*** AMD "Zen" thread (inc AM4/APU discussion) ***

[harney]

Why does 8 cores matter? Did we need better VRMs when we had 125W dual cores over 125W single cores, just because more cores? Did we need much better VRMs to be able to overclock a quad core at 125W over a dual core at 125W? Nope, cheap mobos with 'basic' VRMs performed near identically outside of extreme LN2/phase change overclocks, so why does a 95W octo core need better circuitry than... I don't know, 140W octo core Bulldozers, just because? Cores doesn't matter, power usage does, this uses less power than Bulldozer did at launch by over 40%, but now only a £250 mobo will overclock it?

Agree some thing seems a little off maybe things still need refining with the bios on certain boards re power ....tests i have done myself from cheap end boards to high end boards with same cpu mem psu ect ..all netted about the same in terms of OC ....Yes i understand more phase's help with cleaner more stable power ...but surly a decent b350 should still clock well ..or is it that these zen chips need vast amount of voltage to oc hence would need more phases..

i am hoping its just teething problems with the new arch

 

[DragonQ]

If the cooling is good enough and it will overclock by 200MHz doesn't that mean it will also overclock by 100Mhz or am I missing something?

XFR might only be for a single core. It may never clock to anywhere near 4.2 GHz on all 8 cores unless manually overclocked.

 

[CAT-THE-FIFTH]

It could be quite possible that the R7 1700X and R7 1800X are just leakier chips than the R7 1700,so can actually take more voltage and can hit higher clockspeeds??

Hence,it could be that the R7 1700 hits a voltage wall quicker than the R7 1700X and R7 1800X??

 

[Undesirable]

Seems a pretty crummy and misleading feature if XFR only works on one core.

 

[Zeed]

Totally agree. I made a comment earlier in the thread about the current pre order options of going with a cheap mobo+1800x vs. expensive mobo+1700, to get to 4Ghz and them both being the same price. Whats to say in a few months, after a few bios updates, the B350 boards at the lower end cant OC an 1800x? If you dont need dual front and back case soundblaster support, wifi, "killer" lan, fancy RGB disco going off in your case, why bother with it?

I have a crappy gigabyte £35 quid board in this current rig thats kept a 2500k at 4.2 for years. I'm just not buying it that you need the most expensive boards on the market to overclock these new chips.

4.2 is hardly an overclock on 2500k. I had mine for 3 years running constant 5ghz on maximus 5 gene and board did die Once on warranty

 

[Frozennova]

I imagine it'll be a similar story to Deneb chips. An average C2 stepping 955BEs did 3.6-3.8 where as the C3 steppings more or less always managed 4ghz.

I'd imagine we'll get revised stepping Ryzen chips in a years time

 

[CAT-THE-FIFTH]

4.2 is hardly an overclock on 2500k. I had mine for 3 years running constant 5ghz on maximus 5 gene and board did die Once on warranty

The worst thing is a locked Core i5 7600 probably costs more than a Core i5 2500K and his cheap motherboard.

 

[daman2k]

I guess it's a case of wait and see, I hope the b350 boards clock well as mATX seem all to be b350. I have no interest in having a huge full atx board / case.

 

[d_brennen]

On cheap AM3+ it was more than possible to fry boards that had wonky VRM spec with overclocked 125w+ Phenom IIs.

I did just this for giggles with a cheap 760 matx board.

If Ryzen's power curve is similar to Polaris (owing to process used) an overclocked 95w Ryzen will very quickly be drawing north of 150w. Just saying

 

[Silent_Scone]

Of course the VRM's are important. Overclocking a 8 core 16 thread chip will still require stable voltages across lots of phases.

To be fair, discussing the limitations of any VRM subsystem, let alone one that is yet to be properly investigated isn't really something that's really suited to forum posting. What remains to be seen is how much current these CPU really pull when straying from the path and overclocking, that will be the first step to understanding what is happening. As already pointed out, core count alone really isn't telling you anything here.

 

[harney]

The worst thing is a locked Core i5 7600 probably costs more than a Core i5 2500K and his cheap motherboard.

 

[Chrisc]

i was planning on waiting for the 1600x, ive got well too impatient and ordered the 1700, Im longer over due an upgrade from the 8120

 

[humbug]

Its starting to sound like these things don't really overclock at all, unless on Asus Crosshair or similar £250+ boards and even then only by about 5%, make you wonder why they are even unlocked, its a starting to look like yet another AMD gone wrong somewhere major, they should have saved themselves yet another label of that sort and humiliation and just locked them all, for now.

clearly there is no point in having an unlocked chip if they don't actually overclock, or at least not without *apparently* boiling its motherboards VRMs, just how much power are they using to do that to the VRMs? a recipe for another AMD hotter than the sun running gag? they could have done without a continuation of that, just lock them.

 

[Davedree]

I proved a lot of people wrong on the ability to clock an fx8320 to 4.8/5ghz on a £38 4+1 phase matx gigabyte board. Half of it came down to the fact the chip was golden needing only 1.45v. The thing is if you know what you're doing and the bios is configurable to disable/enable vrm overcurrent protection, then providing you don't take the mickey you should be able to clock within the safety margins of the b350 motherboards mosfet current capabilities. However this is 8c16t and that's a lot of current load when loading it up all cores.
It explains why there's no immediate itx formats yet.
I wouldn't be suprised if the 4c8t is better suited for itx too.

 

[CAT-THE-FIFTH]

Its starting to sound like these things don't really overclock at all, unless on Asus Crosshair or similar £250+ boards and even then only by about 5%, make you wonder why they are even unlocked, its a starting to look like yet another AMD gone wrong somewhere major, they should have saved themselves yet another label of that sort and humiliation and just locked them all, for now.

clearly there is no point in having an unlocked chip if they don't actually overclock, or at least not without *apparently* boiling its VRMs, a resipy for another AMD hotter than the sun running gag? they could have done without a continuation of that, just lock them.

TBF,we are talking about a 33% base clock overclock on a 8C CPU. We need to wait and see what the R7 1700X and R7 1800X bring to the table,but again its a first generation Finfet process,and see the problems we had with IB when Intel first pushed it out.

I expect the R7 1700 are 8C models which can't be validated as a R7 1700X or R7 1800X as they hit a clock and voltage wall quicker.

But TBH I also don't see why this anything different from the Phenom II X6 or FX8150 and FX8350. They all needed fairly decent motherboards to overclock at launch.

But even when the FX8350 launched,the FX6300 needed less cooling and less power delivery to hit a decent overclock,so the 4C and 6C models should hopefully be a bit better.

 

[drunkenmaster]

On cheap AM3+ it was more than possible to fry boards that had wonky VRM spec with overclocked 125w+ Phenom IIs.

I did just this for giggles with a cheap 760 matx board.

If Ryzen's power curve is similar to Polaris (owing to process used) an overclocked 95w Ryzen will very quickly be drawing north of 150w. Just saying

You do get that an overclocked FX8350 would use 200+ watts, and an overclocked Q6600 would use 200+W, and really any overvolted/overclocked CPu will use drastically more wattage than at stock, which again has never been a problem before.

There are 'cheap' boards and there are good value boards. I've never found any £70-100 boards from gigabyte/asus/whoever to struggle with overclocking if they are designed as overclocking boards and put the tinniest bit of effort into it. There are cheap boards that are bad value and skimp on everything, and those are just bad boards full stop.

In general, throughout the past 20 years, the majority of Asus, Gigabyte, MSI, etc, boards that are say £80 and up and designed to support overclocking have a VRM system designed to deliver well over 150W.

 

[xs2man]

Its starting to sound like these things don't really overclock at all, unless on Asus Crosshair or similar £250+ boards and even then only by about 5%, make you wonder why they are even unlocked, its a starting to look like yet another AMD gone wrong somewhere major, they should have saved themselves yet another label of that sort and humiliation and just locked them all, for now.

clearly there is no point in having an unlocked chip if they don't actually overclock, or at least not without *apparently* boiling its motherboards VRMs, just how much power are they using to do that to the VRMs? a recipe for another AMD hotter than the sun running gag? they could have done without a continuation of that, just lock them.

Wait a minute. Gibbo said that on the "lesser" boards he was only seeing 100-200 MHz less than on the "expensive" boards. That seems like a bargain at £100-150 (between boards) for more MHz, especially when you consider the cost difference between pre-binned 5.1 and 5.2 7700k's. That being said, I ran my 4770k for 3 years now at stock 3.5, boosting to around 3.7-3.8. So if I can double my core / thread count at the same speed and roughly the same IPC, then I'm happy enough with the 1700 at 3.8 MHz.

 

[CAT-THE-FIFTH]

You do get that an overclocked FX8350 would use 200+ watts, and an overclocked Q6600 would use 200+W, and really any overvolted/overclocked CPu will use drastically more wattage than at stock, which again has never been a problem before.

There are 'cheap' boards and there are good value boards. I've never found any £70-100 boards from gigabyte/asus/whoever to struggle with overclocking if they are designed as overclocking boards and put the tinniest bit of effort into it. There are cheap boards that are bad value and skimp on everything, and those are just bad boards full stop.

In general, throughout the past 20 years, the majority of Asus, Gigabyte, MSI, etc, boards that are say £80 and up and designed to support overclocking have a VRM system designed to deliver well over 150W.

Agreed - some of the AMD 970 motherboards were very solid with Phenom II X6 and FX6300/FX8300 CPUs but had VRM designs based on 990X motherboards.

 

[humbug]

TBF,we are talking about a 33% base clock overclock on a 8C CPU. We need to wait and see what the R7 1700X and R7 1800X bring to the table,but again its a first generation Finfet process,and see the problems we had with IB when Intel first pushed it out.

I expect the R7 1700 are 8C models which can't be validated as a R7 1700X or R7 1800X as they hit a clock and voltage wall quicker.

But TBH I also don't see why this anything different from the Phenom II X6 or FX8150 and FX8350. They all need fairly decent motherboards to overclock at launch.

But even when the FX8350 launched,the FX6300 needed less cooling and less power delivery to hit a decent overclock,so the 4C and 6C models should hopefully be a bit better.

I hope so because a 4Ghz limit of the best chip is not quite good enough to warrant the outlay of changing platforms, if the IPC is the same i was hoping for 4.2 to 4.4 in the same way my 4690K does 4.6Ghz to 4.8Ghz on a £90 board, its pointless spending £250 on board to get 4Ghz or maybe slightly more.

 

[CAT-THE-FIFTH]

I hope so because a 4Ghz limit of the best chip is not quite good enough to warrant the outlay of changing platforms, if the IPC is the same i was hoping for 4.2 to 4.4 in the same way my 4690K does 4.6Ghz to 4.8Ghz on a £90 board, its pointless spending £250 on board to get 4Ghz or maybe slightly more.

The best chip is the R7 1800X which is 4GHZ already on boost clockspeeds. If you are changing over from a Haswell Core i5 it will be for the fact you want or need 16T,ie,4 times what you have now. If you don't need or want 16T its best to wait for the 6C R5 1600X which is probably going to be £260 and already has a 4GHZ boost clock. If it is anything like the FX6300 and the FX8300 series ranges were,I suspect 6C Ryzen will be easier to overclock and cool.

Like I said a while back an option for you is to consider a secondhand Core i7 4790K as I suspect with what is going to be happening in the next few months,prices should start to drop on older Core i7 chips.