Thursday, March 2nd 2017

AMD Ryzen 7 2700X Spotted With a 3.7 GHz Base Clock, 4.1 GHz Turbo

AMD's next iteration on their very positively received Zen microarchitecture is preparing for take-off in the coming months, and as we draw ever close to the release date, more details are trickling in. This time, it's the appearance of a Ryzen 7 2700X (which supersedes the original Ryzen 7 1700X) on Futuremark's 3DMark database. The Ryzen 7 2700X was paired with an ASRock X370 Taichi motherboard (still considered one of the best ever to grace AMD's new AM4 platform), and its 8 cores and 16 threads are locked into a 3.7 GHz base and 4.1 GHz turbo clocks (respectively 300 MHz higher base and turbo clocks that the 1700X's).

The usage of AMD's XFR 2.0 (eXtended Frequency Range) and Precision Boost 2.0 could mean that the CPU is able, in certain scenarios, to turbo over the specified limit of 4.1 GHz, up to 4.2 GHz, thus delivering an even bigger boost to its performance. The usage of a 12 nm process means AMD has taken the power savings and increased frequency potential that comes from shrinking their original Zen microarchitecture, and put those to increased frequencies across the board, thus increasing their CPU's single-thread performance. Being an X chip,. AMD has kept the package TDP at a still respectable 95 W, much like its 1000 series Ryzens, though we know that this 95 W figure doesn't really spell out just how energy efficient these AMD CPUs really are.
Sources: ComputerBase, WCCFTech
Add your own comment

71 Comments on AMD Ryzen 7 2700X Spotted With a 3.7 GHz Base Clock, 4.1 GHz Turbo

#51
R0H1T
Caqde said:
Actually at these speeds the 2700X would be faster than the stock 1800X and possibly even an overclocked one depending on the workload.

To understand this realize that the 1800X runs at 3.6 to 4.0Ghz (without XFR) the 2700X is 3.7 to 4.1Ghz (No XFR) based on the specs given. This means that the 2700X is 2.4 - 2.7% faster than the 1800X. BUT because of the way turbo works on the 2700X this is not the whole story. Because the 1800X only hits the 4.1Ghz clockspeed during limited single core situations the 2700X with a more granular clockspeed shift between the max and min speeds will have situations where it can run at 4.175Ghz-3.9Ghz while an 1800X would only be running at 3.7Ghz meaning there are situations where the 2700X can be 12.8% faster than the 1800X at stock speeds.
Why yes, XFR now works how it was supposed to, on gen 1 Ryzen. The mXFR, on 2500u & 2700u, & XFR2 are essentially the same & if 12nm GF works as advertised then the gains could actually be even higher. Stock vs stock Ryzen 2xxx should be a pretty decent improvement over the last gen, there's also the possibility of slightly better IPC & much better memory support that could make the upgrade to 2xxx worthwile.
Posted on Reply
#52
phill
Easiest answer to any Ryzen 2 thread at the moment is, I'm looking forward to seeing the reviews and what they make of it.. No point in speculating anything at all as it's all guesswork, why not wait till the facts come in??

I am very very interested in the Ryzen CPUs in general.. If the Ryzen 2's make a decent improvement on what's already here, I'm in!! :)
Posted on Reply
#53
Valantar
R0H1T said:
Why yes, XFR now works how it was supposed to, on gen 1 Ryzen. The mXFR, on 2500u & 2700u, & XFR2 are essentially the same & if 12nm GF works as advertised then the gains could actually be even higher. Stock vs stock Ryzen 2xxx should be a pretty decent improvement over the last gen, there's also the possibility of slightly better IPC & much better memory support that could make the upgrade to 2xxx worthwile.
The IPC won't change, as this is literally the same arch. AMD has said so themselves. Same arch with minor tweaks (improved memory controller, Precision boost 2, some other tweaks) and a new process (which itself is an improved variant of the original 14nm, and not an actual brand new node).

Also, what do you mean "XFR now works how it was supposed to"? You mean by clocking as high as thermals and the die allows within the power budget, with no fixed limit? I suppose that's what we all want, but that's not what AMD made it out to be initially, nor what it is so far. XFR and mXFR are quite different things simply because mXFR lives in the strictly power-limited mobile world, where the inherent clock limits of the chips don't matter much. I suppose we'll see what happens with these chips, but given the probability of relatively minor process improvements, I'm not betting on XFR 2 being an "auto overclock" any more than XFR was.



Also, I have to lend my support to the people here objecting to the notion of upgrading your CPU year-on-year (or generation-on-generation). I mean, even if you excrete money from various orifices, that's meaningless. Pay $300 or more every year for a <10% performance increase in CPU-bound tasks, and far, far less than that in everything else? Yeah, that's dumb alright. Unless you're running a workstation or datacenter where processing time is crucial for profits. I got a 1600X last year after holding on to my Q9450 for almost nine years. That's a bit long, but upgrading your CPU more frequently than every 3+ years? Waste of money.
Posted on Reply
#54
InVasMani
I wonder how much this will shrink the 1080p high refresh rate game performance gap between AMD/Intel since that was the biggest critique knock against Ryzen to begin with. This minor Ryzen refresh defiantly is more appealing than it would be on a antiquated dual or quad core CPU.
Posted on Reply
#55
bug
InVasMani said:
I wonder how much this will shrink the 1080p high refresh rate game performance gap between AMD/Intel since that was the biggest critique knock against Ryzen to begin with.
Why? Outside of benchmarks the difference was small enough to matter only to hardcore gamers. You aren't one (otherwise you'd know the answer already), so really, why bother?
Posted on Reply
#56
Ravenas
I'm hoping for a polished version that handles 1080p and 4k gaming a little better that the initial installment of 1800x.
Posted on Reply
#57
Jism
GhostRyder said:
I am interested to see how far these will go. It would be nice to get 4.5ghz but I think that is a way a long shot. Here's hoping for 4.3ghz!
Knowing AMD the chips proberly pushed close to what the proces is capable of. This means that the headroom for an extra OC will be small.
Posted on Reply
#58
Valantar
Jism said:
Knowing AMD the chips proberly pushed close to what the proces is capable of. This means that the headroom for an extra OC will be small.
Well, considering that the 14nm process Ryzen is made on was never made for high clock rates at all (it's essentially a mobile/low power process), it's nigh on miraculous getting chips to 4.1GHz XFR on that with reasonable power draw. Given that the 12nm process is a refinement of the 14nm process, I expect higher clocks - as that is by far its biggest weakness - but no miracles. How high AMD decides to push clocks depends more on power than frequency, I believe - they'll go as high as they can before the voltage required spikes, to stay within the 95W power envelope. Which is exactly what they did with Ryzen, with excellent results.
Fiery said:
"AMD has kept the package TDP at a still respectable 95 W"

AMD has no reason to stick to 95W in the long run. Just sayin' ... ;)
Really? How about the commitment to Socket AM4 till 2020? Are you suggesting they launch a revised AM4-B made for higher power draw? 'Cause otherwise, they'd be making CPUs that exceed the spec of their own socket, meaning they'd kill the VRMs on low-end boards. That's not a good look, no matter how rare the combination of >120W monster CPU and $60 A320 board might be.

Also, for mainstream chips, >100W TDPs are silly. ~90W strikes a perfect balance of reasonably easy to cool and power with cheap coolers and motherboards, and pushing that just makes you look bad - just look at the whole 125W (or more!) Bulldozer and onward. On the other hand, an 8700K can relatively easily be OC'd to >200W with the right motherboard. Having a process that scales well gives you certain freedoms, but pushing power at stock just makes you look like your base performance is sub-par.
Posted on Reply
#59
TheGuruStud
Valantar said:
Well, considering that the 14nm process Ryzen is made on was never made for high clock rates at all (it's essentially a mobile/low power process), it's nigh on miraculous getting chips to 4.1GHz XFR on that with reasonable power draw. Given that the 12nm process is a refinement of the 14nm process, I expect higher clocks - as that is by far its biggest weakness - but no miracles. How high AMD decides to push clocks depends more on power than frequency, I believe - they'll go as high as they can before the voltage required spikes, to stay within the 95W power envelope. Which is exactly what they did with Ryzen, with excellent results.

Really? How about the commitment to Socket AM4 till 2020? Are you suggesting they launch a revised AM4-B made for higher power draw? 'Cause otherwise, they'd be making CPUs that exceed the spec of their own socket, meaning they'd kill the VRMs on low-end boards. That's not a good look, no matter how rare the combination of >120W monster CPU and $60 A320 board might be.

Also, for mainstream chips, >100W TDPs are silly. ~90W strikes a perfect balance of reasonably easy to cool and power with cheap coolers and motherboards, and pushing that just makes you look bad - just look at the whole 125W (or more!) Bulldozer and onward. On the other hand, an 8700K can relatively easily be OC'd to >200W with the right motherboard. Having a process that scales well gives you certain freedoms, but pushing power at stock just makes you look like your base performance is sub-par.
They can make any TDP chip they want. It was already done with the FX series. Boards were rated at 130W for the juice suckers.

I'm pretty sure every AM4 board except the cheapest are 125W capable and can be rated as such. We see people OCing hard on B350s. The only issue is VRM cooling on the cheapos, but the VRMs are stout enough. Any X370 will safely produce more power than necessary.
Posted on Reply
#60
Caqde
TheGuruStud said:
They can make any TDP chip they want. It was already done with the FX series. Boards were rated at 130W for the juice suckers.

I'm pretty sure every AM4 board except the cheapest are 125W capable and can be rated as such. We see people OCing hard on B350s. The only issue is VRM cooling on the cheapos, but the VRMs are stout enough. Any X370 will safely produce more power than necessary.
They don't need a chip with more than 95W to compete against Intel in the market. So I doubt they would do that. Now chips that are 95W that can achieve overclocks that push the chip to 125W+ that I can see. But we will see for sure in 1 month.
Posted on Reply
#61
TheGuruStud
Caqde said:
They don't need a chip with more than 95W to compete against Intel in the market. So I doubt they would do that. Now chips that are 95W that can achieve overclocks that push the chip to 125W+ that I can see. But we will see for sure in 1 month.
I meant for potentially high clocked zen 7nm. Obviously, this mobile LP process is completely maxed.
Posted on Reply
#62
Fiery
FinalWire / AIDA64 Developer
Valantar said:
Really? How about the commitment to Socket AM4 till 2020? Are you suggesting they launch a revised AM4-B made for higher power draw? 'Cause otherwise, they'd be making CPUs that exceed the spec of their own socket, meaning they'd kill the VRMs on low-end boards. That's not a good look, no matter how rare the combination of >120W monster CPU and $60 A320 board might be.

Also, for mainstream chips, >100W TDPs are silly. ~90W strikes a perfect balance of reasonably easy to cool and power with cheap coolers and motherboards, and pushing that just makes you look bad - just look at the whole 125W (or more!) Bulldozer and onward. On the other hand, an 8700K can relatively easily be OC'd to >200W with the right motherboard. Having a process that scales well gives you certain freedoms, but pushing power at stock just makes you look like your base performance is sub-par.
Yes, really. And even if it is the "same" AM4 socket, they can extend its capabilities at a platform change (e.g. going from X370 to X470, or from X470 to a future X570, etc) by retaining the backward compatibility of the socket. With X470 they can shift it up to for example 110W, and with X570 they can shift it up to 120W if necessary. All they have to do is properly specify that the higher TDP new parts are not compatible with older motherboards, unless the motherboard manufacturer assures that the extended TDP was already calculated into the older PCB design in order to allow a wide overclocking headroom.

BTW, an increased AM4 TDP would also enable some very interesting products, like a 6-core and an 8-core Raven Ridge with the same 11CU iGPU. I wouldn't discount the possibility of such a product, considering the "fact" that an 8-core Coffee Lake is inbound later this year.
Posted on Reply
#63
TheGuruStud
Fiery said:
Yes, really. And even if it is the "same" AM4 socket, they can extend its capabilities at a platform change (e.g. going from X370 to X470, or from X470 to a future X570, etc) by retaining the backward compatibility of the socket. With X470 they can shift it up to for example 110W, and with X570 they can shift it up to 120W if necessary. All they have to do is properly specify that the higher TDP new parts are not compatible with older motherboards, unless the motherboard manufacturer assures that the extended TDP was already calculated into the older PCB design in order to allow a wide overclocking headroom.

BTW, an increased AM4 TDP would also enable some very interesting products, like a 6-core and an 8-core Raven Ridge with the same 11CU iGPU. I wouldn't discount the possibility of such a product, considering the "fact" that an 8-core Coffee Lake is inbound later this year.
You're very optimistic for intel. I wouldn't be surprised if cannon lake is scrapped altogether for performance desktop and they just go to ice lake whenever they can next year. I don't think ryzen refresh warrants another coffee lake sku. They really want their yields right now and keep you paying extra for the 2066 platform.
Posted on Reply
#64
Vayra86
evernessince said:
At 4.5GHz, these chips would pretty much be equal to the 8700K in IPC. The 2900X might be 4.5 GHz if the 2700X is 4.1 GHz.
Eh?? Do you even know what IPC means?

At 4.5 Ghz Ryzen's net performance will rival Intels latest, but the IPC doesnt change...
Posted on Reply
#65
trparky
Vayra86 said:
At 4.5 Ghz Ryzen's net performance will rival Intels latest, but the IPC doesnt change...
Yeah, if you could clock it that high it would be great but unfortunately that's not possible now or even with Ryzen v1.5.
Posted on Reply
#66
Fiery
FinalWire / AIDA64 Developer
TheGuruStud said:
You're very optimistic for intel. I wouldn't be surprised if cannon lake is scrapped altogether for performance desktop and they just go to ice lake whenever they can next year. I don't think ryzen refresh warrants another coffee lake sku. They really want their yields right now and keep you paying extra for the 2066 platform.
Cannon Lake is not planned for the desktop at all, simply because currently Intel's 10nm process fares worse than the 2nd refresh of their 14nm process. So it would actually be an insane move for them to switch to 10nm right now. On the other hand, the 14nm process they use for Coffee Lake works so great by now, that it is not a problem to add a few more bits & pieces (like, I don't know, 2 extra x86 CPU cores) without hurting the TDP or the clock scaling too much. The LGA2066 platform offers 10 to 18 cores that the LGA1151 platform would never match. Not to mention those scenarios where the 8-core part maxes out the available memory bandwidth on the LGA1151 platform. Also, the LGA2066 platform should get a refresh this year where -- I suppose -- 20+ core parts will also be introduced. So an 8-core Coffee Lake would slightly cannibalize the LGA2066 market now, but it wouldn't be a problem by xmas this year.
Posted on Reply
#67
ratirt
IPC and frequency. these two come together but they are not the same. I see that people tend to mix these two up sometimes.
Posted on Reply
#68
bug
Fiery said:
Cannon Lake is not planned for the desktop at all, simply because currently Intel's 10nm process fares worse than the 2nd refresh of their 14nm process.
I believe there was a clear delimitation from the beginning: Coffee Lake for the desktop, Cannon Lake for mobile. Nothing to do with manufacturing (other than capacity).
See: https://en.wikipedia.org/wiki/Coffee_Lake
Posted on Reply
#69
Fiery
FinalWire / AIDA64 Developer
bug said:
I believe there was a clear delimitation from the beginning: Coffee Lake for the desktop, Cannon Lake for mobile. Nothing to do with manufacturing (other than capacity).
See: https://en.wikipedia.org/wiki/Coffee_Lake
Not from the beginning. Intel planned to do a complete refresh, Cannon Lake for desktop (-S), mobile (-H), ultrabook (-U) and tablet (-Y), but due to manufacturing difficulties they scrapped everything but the -U and -Y variants. And even the -U version is now on the verge of being cancelled. Funnily enough, the -Y variant may not hit the market as well in case Ice Lake-Y gets ready the same time Cannon Lake starts to provide acceptable yields in the factories. Intel is so late in fiddling around with their 10nm process that now the 2nd generation 10nm core (Ice Lake) is almost ready.
Posted on Reply
#70
bug
Fiery said:
Not from the beginning. Intel planned to do a complete refresh, Cannon Lake for desktop (-S), mobile (-H), ultrabook (-U) and tablet (-Y), but due to manufacturing difficulties they scrapped everything but the -U and -Y variants. And even the -U version is now on the verge of being cancelled. Funnily enough, the -Y variant may not hit the market as well in case Ice Lake-Y gets ready the same time Cannon Lake starts to provide acceptable yields in the factories. Intel is so late in fiddling around with their 10nm process that now the 2nd generation 10nm core (Ice Lake) is almost ready.
I don't recall any of that (but then again, Cannon Lake was always an oddity). Do you have any sources to share?
Posted on Reply
#71
Fiery
FinalWire / AIDA64 Developer
bug said:
I don't recall any of that (but then again, Cannon Lake was always an oddity). Do you have any sources to share?
Not any public sources for sure ;)
Posted on Reply
Add your own comment