considering the surprisingly pitiful numbers we've seen on air for Ryzen, how optimistic are you that we'll see a genuinely aggressive stock clock like 4.2GHz+?

I can't say I'm optimistic. You saw what happened when he tried to push that 1800X to a 4.1GHz OC on Noctua coolers. I have a feeling that the 1700X/1800X won't be upended by the R5 release in the same way that the 7700K towers over the 6900K for gaming.

I don't think everybody else is seeing what I'm seeing (if Amazon sales positioning is correct: 1700 = #6 ; 1800X = #8; 1700X = #12. I espy AMD repeating one strategy of bullshit marketing. They're selling the exact same processor under two different labels, and they don't appear to even have any sort of practical "binning" distance between them.

Enthusiast gamers are buying the WRONG Ryzen processors! The 1700 and 1800X. These are NOT the best purchases. The best purchase is the 1700X.

I've now seen three separate reviews where both 1700X and 1800X topped out at 4.0GHz on air. Maybe the 1800X can separate on liquid, but in terms of architecture, it's identical to the 1700X. So at 4.0GHz, there will be effectively zero difference between the two. That's not how binning is supposed to work! The higher-binned chips always are believed to have less theoretical headroom, but in practical testing, they always hold as least a portion of their lead from that natively higher clock.

This is the FX 9xxx series all over again! Except that they launched them as part of the same line on the same day. Stop buying the 1800X! You're getting the same damn chip for $100 less. The 1700X actually opens up the throttle over the 1700 (a higher TDP is typically a surefire indication a chip can sustain higher voltages, and that appears to be the case from everything I've seen posted so far).

This is how I would rank purchase strength based on the numbers we've seen:
  1. 1700X
  2. 1700
  3. don't do it
  4. 1800X
Click to expand...

I don't know how optimistic I am at this point. AMD seems to have the a talent for being brilliant and shit at the same time. The only thin I can see that looks to have some hope for the R5 is the base clock are reported to be 4.0. So unless AMD has binned them right up to the limit, which is possible, you might be able to squeeze more out of them.

As for of the lineup is the best, so far I've seen the 1700 overclocked matching and beating the 1800X. With it being $70 less then the 1700X I think people have it right in this case.
 
m52nickerson said:
...because cache is not a good measure. An i5-7500 has a lot less cache then a i7-6950X and the i5 is better at single thread applications, and thus gaming.
Click to expand...
That example is the exception because the 6900 series are shit for gaming, the budget i7 also beats it in single core.
 
TSO said:
That example is the exception because the 6900 series are shit for gaming, the budget i7 also beats it in single core.
Click to expand...
That example isn't the "exception". It demonstrates the principle, TSO.

Using the same architecture, more cores = more heat. You get more overall performance for each 100Mhz you push up the clock if there are more cores, as I noted earlier in the thread, but because you generate more heat (presuming all cores are active), you will NOT be able to stabilize your clock as high as you would a chip build on the same architecture with half the cores. That's why overclocking world champions are always overclocking only a single core of the best overclocking chips.

Because of this, there's a trade-off. The extra cache will hope in most situations, certainly, but it won't matter as much as muscling through with more CPU cycles (because that's more total instructions).

So there's a tension between cache, cores, and clock (with IPC the elephant in the room). Right now, the most powerful gaming CPU's find that premium balance between the i5-7600K and i7-7770K where the gamer niche achieves its best performance.
 
Madmick said:
That example isn't the "exception". It demonstrates the principle, TSO.

Using the same architecture, more cores = more heat. You get more overall performance for each 100Mhz you push up the clock if there are more cores, as I noted earlier in the thread, but because you generate more heat (presuming all cores are active), you will NOT be able to stabilize your clock as high as you would a chip build on the same architecture with half the cores. That's why overclocking world champions are always overclocking only a single core of the best overclocking chips.

Because of this, there's a trade-off. The extra cache will hope in most situations, certainly, but it won't matter as much as muscling through with more CPU cycles (because that's more total instructions).

So there's a tension between cache, cores, and clock (with IPC the elephant in the room). Right now, the most powerful gaming CPU's find that premium balance between the i5-7600K and i7-7770K where the gamer niche achieves its best performance.
Click to expand...
So if you overclock just the first two cores wouldn't you have more room to overclock and stay stable since you wouldn't be generating more heat but would still have all the extra cache?

I've only just started playing with manually over clocking with this chip, all my other stuff I used the mobo dummy over clock if anything. But on my ryzen I've been able to overclock just the first two cores and leave the others alone and it didn't seem to affect my temp at all.
 
TSO said:
So if you overclock just the first two cores wouldn't you have more room to overclock and stay stable since you wouldn't be generating more heat but would still have all the extra cache?
Click to expand...
Yeah. Nevrtheless, generally speaking, L3 cache is shared, but I believe it relies on the added pipelines from each core's architecture to maximize performance drawing on that cache. Here's a blueprint of a Nehalem die (old Intel):

prXuv.jpg


Here's AMD"s Bulldozer design:

amd_diag_1.jpg


As you can see, there appears to be greater sequestration of the L3 cache by core in the AMD design, but effectively it's the same design strategy. The problem is that I think even a shared pool still generates heat, and also might be more likely to become bottlenecked by a lack of cores firing to maximize the lanes of data exchange. Basically, the L3 Cache is distributed across the cores even though it's shared, and that's why CPU's with more cores tend to carry more overall cache of every type. There's more space on the die, and more processing power requiring more intense cache demands to sometimes keep up.
I've only just started playing with manually over clocking with this chip, all my other stuff I used the mobo dummy over clock if anything. But on my ryzen I've been able to overclock just the first two cores and leave the others alone and it didn't seem to affect my temp at all.
Click to expand...
Theoretically, yes, but practically, the way everything is set up, everything will fight you. More to the point, it's just not worth it. Not a good idea to disable cores. Better strategy to buy a CPU with half the cores and a higher IPC like the 7700K if that is what you really desire. You're not the only one who has considered this strategy. Already came up in the Linus forums:
https://linustechtips.com/main/topic/746642-ryzen-overclocking-disabling-cores/

I'm seeing those Noctua overclocks get to 3.9GHz with the 1700. If you have an air cooler then it becomes probably the best value of the three, so you're good. It's also identical to the 1700X and 1800X in terms of pipelines, with the only difference that lower TDP and native clocking, so if it gets to 3.9GHz, then it should be roughly identical to the 1700X or 1800X at 3.9GHz, and just barely (~2.5%) inferior to them since they were able to achieve 4.0GHz on the same Noctua cooler.
 
This is fuuny AMD at their 32 ZEN core server chip announcement they gave details on ZEN 2 and ZEN 3.

Apparently they are going to ring out 10 to 15 percent better IPC then the ZEN chip. They say you can expect the ZEN 2 to ship at end of the year.

Zen 3 not till mid to end of 2018. The CEO talked about a massive increase in chip engineers working on the updates. She hinted at over 200 engineers and researchers working on the updates.

The 32 core chip could feature as many 19 billion transistors. It is targeted at purely server processing. They apparently did not show the chip but had a 64 cpu 128 thread machine crushing Intel's 24 core monsters.
 
Can someone explain to me the radical discrepancies we're seeing between the terrible average air and consumer market overclocks I'm seeing for these processors (ex. apart from what I've posted 4.1GHz is still the peak score I'm seeing on UserBenchmark) and what is apparently happening in the competitive custom liquid nitrogen overclocking world? It doesn't look to be just an issue of core count because it isn't only performing in single core feats. I'm genuinely confused as shit.
Hot Hardware > AMD Ryzen 7 1800X Overclocked To 5.8GHz On LN2, Benchmark World Record Shredded At 5.4GHz
Hot Hardware said:
It was only last week when the Internet was abuzz with an overclocker taking an AMD Ryzen 7 1800X processor and revving it all the way to 5.2GHz on LN2 cooling. That paved the way for a new Cinebench R15 world record, though that record no longer stands. Now barely a week later, an overclocker has pushed a presumably delidded Ryzen 7 1800X processor to 5.8GHz, the highest officially recorded clockspeed to date for a Ryzen CPU. And that Cinebench R15 score? Somebody has topped that, too.

The new frequency record was set by none other than Der8auer, a renowned overclocker who recently posted a video showing how to (and warning against) delidding a Ryzen processor. He explained in the video that he had destroyed two previous Ryzen 7 1700 processors in his attempts to remove the integrated heat spreader (IHS), and that he hoped the third time would be the charm. It was, and near the end of the video he talked about plans to attempt the same thing on a Ryzen 7 1800X processor to see if the effort to pry off an IHS is worth the gains that can be had from direct die cooling.

Ryzen_CPU-Z.jpg


It looks like he was successful. There is a picture of a delidded Ryzen processor accompanying his validated overclock, which sits at precisely 5,802.93MHz (5.8GHz) using a 44.5 multiplier and 130.4MHz bus speed. That is a 61.19 percent jump over the processor's stock clock, and as is usually the case in these world record attempts it took liquid nitrogen to make the feat possible.
giphy.gif
The overclock was achieved in an ASUS ROG Crosshair VI Hero motherboard based on AMD's X370 chipset with the CPU voltage cranked to 1.97V (not 0.787V as shown in the CPU-Z shot above). The setup also contained 16GB of DDR4-2400MHz memory. Der8auer says he did not disable any cores or threads during his record run.



In related news, browsing HWBot's website reveals that there's a new Cinebench R15 record. It was set by another renowned overclocker, Elmor, who now holds the record with a score of 2,454. It was set using a Ryzen 7 1800X processor overclocked to nearly 5.4GHz (5,346MHz), which was also achieved with an ASUS ROG Crosshair VI Hero motherboard.

Obviously all of this extreme overclocking falls into another realm well beyond every everyday use. However, it's interesting to see how far Ryzen's legs can stretch.
Click to expand...
Direct link to the 8-core Cinebench World Record Chart:
http://hwbot.org/benchmark/cinebench_-_r15/rankings#start=0#interval=20#cores=8
As you can see, it's legit. The Ryzen 1800X has the i7-5960X and i7-6900K both beat in Cinebench for bragging rights to being the #1 octacore processor in the world in that benchmark.

HWBOT Hall of Fame 8-Core World Records
  • wPrime 32m
    • 1.874
    • der8auer Jul 19th, 2015
    • Intel Core i7 5960X
    • ASUS Rampage V Extreme
  • wPrime 1024m
    • 57.312
    • der8auer Jul 19th, 2015
    • Intel Core i7 5960X
    • ASUS Rampage V Extreme
  • XTU
    • 3164
    • Dancop Jan 12th, 2016
    • Intel Core i7 5960X
    • ASUS Rampage V Extreme
    • G.SKILL Trident Z Seasonic Platinum
  • Cinebench R11.5
    • 27.49
    • Mr. DB! Mar 13th, 2017
    • AMD Ryzen 7 1800X
    • ASUS ROG Crosshair VI Hero
    • G.SKILL Ripjaws V
  • Cinebench R15
    • 2454
    • elmor Mar 2nd, 2017
    • AMD Ryzen 7 1800X
    • ASUS ROG Crosshair VI Hero
  • GPUPI for CPU 1B
    • 130.266
    • der8auer Mar 2nd, 2017
    • AMD Ryzen 7 1800X
    • ASUS ROG Crosshair VI Hero
  • Geekbench3 Multi Core
    • 46144
    • steponz Sep 24th, 2015
    • Intel Core i7 5960X
    • GIGABYTE X99-SOC Champion
  • HWBOT Prime
    • 9848.93
    • elmor Nov 17th, 2015
    • Intel Core i7 5960X
    • ASUS Rampage V Extreme
 
Last edited:
Wow5.8 ghz. I'm running my 1700 at 3.4 and that's where it will be staying.
 
Excited for the ryzen 5 cpu's coming out next month. It gives me a reason to upgrade my B350 to an X370 for my 1700 and use the B350 for a 1600/1500 budget build.
 
Raven Ridge, their next APU, won't be released for desktops until 2018
 
TSO said:
Excited for the ryzen 5 cpu's coming out next month. It gives me a reason to upgrade my B350 to an X370 for my 1700 and use the B350 for a 1600/1500 budget build.
Click to expand...
 
So I've been collecting parts for an ITX build and just got a Z270 ITX board with plans on getting a 7600K. Should I send this shit back and get the 1600X instead when it drops or just wait to see if the 7600K price goes down? I don't see any Ryzen ITX boards yet either so I don't want to wait forever.

This build will be for photoshop/illustrator and gaming.
 
Snubnoze707 said:
So I've been collecting parts for an ITX build and just got a Z270 ITX board with plans on getting a 7600K. Should I send this shit back and get the 1600X instead when it drops or just wait to see if the 7600K price goes down? I don't see any Ryzen ITX boards yet either so I don't want to wait forever.

This build will be for photoshop/illustrator and gaming.
Click to expand...
If you were looking for the top performance from a quad-core processor...you've already got it. The R7 1600 isn't going to be superior to any of the Kaby Lake i5's for gaming. That's what insight into the IPC gives us.

When Nickerson says that it might beat the i5 on value he means that it will give you more bang for the buck, not that it will give you more gaming bang. No way in the world, especially considering how quickly we're seeing the 1700/1700x/1800x hit the heat transfer wall (~4.2GHz) on air coolers. That is a strong indication that them just removing 2-4 cores with Zen architecture won't allow them to push up the frequency a massive margin, so don't expect to see a 4.5GHz core clock on the quad cores when they come out or anything (and if you see that it will probably be because they're repeating the FX-9xxx snake oil strategy).


*Edit*
In fact, as was indicated by Paul's review and several others, it looks like my earlier prediction was likely wrong, and the R7 1700 is the best purchase of the three. After all, the $85 Noctua D15 reportedly will take the 1700 to 3.9GHz vs. 4.0GHz for the 1700X/1800X, or possibly the same 4GHz. Regardless, if you used a 240mm/360mm Liquid AIO Cooler it won't matter because you shouldn't count on doing better than 4.1GHz-- if that-- with any of them.
So factory quoted TDP doesn't matter if that's the end difference.

Hell, even looking at UserBenchmark, here, a week down the road, with many more real-world tests ran, here are the best Peak Overclocked Scores for each of the three:
http://cpu.userbenchmark.com/Compare/AMD-Ryzen-7-1800X-vs-AMD-Ryzen-7-1700/3916vs3917
http://cpu.userbenchmark.com/Compare/AMD-Ryzen-7-1700X-vs-AMD-Ryzen-7-1700/3915vs3917
  • R7 1700
    • $310
    • 624 benchmarks
    • OC SC = 117 pts
    • OC MC = 1325 pts
  • R7 1700X
    • $370
    • 384 benchmarks
    • OC SC = 119 pts (+1.7%)
    • OC MC = 1347 pts (+1.7%)
  • R7 1800X
    • $470
    • 348 benchmarks
    • OC SC = 121 pts (+3.4%)
    • OC MC = 1365 pts (+3.0%)
  • i7 7700K
    • $310
    • 11,376 benchmarks
    • OC SC = 154 pts (+31.6%)
    • OC MC = 825 pts (-60.6%)
  • i5-7600K
    • $215
    • 3,487 benchmarks
    • OC SC = 151 pts (+29.1%)
    • OC MC = 596 pts (-122.3%)
That's a 2% & 3% difference, respectively, in horsepower when the R7 1700 lists for $310 from the most successful electronics reseller on eBay (antonline) if you don't want to to pay $330 to get it directly from Amazon. This pricing is identical to the 7700K. Let's not forget that the latter was $380 on Amazon a mere few weeks ago:
http://www.ebay.com/itm/like/292046914812?vectorid=229466&lgeo=1&item=292046914812&rmvSB=true
Pair that with a Nocta D15 SE-AM4, Noctua D15S (+2nd Noctua AF15 fan), or a compatible 240mm+ Liquid AIO Cooler. We're talking a price range for coolers that runs from $80-$200, here.That puts your total anywhere from $390-$530 depending on your choices.

Show me the Intel that can be argued as a competitive overall processing value at the highest end against an R7 1700 clocked @4.0GHz for under $370; as that is possible according to that PCPer reviewer who used a $58 Noctua U12S to hit the mark. Argument will come down to one of two philosophical strategies:
  • i7-7700K + that same Noctua cooler, or screw $400 and invest in something like the Corsair H100v2/H110i/H115 or NZXT Kraken X61/X62 to achieve the absolute peak gaming performance by supreme single core output in a 4+ core processor (sans custom cooling).
    • For the guys who want a CPU that can be part of a build claiming bragging rights to the top gaming machine out there.
  • i5-7600K + Cryorig H7 (or similar budget cooler) for the best gaming performance value at the highest end-- specifically only that as a focus.
    • For the guys who are willing to shave a few real-world fps off what the above guys are getting in pretty much everything, for a $150-$200 lower purchase price, and probably think the above guys are idiots for thinking their epeen stretches far when the only substantially superior capabilities of the i7's owned by them are pitiful when measured against what Ryzen owners possess for the same price.
Neither will touch the editing capabilities of an OC'd 1700. Also, some gaming CPU tasks can be elastic in terms of processing demand. Some of the most CPU intensive tasks in games are AI, physics, population & traffic density (major concern for MMO gamers), draw distance, view distance, level of detail, particle detail, reflections, and shadowing.

Population & Traffic + Draw & View Distances often aren't cranked up to true maximums or meaningfully tested in benchmarks (i.e. they didn't visit the busy capital city on a server, or they didn't apply customized settings to achieve a true peak draw/view distance as is commonly possible).
 
Last edited:
Status
Not open for further replies.

Similar threads

Madmick
Nintendo Nintendo Switch 2
2
Replies
24
Views
2K
method115
method115

Latest posts

Forum statistics

Threads
1,236,898
Messages
55,453,394
Members
174,785
Latest member
ljae89

Share this page

Back
Top