QOTW: Has AMD Ryzen 3rd Gen lived up to your expectations?

[Corky34]

AMD said ages ago that the Ryzen 3 chips were not going to be great overclockers. They just cannot take the massive hikes in TDP that Intel can to overclock. I mean what's not to like about 6c12t and 65W which is really 65W

Not disagreeing but i suspect PBO is a little broken ATM, theoretically we should be seeing slightly higher boost clocks when using PBO than manual OC'ing but that doesn't seem to be the case.

[tonyd2233]

Yes. My decision to buy the 2700x and Crosshair Hero a year ago has turned out worthwhile - i can drop a new cpu into my existing rig and get a performance boost... Will be very interested in the 3800x reviews as I've not seen one yet...

[3dcandy]

Not disagreeing but i suspect PBO is a little broken ATM, theoretically we should be seeing slightly higher boost clocks when using PBO than manual OC'ing but that doesn't seem to be the case.

Yeah deffo will be better when yields and BIOS's are a bit more up to date. Yes X570 boards are expensive but with Asus saying many pcie gen4 functions work on X470 boards I think I'll go that way to start for my long awaited work puter replacement. Currently on an X5670, 24 gigs 480 gig ssd R9380X setup. The Radeon is staying as no need for a better gaming card and the opencl performance is fine for now.
A nice Ryzen 3 with a decent nvme ssd and say 16 gigs ram to start will do me

[OldManInABoat]

I got a Ryzen 3600 this week, on a B450 motherboard, to replace a Core i5 4th gen.
A huge improvement in gaming, in PUBG i used to get massive stutters all over the place - it wasnt my graphics card (1660ti), but my CPU unable to process data quickly enough. Now i'm super smooth - i suspect my top end frames per seconds havent improved much, but my average has gone up a little, and my 1% lows will have climbed massively.
The experience of actually playing the game has changed from 'ok but annoying' to 'smooth and satisfying'.
Although i still cant shoot accurately.

[Corky34]

Yeah deffo will be better when yields and BIOS's are a bit more up to date. Yes X570 boards are expensive but with Asus saying many pcie gen4 functions work on X470 boards I think I'll go that way to start for my long awaited work puter replacement. Currently on an X5670, 24 gigs 480 gig ssd R9380X setup. The Radeon is staying as no need for a better gaming card and the opencl performance is fine for now.
A nice Ryzen 3 with a decent nvme ssd and say 16 gigs ram to start will do me

Personally i think X570 boards a fairly reasonable, at the low end they're around £50 more than X470 and around £100 more in the high to mid, boards with LN2 features are silly for norms (IMO) and priced accordingly. Most X570 boards have better VRMs than X470 so you can be pretty certain they'll take whatever AMD release in the next two years, possibly longer, and with PCIe 4.0 there's no need to worry if some future hardware is going to gimped, even though i think 'future proof' is a bit of a misnomer i think X570 will be fairly long lived.

[el_raberto]

To be honest I'm a little old and jaded to hold expectations about anybody's new hardware releases. A product arrives, reviews well and is in my budget, hey let's go for it. Ryzen 3600 would suit my needs for my next new build but not until new model sub £100 mobos arrive. I'm not messing about on the crap shoot of older boards being "compatible" but maybe only if your lucky to buy it shipping with the right BIOS. The new boards available right now are way too rich for my blood.

[msroadkill612]

It is simply testimony to how good their auto tuning hardware and software is. Get over it humans. Machines can do many things better these days.

[edmundhonda]

The total lack of overclocking headroom is the only thing I'm disappointed with, and that's not a dealbreaker ultimately.

I'm not messing about on the crap shoot of older boards being "compatible" but maybe only if your lucking to buy it shipping with the right BIOS.

As an example: the MSI B450 Tomahawk (£90 on scan, well reviewed and generally known as a good option) has the ability to flash the BIOS from USB without a CPU installed.

[3dcandy]

The total lack of overclocking headroom is the only thing I'm disappointed with, and that's not a dealbreaker ultimately.



As an example: the MSI B450 Tomahawk (£90 on scan, well reviewed and generally known as a good option) has the ability to flash the BIOS from USB without a CPU installed.

Yup - as I said above AMD said months ago the Ryzen 3 chips are not focused on headroom but IPC and efficiency. Again many chips run extremely well in a 65W envelope

[watercooled]

Not really. It means they can advertise a product that is "pushing the boundaries" but really they aren't producing the product on high quality materials and processes. How do we know even now that the silicone that is being used for zen2 is a good batch, where is the comparison? we wont know, because we don't even know what should be getting as the consumer. Back to the point at hand, great value and i just need biostar to sort out my bios issues

Silicon not silicone - two entirely different materials.

What do you mean about 'high quality materials and processes' exactly? Because that claim makes no sense whatsoever, TSMC's 7nm node is amongst, if not *the*, most advanced fabrication node currently in existence. Simply expecting insane clocks just because it's a smaller node is, well wrong. Just ask Intel about their 10nm node. Even by their own publications they're expecting clock regressions moving from 14 to 10.

By a 'good batch' I suspect you're implying they're sending 'good' dies to EPYC? That implies that die quality is a scalar quantity which isn't entirely true - dies for high-end Ryzen are intended to be clocked highly which doesn't necessarily overlap with binning for operation at lower powers which is more in line with what EPYC processors demand. Essentially, you can't really sum up die 'quality' with a single number in much the same way you can't sum up CPU performance in a single number/benchmark.

I don't understand how some are painting lack of manual OC headroom as a negative TBH. It probably arises from the fallacy that Intel have ended up creating by specifically binning special 'overclocking' CPUs, making you pay for that 'free' performance, and just disabling overclocking on the rest.

For those whinging about 'but but not 5GHzzzzz!!!' (not quoted poster I know) - that's just downright silly. I can't believe I've had to explain this to so many people! Clock speed in isolation is a meaningless number - 5GHz on one CPU is not the same as 5GHz on another. Zen2 manages to be faster than Skylake at 5GHz (or above) at many tasks and at least on-par for majority of the rest. So why is its absolute clock speed remotely important? Go buy a FX-9590 if you want 5GHz!!

In response to the OP, yes Zen2 is at least in line with what I expected based on recent announcements. Latency incurred by the multi-die approach turned out to be a non-issue, performance in most cases exceeded what I expected and game performance fell in line with what I reasonably expected with very few sticking points. Efficiency is excellent. Funny how people have gone silent over the whole power consumption argument again! Oh and memory speed seems to be far less critical than some feared. There's still something to gain by using faster memory, but it doesn't seem to have a major impact for most real workloads.

All in all a very solid launch for AMD in my opinion! Makes a nice change from AMD historically managing to get a good product to look bad on release for silly reasons! Release BIOS seemed sub-optimal but it reviewed very well despite that.

[ddr4000]

Help, I'm unable to activate my account. It keeps saying It's already activated but I need to log in, however I'm unable to do anything about it. I can't even contact forum staff because I don't have the permission without activating.

[FRISH]

For those whinging about 'but but not 5GHzzzzz!!!' (not quoted poster I know) - that's just downright silly. I can't believe I've had to explain this to so many people! Clock speed in isolation is a meaningless number - 5GHz on one CPU is not the same as 5GHz on another. Zen2 manages to be faster than Skylake at 5GHz (or above) at many tasks and at least on-par for majority of the rest. So why is its absolute clock speed remotely important? Go buy a FX-9590 if you want 5GHz!!

I don't think anyone here was actually whinging about not getting 5GHz. That doesn't mean it wouldn't have been nice for it to clock higher, particularly since AMDs video about precision boosting for AMD was not only talking about the advertised boosts, but mentioned going higher with 4.75GHz as an example. Sure it's something to take with a grain of salt, but still zen 2 isn't "destroying" intel as much as was lead on by those sensationalising the hype. They're worth recognising for providing the best competition in a decade, but at the same time, the gap would be larger if intel didn't fail to bring us 10nm in a timely manner. Sadly a big reason why they tend to cost more is because they can. Although frankly I am kind of surprised that the string of vulnerabilities hasn't hurt their reputation more.

[3dcandy]

I don't think anyone here was actually whinging about not getting 5GHz. That doesn't mean it wouldn't have been nice for it to clock higher, particularly since AMDs video about precision boosting for AMD was not only talking about the advertised boosts, but mentioned going higher with 4.75GHz as an example. Sure it's something to take with a grain of salt, but still zen 2 isn't "destroying" intel as much as was lead on by those sensationalising the hype. They're worth recognising for providing the best competition in a decade, but at the same time, the gap would be larger if intel didn't fail to bring us 10nm in a timely manner. Sadly a big reason why they tend to cost more is because they can. Although frankly I am kind of surprised that the string of vulnerabilities hasn't hurt their reputation more.

For many people it has.... but again many not going for a upgrade/crossgrade at this moment in time either

[watercooled]

I don't think anyone here was actually whinging about not getting 5GHz.

I saw a couple, but it was more of a general rant.

That doesn't mean it wouldn't have been nice for it to clock higher,

Isn't that a bit redundant?

As for the hype, that's why you wait for the actual product before jumping to conclusions. For me personally, performance is at least where I expected based on AMD's official information, and achieved it at lower clocks (so higher IPC) than I expected. The absolute clocks are totally irrelevant. Clock speed has an awful lot to do with microarchitecture/layout choices, not just the lithography. There's a reason GPUs don't clock nearly as high as CPUs on an identical node, and two different CPU cores produced on the same node will have difference achievable clock speeds. Heck, even different core layouts of the same core on the same node can be tuned for different clock speeds, just look at mobile SoCs! But there are trade-offs to make e.g. power efficiency and area.

In order to get a hypothetical Zen2 core to hit 5GHz reliably on the current 7nm node, AMD may (or may not, of course) have had to trade-off a considerably amount of efficiency, extend the pipeline and therefore increase misprediction penalties and instruction latency, make the caches smaller and therefore reduce cache hits, make the dies larger and therefore more expensive and lower-yielding, etc. Get my point? Worst case, designing the core to hit 5GHz whatever the cost may have resulted in inferior real performance, efficiency and/or cost. Of course, it's not that far off that all of that would be likely required, but you get the point, it's a trade-off, and one AMD will have considered. Modern CPU cores aren't just designed in a vacuum then chucked onto whatever node is available, they're very closely tied together.

[FRISH]

I saw a couple, but it was more of a general rant.

Isn't that a bit redundant?

As for the hype, that's why you wait for the actual product before jumping to conclusions. For me personally, performance is at least where I expected based on AMD's official information, and achieved it at lower clocks (so higher IPC) than I expected. The absolute clocks are totally irrelevant. Clock speed has an awful lot to do with microarchitecture/layout choices, not just the lithography. There's a reason GPUs don't clock nearly as high as CPUs on an identical node, and two different CPU cores produced on the same node will have difference achievable clock speeds. Heck, even different core layouts of the same core on the same node can be tuned for different clock speeds, just look at mobile SoCs! But there are trade-offs to make e.g. power efficiency and area.

In order to get a hypothetical Zen2 core to hit 5GHz reliably on the current 7nm node, AMD may (or may not, of course) have had to trade-off a considerably amount of efficiency, extend the pipeline and therefore increase misprediction penalties and instruction latency, make the caches smaller and therefore reduce cache hits, make the dies larger and therefore more expensive and lower-yielding, etc. Get my point? Worst case, designing the core to hit 5GHz whatever the cost may have resulted in inferior real performance, efficiency and/or cost. Of course, it's not that far off that all of that would be likely required, but you get the point, it's a trade-off, and one AMD will have considered. Modern CPU cores aren't just designed in a vacuum then chucked onto whatever node is available, they're very closely tied together.

Well yes, but we already knew IPCs were higher so it's not entirely fair to expect much more in that regard, and there's no complaints about core counts, that leaves frequency. Intel can unlock more performance, AMD less so. While AMD win in a lot of aspects, it's not a flat out winner across the board as hoped by many. For example while the 3600 and somewhat 3900x seem like better options due to their budget and multi-core performance, the 3700x and 9700k are more down to preference. The 3600x and probably 3800x seem like less attractive buys, but personally I'm fine with that since imo it enhances the value of the other CPUs. There's also been talk about making sure fixes for intel are included for benchmarks, a fair ask, but I do think those people may have expected intel to get hammered more than they have. I think one thing that has really helped AMD, is intel's stagnation. I think it's fair to have expected a little more, but that doesn't mean they're not worth buying, and I'm definitely interested in seeing AMD continue to compete with intel more than the past decade.

[=assassin=]

I'm pleased with the performance of AMD's new processors, but a little disappointed that the X570 boards mostly have tiny little fans on rather than being passively cooled, which I expect will result in some annoying whiny sound.