AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[michty_me]

I'm banking on X570, PCIE 4.0, NVME, high core/decent clocking CPU with 16/32GB RAM lasting many generations of GPU. We'll see if it lasts as well as Sandybridge, I think it will as clock speed is reaching a limit and there are only do many cores you can add and supply power to. Progress will inevitably be slower and systems longer lasting imo.

Which is good for the consumer!
I remember back in the day, Things progressed rather quickly with updates that your system would be obsolete fairly quickly.

 

[smilingcrow]

Gaming consoles should be on 8 cores until 2025 or later so 12 to 16C for the desktop should be good for the majority for 6 years or longer.
This for the high end only as for the rest 8C is enough.

 

[kitfit1]

LOL

My only issue is that this is going to push me to reinstall windows and move to Win 10, keeping the 8.1 install from haswell to coffeelake was a fair bit of work but I got away with it, however going from intel to AMD I think will be a step too far to pull it off.

Sometimes biting the bullet and doing a fresh install is the only way to accept progress and move on. As you know, Ryzen won't run win8.1 natively, so win10 or Linux are the only options. It will almost certainly be the case that the next Intel architecture will need win10 as well.

 

[Perfect_Chaos]

3700x is almost an instant buy if the specs are true. Now where are the leaks to see if the rumours are rubbish or not

 

[Mercutio]

Whatever has the fastest boost clocks for me. If it's that 3850x with 16c/32t, so be it.

 

[CAT-THE-FIFTH]

Threadripper 3 might be delayed:

https://www.tomshardware.com/news/amd-third-gen-threadripper-roadmap,39254.html

 

[Mercutio]

I do wonder (threadripper) if there's perhaps been a case for "is it needed?"
Rather than copying Intel and neatly segmenting their CPU's.... could 40 lanes maybe being considered enough? There's only extra memory channels really making a large difference at that point(?).
Ryzen 3/5/7/9 maybe being considered sufficient to cover everyone?
Maybe my HEDT understanding is weak and I'm missing other key points. I kinda always thought it was more cache, more lanes, more memory channels.

 

[Potatowithearsontheside]

The TR4 platform with PCIe4 might just be overkill, and there's been no reports of any boards having been developed for it. Without X599 boards any new TR would be leaving something on the table.

 

[smilingcrow]

They could just add a 32C Zen 2 chip to the TR platform to give it a faster, more power efficient and hopefully cheaper flagship.
Might be too much effort just for one SKU though.

 

[DaGooN]

Well after 10 years i am willing to upgrade my UD5+x5670 to ryzen 7 or 9.
Been a long time since i used new hardware, so looking forward to what they show at computex.

I still own two Amd platforms that work
1.Abit Fatal1ty AN8 939 + Athlon 64 X2 4200+ inbox
2.Gigabyte GA-MA790FX-DS5 + X6 1065T inbox

 

[Journey]

If they delay/cancel TR3 CPU's it would seriously impact the uptake in the workstation market of AMD products.

If they could make the AM4 support 40 lanes+ of PCI-E 4.0 split down in to 80 PCI-3.0 lanes and ensure that 128GB of ECC RAM works well at 2933+ turn they may be able to swing it, if the boards are as good as are being led to believe.

I've loved developing TR4 based systems over the last two generations, and it opened up so many possibilities that were not there with Xeon, and having 64 PCI-E lanes was a huge part of that, I really hope they don't leave it behind.

 

[ltron]

3700x is almost an instant buy if the specs are true.

Same here.

 

[EsaT]

Overheating with official specs and causing instability?

Or just overheating compared to temps typically seen on boards?

Although I suppose no surprise with asus who seem to be having a habit now of been best on top end boards, but lack lacklustre on mainstream.

Review didn't mention instability, but throttling CPU to 550MHz / 0,750V after 106C VRM temperature.

Gigabyte B450 Aorus Pro also had problems with (identical components) VRM reaching very hot 101C.
Also Precision Boost Overdrive led to problems with even 2600X.

Here's whole review if you feel lucky trying Google Translate:
https://www.io-tech.fi/artikkelit/testissa-amd-b450-emolevyt-asrock-asus-gigabyte-msi/

 

[EsaT]

Surely they test the cores before making the CCXs up? Otherwise I don't see the benefit over one monolithic chip, both could have defective cores.
I imagined they take all the cores, test them, rank them then pick 8 of the best for Epyc, next best for Threadripper and so on. Would they not just use 6 good cores for 6 core cpu rather than 2 bad + 6 good?

EPYC is all about power efficiency and not about highest clocks.
Those will use computing dies with the lowest power consumption at moderate clocks.
Lower core count models could also use dies with broken cores.

Threadripper is certainly going to need high clocking dies.
But that's coming late in this year anyway and hence doesn't need diverting all high clocking dies away from Ryzen.

 

[chrcoluk]

If there will be driver support issues with Windows 10 LTSC 1809, then I'd skip the 3000 series. Running the consumer version of Windows 10 is not a viable option for me.

yeah I am looking into LTSB myself also, it seems tho getting a legal license as a consumer is not a thing? Microsoft seem to consider LTSB a enterprise only feature.

To me doing feature updates every 6-18 months is not viable.

 

[chrcoluk]

EPYC is all about power efficiency and not about highest clocks.
Those will use computing dies with the lowest power consumption at moderate clocks.
Lower core count models could also use dies with broken cores.

Threadripper is certainly going to need high clocking dies.
But that's coming late in this year anyway and hence doesn't need diverting all high clocking dies away from Ryzen.

This is what is frustrating me with AMD, that they tieing in highest clocks with highest cores, there is not a model with highest clocks and low core count.

 

[EsaT]

This is what is frustrating me with AMD, that they tieing in highest clocks with highest cores, there is not a model with highest clocks and low core count.

At least in current GloFo made Ryzens those high clocks needed some serious volts meaning, which might not fit inside TDPs of lower models.

 

[chrcoluk]

Review didn't mention instability, but throttling CPU to 550MHz / 0,750V after 106C VRM temperature.

Gigabyte B450 Aorus Pro also had problems with (identical components) VRM reaching very hot 101C.
Also Precision Boost Overdrive led to problems with even 2600X.

Here's whole review if you feel lucky trying Google Translate:
https://www.io-tech.fi/artikkelit/testissa-amd-b450-emolevyt-asrock-asus-gigabyte-msi/

Its a long wall of text and translated I will admit guilty of lazyness, anything from a english reviewer with more directness of information?

From what I can gather the test was a full load of prime95, which isnt realistic in terms of the use case of the boards, I didnt see anything indicating if TDP limits were been honoured or overidden in settings. I find it hard to believe a 95 watt load was making a cpu throttle down to 550mhz. IF its true then why this isnt all out in the open is shameful from reviewers, but I do find it hard to believe at the moment, usually VRM issues are related to overclocking that breaches spec of the cpu.

 

[EsaT]

Its a long wall of text and translated I will admit guilty of lazyness, anything from a english reviewer with more directness of information?

From what I can gather the test was a full load of prime95, which isnt realistic in terms of the use case of the boards, I didnt see anything indicating if TDP limits were been honoured or overidden in settings. I find it hard to believe a 95 watt load was making a cpu throttle down to 550mhz. IF its true then why this isnt all out in the open is shameful from reviewers, but I do find it hard to believe at the moment, usually VRM issues are related to overclocking that breaches spec of the cpu.

Primary method for testing VRM temps etc was Prime, but already X264 encoding with 2700X overheated Strix B450-F's VRM after 12 minutes.
And even if overheating and throttling isn't triggered that Asus and Gigabyte VRMs run such hot that endurance would be questionable with 95W TDP CPU.

And lots of reviewing is pretty much disguised advertising.
I'm sure you've often noticed in reviews of most things how reviewers just hype RGB bling bling and other fashion cruds.

 

[StriderX]

Am I right to assume that 5GHz is basically the thermal limit (within reason)? Where exactly does intel intend to go if this is the case?