Looks like, potenitally as a result of Ryzen [though no doubt Intel will claim t'was their intentional all along] we'll finally see the back of 2c i3s
So they have rebranded the I5 an I3. Let's see if the unlocked one costs what the I5k used to. I also find it rather insulting that they are still selling locked CPUs. Old habits are going to die hard it would seem.
Thing is you could go to town on water cooling a Ryzen and still save yourself £200 over the Intel. Mind you having said that the Intels won't be that much better IMO. The 12 core one might be well behaved, but then it only runs 100mhz faster than my 14 core.
Got to have a usage for all these cores, most end users don't. Fun to talk about though, enough I think it will be many years before I own a cpu with 12 cores or more let alone 16 or 32 some of these have. It's funny to see Intel showing a cinebench score. The end of the duel core will be a sad thing though.
It's great that we've got real options out there for CPUs, but just a bit of a shame that for most people, CPUs from several gens ago are still perfectly adequate for most tasks. That might change once "more cores" becomes genuinely useful for most. In the meantime, this thread should probably just be renamed the "AMD Good, Intel Bad" thread based on the last couple of pages
Indeed, no matter the size of Intels R&D budget, the laws of physics still apply to them, so 18 cores at any half decent speed will definitely end up as a massive heatsource.
Yeah totally. My FX 8320 was perfectly behaved at stock. Crank it to 5ghz and the motherboard became a grill. Man, those VRMs used to get hot !
Do most of us need TR? Nah, I can get by just fine with the eight cores I have now, but the X370 platform is just going to be a limit in the future due to the pci-e lane shortage and so on, so I'll happily take the extra four cores of the 12 TR (no point upgrading from an eight core to an eight core TR), but I'll give TR a month or two before I jump on it so there is a decent selection of coolers as well as bios versions having matured a bit. I agree that for "normal" people outside the HW enthusiast realm the HEDT platforms don't have much of an appeal, but does that matter? It just gives us more choices which is never a bad thing.
Navi will use more than one GPU core on a die. So basically a quad core GPU, for example. I'm not sure how that will pan out PCIE lane wise, but you could find GPUs using far more than just 16 lanes as we go into the future. I think AMD (and Nvidia, they have now mentioned it too) are getting sick of making big single dies when multiple cored GPUs could work so well. IIRC the next iterations of the PS4 and Xbone will use some form of Navi.
Here's my final standing on my rig... I won't be tuning it much more. I can get to 3200MHz on the memory with 12-11-11-31-45-2T timings, but it needs a hefty 1.55v (perhaps more) to be able to pass some tests. Nah, I'll stick with this and 1.46v or so on the memory.
So I was actually reading some Threadripper reviews on my lunch break (instead of just skipping to the graphs and mentally shouting "Hah! Take that, Intel!"). Seems like AMD's claims of only using the top binned chips looks fairly accurate. Most are happy to do similar clocks to the 1700/1800X, but at a considerably lower voltage, usually between 1.35 and 1.375. The huge temperature offset of 27 degrees also makes the temperate results a lot better than the charts suggest... Probably helped by that lower voltage, too. Be interesting to see how the platform matures over the next year, but it could force my hand to start a shopping list a bit earlier than planned (but still not until pay-rise month 2018 haha).
CLICK HERE FOR BIGGER IMAGE Happy with mine, Cpu can do 4ghz but i choose to run 3.9ghz as it requires a bit less voltage. Ram is samsung b die based kit, running at 3600mhz with tweaked sub timings 1.45v vdimm and 1.1v SOC.
Are we getting a straight performance improvement with the faster RAM, if so is this going drive memory speed up again?
You do get direct speed improvements, yes. The faster your memory operates, the faster the CCX (link between cores) works. 2133 to 3200 is over 10% improvement alone, so the extra little bit will continually increase the speed of it too.
You get a performance boost yes, but only if you can find a decent balance between ram mhz and timings. If you have fast ram and abysmal loose timings you don't really see any improvement over a slower tighter kit. Sometimes its better to run a slower kit with tighter timings eg: 3200mhz c13/c14 with tweaked sub timings will be faster than 3600mhz c16 with the sub timings left on auto. RYZEN performs better with faster memory due to what TheMadDutchDude mentioned above, but it only does up to a point, That's due to Ram timings that need to be tweaked manually. In my testing on RYZEN it actually seem's better to run your ram at about 3200mhz with tight timings as it requires less ram voltage, less SOC voltage and you get around the same performance. I only run at 3600mhz because my system only requires 1.1v on the soc with 1.45v vdimm to be stable, most systems wont be able to run that at those volts if at all.
Unfortunately, uneven initial CAS timings aren't a thing on AMD Ryzen just yet, so you can't choose C13 or C15 for example, that's why I am now running 3476 (when my board returns from RMA) at 14-13-13-28-45-2T. I can run at 1T but the cold boot training is rather hit and miss, and I can't be bothered with it for the tiny bit of improvement it offers. I can get away with those RAM timings and frequency at 1.4v (BIOS set) which gives me approx 1.46v after LLC (cannot be controlled, annoyingly!) kicks in.
