The Ryzen vs Coffee Lake choice.

[Abwx]

8600K would let you OC to erase that deficit, but it costs more, and needs a more expensive MB.

Most annoying is that it s a i5 and Intel s segmentation has been very tricky in the past for this line of CPU, to sumarize they do not hold a single candle when it comes to multitasking when compared to the i7 or even AMD s FXs, dunno if they kept segmenting this way with CFL i5s...

So far such tests have not been done by Computerbase on recent CPUs since Ryzen was released, but i would be curious to check what it is about with the current most sold CPUs.

 

[BSim500]

Overclocking with a K chip does not disable turbo. In effect you are overriding the max turbo frequency with what you set.

"Override" is what I meant. Thanks for the correction.

These speeds are not guaranteed.

Yes they are if the chip doesn't hit thermal / platform power restrictions as every review of the chips so far have shown and 0% of the people claiming otherwise have demonstrated.

 

[PeterScott]

A whole lot of confusion here between Base vs Turbo vs MCE vs Overclock:-

- Base clocks on i5//i7 Turbo are applicable only if Turbo Boost is disabled in the BIOS or if the motherboard is set to have a power / time limit for Turbo for thermal / power envelope reasons (often done on laptops / NUC's to ensure, eg, constant maxed out Turbo for hours on end doesn't fry a 55w power brick PSU even if the CPU cooler can handle it). For desktops, as long as the thermals are good, the Turbo's are:-

i7-8700K = 3.7GHz Base. 4.7 (1C) / 4.6 (2C) / 4.4 (4C) / 4.3 (6C)
i7-8700 = 3.2GHz Base. 4.6 (1C) / 4.5 (2C) / 4.3 (4C) / 4.3 (6C)
i5-8600K = 3.6GHz Base. 4.3 (1C) / 4.2 (2C) / 4.2 (4C) / 4.1 (6C)
i5-8400 = 2.8GHz Base. 4.0 (1C) / 3.9 (2C) / 3.9 (4C) / 3.8 (6C)
i3-8350K = 4.0GHz Base (No Turbo but Unlocked)
i3-8100 = 3.6GHz Fixed

- These Turbo's are guaranteed to work at those freqs under constant load, ie, an i5-8400 is pretty much a 3.8-4.0GHz chip and an i7-8700 a 4.3-4.6GHz chip, as long as temps aren't hitting throttle limit. That's what separates a Turbo from an OC - Turbo is officially guaranteed (even on non-K chips), an OC isn't (even on K chips).

- Z boards allows overclocking (and raising clocks on a K chip to above Turbo speeds will disable override default Turbo's anyway), but all boards including cheaper H and B boards can hold Turbo at rated speeds just fine.

- MCE forces "all load cores" to the highest 1T core frequency but this isn't an official feature and whether this works or not may be motherboard vendor specific

- CPU speeds normally fluctuate under non maxed gaming loads due to SpeedStep and you'll often see 3-4GHz variations as cores are lightly loaded (down to 1GHz for lightweight / older games). This isn't the same as thermal / power / Turbo throttling. Disabling this whilst benchmarking (High Performance Power Option or at least advanced setting "Minimum Processor State" to 100%) can clear up confusion.

- There's nothing abnormal about low listed base speeds on the slowest i5 variant and is simply how Intel's marketing works on all previous chips too (eg, i5-7400 = 3.0GHz, i5-6400 = 2.7GHz, i5-4430 = 3.0GHz, i5-3330 = 3.0GHz, i5-2300 = 2.8GHz, etc). It has nothing to do with "binning limitations because there's a big gap". Remember prior to Haswell, both Sandy + Ivy Bridge chips had that +4-bin limited OC on non-K chips that allowed +400MHz on top of max Turbo which created similar large gaps, eg, i5-3450 = 3.1GHz Base / 3.9GHz max Turbo.

Pretty much any software monitor that polls then logs CPU freq several times per second (eg, HWInfo64) under an all core load shows this up quite clearly. As for 8400's "real speeds", it's exactly the above Turbo speeds as multiple review sites have already checked : For 6C loads (Examples 1, 2), 2-4C loads (Example 1) and 1C loads (Examples 1, 2).

I guess the only unanswered question is what MCE will do with non-K processors, some indicate this used to work on Non-K processors. But remember a review were it did NOT work on the 8400, so I am thinking Intel closed this loophole, though it is only the difference between 3.8GHz and 4.0GHz in the case of the 8400.

 

[jpiniero]

MCE should be disabled for Non-K. I don't remember when this changed, might have been Skylake.

 

[Smoblikat]

Neither, im waiting for Ryzen 2 to come out. My 3770K is doing just fine, especially with a 980Ti at 1080P.

 

[BSim500]

I guess the only unanswered question is what MCE will do with non-K processors, some indicate this used to work on Non-K processors. But remember a review were it did NOT work on the 8400, so I am thinking Intel closed this loophole, though it is only the difference between 3.8GHz and 4.0GHz in the case of the 8400.

Based on previous generations, MCE is one of those things that's been very motherboard specific in the past. There's no real way of knowing for sure in advance that a certain board you're interested in works until someone with one tests it.

 

[PeterScott]

Based on previous generations, MCE is one of those things that's been very motherboard specific in the past. There's no real way of knowing for sure in advance that a certain board you're interested in works until someone with one tests it.

Sure, But even on MB with MCE working, from the sound of things, it now only works on K processors.

IMO the only benefit of MCE would be if you could use it to essentially OC, a Non K processor.

On a K processor I think you are much better off manually overclocking.

 

[epsilon84]

A tentative vote to CFL because I'm still waiting to see what Ryzen 2 brings to the table, but I don't expect AMD to be able to completely bridge the gap in gaming (happy to be proven wrong though) Hopefully by then nVidias Volta (and new Vega GPUs) are released and I can finally build a kickass gaming system and retire my 2500K rig.

 

[PeterScott]

A tentative vote to CFL because I'm still waiting to see what Ryzen 2 brings to the table, but I don't expect AMD to be able to completely bridge the gap in gaming (happy to be proven wrong though) Hopefully by then nVidias Volta (and new Vega GPUs) are released and I can finally build a kickass gaming system and retire my 2500K rig.

Ryzen 2 is due in 2019 AFAIK. I may wait a couple of months, but not a couple of years.

 

[epsilon84]

Ryzen 2 is due in 2019 AFAIK. I may wait a couple of months, but not a couple of years.

Ahh, I meant the Ryzen refresh aka Pinnacle Ridge. I suppose that will still be sold as a Ryzen chip, and not Ryzen 2?

Anyhow, PR is supposed to bring higher clockspeeds, not sure about IPC though? As I said earlier, I doubt it will be enough to dethrone Intel at gaming, but I'm holding on to my current system for a few more months so am happy to see final performance numbers before making a decision.

 

[PeterScott]

Pinnacle Ridge is has no architecture change AFAIK, it's a process tweak, but yeah, there should be some clock speed boost.

Now that I have decided to upgrade, I won't be waiting for Pinnacle Ridge, Unless AMD surprise announces it's coming soon, which I don't expect.

Future chip products, tend to be late more than early in recent years.

 

[Topweasel]

I think this will be the confusion with the CPU name being so close to the codename.

What we know of as Zen+ will likely be called Ryzen 2 (or specifically they will be the Ryzen 2k series). Zen 2 being Ryzen 3K.

 

[TahoeDust]

How much can AMD increase clock speed realistically? The problem is, until they match IPC and clock speed, their only way to compete is more cores. If AMD gets too close, intel can just match the core count in the mainstream. It would cut into their HEDT line...but only picking off the 8 core chip. Their HEDT line goes all the way up 18 cores now. It would leave plenty of skews there.

AMD is pushing intel to give use more cores per dollar and that is great. Thanks AMD.

 

[Eric1987]

Pretty sure we'd get the IPC from the clock speed.

 

[raghu78]

How much can AMD increase clock speed realistically? The problem is, until they match IPC and clock speed, their only way to compete is more cores. If AMD gets too close, intel can just match the core count in the mainstream. It would cut into their HEDT line...but only picking off the 8 core chip. Their HEDT line goes all the way up 18 cores now. It would leave plenty of skews there.
AMD is pushing intel to give use more cores per dollar and that is great. Thanks AMD.

I think with Ryzen 2000 series we could see clocks speeds around 4.5-4.6 Ghz as the 12LP process has transistor level enhancements to improve performance and AMD could further optimize the physical design to hit higher frequencies. AMD could also tweak fabric speeds to improve memory latency and improve gaming perf which is their biggest weakness against CFL. But I agree that Intel's ST and gaming perf lead might reduce but they will still be ahead. I think the 2019 contest could end up very close as AMD Zen 2 is the first big microarchitectural improvement from Zen and AMD will have significant room for IPC improvements given that Zen is a brand new core and there is bound to be lessons learnt and weaknesses identified in the current Zen core. Zen 2 will most likely be fabbed at GF 7LP which is a high performance node designed for AMD and IBM. GF 7LP is very close to Intel 10nm in transistor density and should allow very high frequency CPU design given that the process is designed for 5 Ghz operation. AMD will have its best opportunity with Zen 2 on GF 7LP in 2 decades to compete against Intel.

 

[DooKey]

I think with Ryzen 2000 series we could see clocks speeds around 4.5-4.6 Ghz as the 12LP process has transistor level enhancements to improve performance and AMD could further optimize the physical design to hit higher frequencies. AMD could also tweak fabric speeds to improve memory latency and improve gaming perf which is their biggest weakness against CFL. But I agree that Intel's ST and gaming perf lead might reduce but they will still be ahead. I think the 2019 contest could end up very close as AMD Zen 2 is the first big microarchitectural improvement from Zen and AMD will have significant room for IPC improvements given that Zen is a brand new core and there is bound to be lessons learnt and weaknesses identified in the current Zen core. Zen 2 will most likely be fabbed at GF 7LP which is a high performance node designed for AMD and IBM. GF 7LP is very close to Intel 10nm in transistor density and should allow very high frequency CPU design given that the process is designed for 5 Ghz operation. AMD will have its best opportunity with Zen 2 on GF 7LP in 2 decades to compete against Intel.

I'm not going to hold my breath for any of that considering your overly optimistic predictions of the past.

 

[epsilon84]

How much can AMD increase clock speed realistically? The problem is, until they match IPC and clock speed, their only way to compete is more cores. If AMD gets too close, intel can just match the core count in the mainstream. It would cut into their HEDT line...but only picking off the 8 core chip. Their HEDT line goes all the way up 18 cores now. It would leave plenty of skews there.

AMD is pushing intel to give use more cores per dollar and that is great. Thanks AMD.

Realistically, I think PR will enable AMD to get a few more hundred MHz. So maybe we will look at stock 4GHz CPUs (talking base/multi core turbo btw, I know XFR already exceeds 4GHz)

A stock 4GHz Ryzen will beat a stock 8700K in most MT tests, but still be behind in gaming due to the IPC and clockspeed deficit, as you mentioned.

 

[TahoeDust]

I think with Ryzen 2000 series we could see clocks speeds around 4.5-4.6 Ghz as the 12LP process has transistor level enhancements to improve performance and AMD could further optimize the physical design to hit higher frequencies. AMD could also tweak fabric speeds to improve memory latency and improve gaming perf which is their biggest weakness against CFL. But I agree that Intel's ST and gaming perf lead might reduce but they will still be ahead. I think the 2019 contest could end up very close as AMD Zen 2 is the first big microarchitectural improvement from Zen and AMD will have significant room for IPC improvements given that Zen is a brand new core and there is bound to be lessons learnt and weaknesses identified in the current Zen core. Zen 2 will most likely be fabbed at GF 7LP which is a high performance node designed for AMD and IBM. GF 7LP is very close to Intel 10nm in transistor density and should allow very high frequency CPU design given that the process is designed for 5 Ghz operation. AMD will have its best opportunity with Zen 2 on GF 7LP in 2 decades to compete against Intel.

Thanks for a real answer. It makes me excited for 2019.

 

[crashtech]

Pretty sure we'd get the IPC from the clock speed.

No, IPC is instructions per clock and won't vary with clockspeed, you are probably thinking of something better referred to as throughput.

 

[TahoeDust]

Pretty sure we'd get the IPC from the clock speed.

You can't "get" IPC from increasing clock speed.

 

[epsilon84]

Pinnacle Ridge is has no architecture change AFAIK, it's a process tweak, but yeah, there should be some clock speed boost.

Now that I have decided to upgrade, I won't be waiting for Pinnacle Ridge, Unless AMD surprise announces it's coming soon, which I don't expect.

Future chip products, tend to be late more than early in recent years.

Well you have to pull the trigger at some point right? To be honest I would build a new rig now if not for the exorbitant DDR4 prices. Yes, I can get cheaper low grade DDR4 but I refuse to get DDR4-2400 class RAM when I know DDR4-3200+ is where its at for gaming, in fact Intel can take significant advantage of up to DDR4-4000 which is damn expensive right now! https://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

 

[PeterScott]

Well you have to pull the trigger at some point right? To be honest I would build a new rig now if not for the exorbitant DDR4 prices. Yes, I can get cheaper low grade DDR4 but I refuse to get DDR4-2400 class RAM when I know DDR4-3200+ is where its at for gaming, in fact Intel can take significant advantage of up to DDR4-4000 which is damn expensive right now! https://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

RAM price is really killing me too. Which in a way tilts me toward the i5-8400 on a cheaper MB, which I assume won't let you OC the RAM anyway. Cheaper CPU, Cheaper MB, and Cheaper RAM.

Then there are overpriced GPUs as well. I might try to run on the IGP for a while, until GPU prices come back down to reality.

How does a current iGPU compared to an old NVidia 8800GT?

 

[raghu78]

I'm not going to hold my breath for any of that considering your overly optimistic predictions of the past.

Who is asking you to believe in anything what I say ? Its just speculation and its not some law of nature which I am stating. 2018 will show whether GF is improving execution or not as GF ramps 12LP. So we will know sooner rather than later how GF is doing and how its positioned for the future. btw any body can mess up at the leading edge including the mighty Intel which is struggling with their 10nm process delays. are you still hoping for Icelake in 2018 ?

 

[Campy]

Yes, I can get cheaper low grade DDR4 but I refuse to get DDR4-2400 class RAM when I know DDR4-3200+ is where its at for gaming, in fact Intel can take significant advantage of up to DDR4-4000 which is damn expensive right now! https://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

Keep in mind that they are using 980Ti SLI to achieve those results with 4000mhz memory. There was practically no scaling when using a single 980Ti. I agree 3200Mhz CL14 is a nice spot to be in for gaming.

 

[Paratus]

Jay two cents redid his 8700K vs 1700X review. This time with MCE off and his Ryzen mem actually at 2933 instead of default speeds.

At stock they are pretty close and even similar at OC. (Main exception being gaming )

(There was a bit of a back and forth with AdoredTV in the comments of both videos about MCE)