Question The new threadrippers 7980X and 7970X

[iamgenius]

Quote from Steve from GN


LOLOL....

However what if primary is PCI-E lanes and secondary overclocking, and third Gaming?

The only thing holding me back from getting one right now, is probably the board selection, as there is no real board that i like at the moment.

But paying a bit more for marginal gaming performance and having double the PCI-E Gen5 Lanes along with 8 channel DDR5 support, also makes it hard to ignore the W5-3435X, unless i really wanted those more efficient cores, which if you look down my list, doesn't show up until you probably go to like point 5 or 6.

So i am still in Limbo on what my next HEDT will be.


W5-3435X with 8 channel DDR5 + 128 PCI-E Gen5 goes for around $1500 if you can find them at non scalper prices.

https://www.newegg.com/intel-xeon-w5-3435x-lga-4677/p/N82E16819118452

I think this is the sweet spot for Intel HEDT competition with Thread Ripper.
Not the 2400 line where its sort of gimped compared to the 3400 series.
The 2400 series has no real chance against the 7000 series.
Its a no brainer IMO, unless you absolutely really need those PCI-E lanes, which you don't even have as many compared to the 3400 series.
But if your that much in a budget / debt, then you should be getting last gen's EYPC IMO, and not a HEDT period.



Are you in the US or EU?
Because in the US, its 240V, for a L1+L2 combine.
And yes your math is correct. You get double the wattage at half the amperage with double the voltage.

I really don't recommend you doing a 240V dedicated for PC's unless your ASIC farming, or its next to a garage where you have your EV charger port at.

Wiring codes, and permits are a PITFA, unless you do it illegally, and 240V is not really something you want to mess with unless your a liscensed electrictian.

But if your EU, where nothing is 120v, then yea, im all in for 240V PSU's as they use half the amperage, and also more efficient because of it.



Note im hearing from sources, that if you overclock a threadripper, you fuse a register, which shows you overclocked it, and void all forms of warrenty on the CPU.

This does not sit well with me, which is why im also waiting, to see if these guys will melt a hole though the motherboard like the 7000X3D series did.

9 New Things About Overclocking Ryzen Threadripper 7000 - SkatterBencher

Here are 9 things you need to know before overclocking the Ryzen Threadripper 7000 "Storm Peak" processors.

skatterbencher.com

1- What is it that is missing from current Threadrippers 7000 series boards that you are looking for ?

2- I'm in the middle east, UAE specifically where cop28 is taking place now. Our wall outlets are 220v and 13 amps. I don't understand why you don't recommend dedicating once circuit for one PC. I mean you can more devices using the same outlet, but why would it be bad if only the PC is running using this specific circuit? A PC with an overclocked threadripper will almost saturate, so better be safe than sorry. Did I miss something?

3- What? If my cpu goes bad they will not cover it under the warranty if it has been overclocked even for 1 second? That's really really bad. I mean this can be a deal breaker. Are you waiting on some workaround against this?

 

[StefanR5R]

Just to be sure, so sometimes you will have to connect both 24-pin connectors to the motherboard, and sometimes not depending on the motherboard manual and what it says. If you are going to get a motherboard that requires the two 24-pin connectors to be connected, then you will either have to use two power supplies or use one power supply that comes with two 24-pin connectors, right?

I mean, it depends on the motherboard how it provides power to the threadripper being overclocked.

The info which the two vendors give for the two only(?) announced boards is rather thin.

ASRock TRX50 WS
product page - https://www.asrock.com/MB/AMD/TRX50 WS/
manual - https://download.asrock.com/Manual/TRX50 WS.pdf

Asus Pro WS TRX50-Sage WIFI
product page - https://www.asus.com/motherboards-components/motherboards/workstation/pro-ws-trx50-sage-wifi/
manual - https://dlcdnets.asus.com/pub/ASUS/..._WIFI_EM_WEB.pdf?model=Pro WS TRX50-SAGE WIFI
quick guide - https://gzhls.at/blob/ldb/7/c/9/d/5f1dfd915c5b3c2561d01fda874084028ad3.pdf

My understanding:

I suspect that the second 24-pin ATX12V connector on the Asus board fulfills no other function than the one which the little 3-pin header on the ASRock board (to which an adapter cable to 24-pin ATX12V would be attached) fulfills: All it does is to distribute the power-on signal to a secondary PSU when you power on the primary PSU via the mainboard's power-on signal. In theory, the two +12V pins of Asus' secondary 24-pin ATX12V connector could be used for extra power supply, but I very much doubt that because their contribution would be minimal.

Rather, it's the various 8-pin EPS12V and 8- or 6-pin PCIE connectors through which the big Wattage goes. It is mandatory to connect at least the first two 8-pin EPS12V connectors, the others are optional. But if you don't populate any of the optional power inputs, you shouldn't overclock too much. Just the primary two 8-pin EPS12V connectors could easily be sufficient for at least 600 W continuous power to the CPU socket, going by the SSI EPS12V specification. But how much can go through these two connectors on the particular mainboards also depends on how the boards are built.

If you want to supply more than these circa 600 W to the CPU, populate some or all of the optional power inputs (8-, 6-pin PCIe, extra 8-pin EPS12V).

Unfortunately the linked manuals do not provide any tables which tell how much power can safely be drawn with which set of connectors.

However, it is totally immaterial if you attach these various power inputs to a single PSU, or to two PSUs. (The only condition is that the primary two EPS12V connectors need to be supplied from a single rail, according to the Asus manual.) Now, most of the available PSUs give you only two EPS12V cables. There are some PSUs which have three EPS12V cables. (These were designed for workstations with a professional GPU with EPS12V power input instead of the more common 6- and 8-pin PCIE power inputs.) I don't know of any PSUs with four EPS12V cables. But you may be able to get additional modular cables from the customer support of the PSU maker, or may find compatible 3rd party cables from certain case modding brands. --- The number of 8- and 6-pin PCIE cables on the other hand shouldn't be an issue, especially on bigger PSUs... until you run out of these if you also plan to put one or two or n big GPUs into the system.

That is, at some point you might run out of cable connections from your PSU, and that's when you need to look for additional cables (if there are still spare modular ports on the PSU) or need to add a second PSU.

Also note, one 6-pin PCIE connector on the ASRock board seems to power the PCIe slots, not the CPU. It is not clear from Asus' manual whether or not any of the PCIE connectors on their board is likewise dedicated to PCIe slots instead of the CPU. Furthermore, Asus' manual advises that, if 2 PSUs are used, they should be the same make and model.

Anyhow.
The need of a second PSU should only arise with LN2 overclocking, overclocking of the 64- and 96-core Threadrippers, or if you need a lot of power for dual/ triple/ quadruple GPUs.

 

[StefanR5R]

I'm in the middle east, UAE specifically

Aren't you having high humidity over there throughout most of the year? In which case it may be wise to stay below the Amperage rating of a given 220 V circuit, depending on how long ago the wiring in your building was installed. (Humidity could promote contact corrosion in wall outlets, junction boxes and such. Corroded clamps/ wire terminals etc. with respective undesired resistance could get hot when highly loaded.)

What? If my cpu goes bad they will not cover it under the warranty if it has been overclocked even for 1 second?

"Improper use" has always been excluded from warranty. The only change with Threadripper 7000, *if* what the linked article claims is true, would be that AMD would have an easier time to check a returned CPU for whether or not it had been overclocked. However, that a damaged CPU had been overclocked doesn't necessarily mean it got its damage from the overclock. Ask AMD's customer support or/and your lawyer... :-)

 

[aigomorla]

1- What is it that is missing from current Threadrippers 7000 series boards that you are looking for ?

2- I'm in the middle east, UAE specifically where cop28 is taking place now. Our wall outlets are 220v and 13 amps. I don't understand why you don't recommend dedicating once circuit for one PC. I mean you can more devices using the same outlet, but why would it be bad if only the PC is running using this specific circuit? A PC with an overclocked threadripper will almost saturate, so better be safe than sorry. Did I miss something?

3- What? If my cpu goes bad they will not cover it under the warranty if it has been overclocked even for 1 second? That's really really bad. I mean this can be a deal breaker. Are you waiting on some workaround against this?

The plentiful physical PCI-E 16x lanes that can even utilize that many pci-e lanes.
So far no board ive seen has that many physical slot pci-e lanes.

My comment only applies if your US.
If your power grid is on 220V/240V, then its probably better for you to get a 240V PSU as they run half amperage and hence run more efficient.

Im worried about that last comment.
Intel also said they would void warrenty from overclocking, but never rejected any RMA's.
Even on the Wolfdale, and Yorkfield which many many people killed from excessive voltage, because they thought it would be like a kentsfield and be impossible to kill.

But intel never went as far as to fuse a register.
This sort of really doesn't sit well with me, because AMD has issues with EXPO still on the 7000X3D series, although not as bad as when they first came out.

That means if you overclock fuse a register, and run into a EXPO accident, even if its not your fault, both the motherboard vendor and the cpu vendor now have grounds to say you caused it, and were not liable for it.

 

[StefanR5R]

Facts:
1. AMD states that the TRX50 and WRX90 platforms have overclocking support. (source)
2. Damage from "improper use" ("misuse, neglect, improper installation or testing") is not, and never has been, covered by warranty. (source)

If your Threadripper is damaged by one of the cases which are covered by its limited warranty, AMD is liable to fulfill that warranty.

It's obvious that this matter is more important to both the customer and the maker if it's about a $5000 product, rather than a $250 product.

 

[iamgenius]

The info which the two vendors give for the two only(?) announced boards is rather thin.

ASRock TRX50 WS
product page - https://www.asrock.com/MB/AMD/TRX50 WS/
manual - https://download.asrock.com/Manual/TRX50 WS.pdf

Asus Pro WS TRX50-Sage WIFI
product page - https://www.asus.com/motherboards-components/motherboards/workstation/pro-ws-trx50-sage-wifi/
manual - https://dlcdnets.asus.com/pub/ASUS/..._WIFI_EM_WEB.pdf?model=Pro WS TRX50-SAGE WIFI
quick guide - https://gzhls.at/blob/ldb/7/c/9/d/5f1dfd915c5b3c2561d01fda874084028ad3.pdf

My understanding:

I suspect that the second 24-pin ATX12V connector on the Asus board fulfills no other function than the one which the little 3-pin header on the ASRock board (to which an adapter cable to 24-pin ATX12V would be attached) fulfills: All it does is to distribute the power-on signal to a secondary PSU when you power on the primary PSU via the mainboard's power-on signal. In theory, the two +12V pins of Asus' secondary 24-pin ATX12V connector could be used for extra power supply, but I very much doubt that because their contribution would be minimal.

Rather, it's the various 8-pin EPS12V and 8- or 6-pin PCIE connectors through which the big Wattage goes. It is mandatory to connect at least the first two 8-pin EPS12V connectors, the others are optional. But if you don't populate any of the optional power inputs, you shouldn't overclock too much. Just the primary two 8-pin EPS12V connectors could easily be sufficient for at least 600 W continuous power to the CPU socket, going by the SSI EPS12V specification. But how much can go through these two connectors on the particular mainboards also depends on how the boards are built.

If you want to supply more than these circa 600 W to the CPU, populate some or all of the optional power inputs (8-, 6-pin PCIe, extra 8-pin EPS12V).

Unfortunately the linked manuals do not provide any tables which tell how much power can safely be drawn with which set of connectors.

However, it is totally immaterial if you attach these various power inputs to a single PSU, or to two PSUs. (The only condition is that the primary two EPS12V connectors need to be supplied from a single rail, according to the Asus manual.) Now, most of the available PSUs give you only two EPS12V cables. There are some PSUs which have three EPS12V cables. (These were designed for workstations with a professional GPU with EPS12V power input instead of the more common 6- and 8-pin PCIE power inputs.) I don't know of any PSUs with four EPS12V cables. But you may be able to get additional modular cables from the customer support of the PSU maker, or may find compatible 3rd party cables from certain case modding brands. --- The number of 8- and 6-pin PCIE cables on the other hand shouldn't be an issue, especially on bigger PSUs... until you run out of these if you also plan to put one or two or n big GPUs into the system.

That is, at some point you might run out of cable connections from your PSU, and that's when you need to look for additional cables (if there are still spare modular ports on the PSU) or need to add a second PSU.

Also note, one 6-pin PCIE connector on the ASRock board seems to power the PCIe slots, not the CPU. It is not clear from Asus' manual whether or not any of the PCIE connectors on their board is likewise dedicated to PCIe slots instead of the CPU. Furthermore, Asus' manual advises that, if 2 PSUs are used, they should be the same make and model.

Anyhow.
The need of a second PSU should only arise with LN2 overclocking, overclocking of the 64- and 96-core Threadrippers, or if you need a lot of power for dual/ triple/ quadruple GPUs.

It is a bit complicated I see. I have no plans to use more than one GPU, so that's one thing. And no LN2 of course because I want a usable overclock, but I still think a large PSU is in order. I think I will wait and see if more things clear up later on. And you can always buy more PSU cables.

 

[iamgenius]

Aren't you having high humidity over there throughout most of the year? In which case it may be wise to stay below the Amperage rating of a given 220 V circuit, depending on how long ago the wiring in your building was installed. (Humidity could promote contact corrosion in wall outlets, junction boxes and such. Corroded clamps/ wire terminals etc. with respective undesired resistance could get hot when highly loaded.)


"Improper use" has always been excluded from warranty. The only change with Threadripper 7000, *if* what the linked article claims is true, would be that AMD would have an easier time to check a returned CPU for whether or not it had been overclocked. However, that a damaged CPU had been overclocked doesn't necessarily mean it got its damage from the overclock. Ask AMD's customer support or/and your lawyer... :-)

Yes humidity is a problem here. But the house is relatively new. We will need to wait to see if this overclocking thingie is true or not and whether a workaround can be made to it. Thanks.

 

[iamgenius]

The plentiful physical PCI-E 16x lanes that can even utilize that many pci-e lanes.
So far no board ive seen has that many physical slot pci-e lanes.

My comment only applies if your US.
If your power grid is on 220V/240V, then its probably better for you to get a 240V PSU as they run half amperage and hence run more efficient.

Im worried about that last comment.
Intel also said they would void warrenty from overclocking, but never rejected any RMA's.
Even on the Wolfdale, and Yorkfield which many many people killed from excessive voltage, because they thought it would be like a kentsfield and be impossible to kill.

But intel never went as far as to fuse a register.
This sort of really doesn't sit well with me, because AMD has issues with EXPO still on the 7000X3D series, although not as bad as when they first came out.

That means if you overclock fuse a register, and run into a EXPO accident, even if its not your fault, both the motherboard vendor and the cpu vendor now have grounds to say you caused it, and were not liable for it.

Thankfully I don't need many lanes. I remember you aigomorla use so many SSD's for gaming!

Facts:
1. AMD states that the TRX50 and WRX90 platforms have overclocking support. (source)
2. Damage from "improper use" ("misuse, neglect, improper installation or testing") is not, and never has been, covered by warranty. (source)

If your Threadripper is damaged by one of the cases which are covered by its limited warranty, AMD is liable to fulfill that warranty.

It's obvious that this matter is more important to both the customer and the maker if it's about a $5000 product, rather than a $250 product.

I just hope it is not an issue. I mean if you allow overclocking and advertise that your product is overclockable and be proud about it, but then make customers lose their piece of mind if they use it. I think that's a bit unethical.

 

[aigomorla]

Thankfully I don't need many lanes.

The 7950X and the 14900K are both significantly faster in everything unless you need that many cores.
Unless you need the lanes, Threadripper / W-Series Xeon processors should not even be on your list.

Also I hear Meteor Lake will drastically change how we think of modern processors, so it may be in your best interest to just wait until Metor Lake arrives and see if the claims are even half correct from what im hearing.

 

[Markfw]

The 7950X and the 14900K are both significantly faster in everything unless you need that many cores.
Unless you need the lanes, Threadripper / W-Series Xeon processors should not even be on your list.

Also I hear Meteor Lake will drastically change how we think of modern processors, so it may be in your best interest to just wait until Metor Lake arrives and see if the claims are even half correct from what im hearing.

I would agree, however, Zen 5 may also be a groundbreaker. When is Meteor lake desktop supposed to launch ? as in reviews/for sale. ETA for Zen 5 I hear is early April.

 

[iamgenius]

The 7950X and the 14900K are both significantly faster in everything unless you need that many cores.
Unless you need the lanes, Threadripper / W-Series Xeon processors should not even be on your list.

Also I hear Meteor Lake will drastically change how we think of modern processors, so it may be in your best interest to just wait until Metor Lake arrives and see if the claims are even half correct from what im hearing.

Lanes are to connect devices. I don't want to connect so many devices but I want the cores for heavy multitasking and running so many VMs at the same time. Why do you say the 7950X is faster in everything? It is for gaming mainly.

Meteor lake? I didn't read about it. What did you hear? Is it gonna be like when Intel blew AMD away years ago with their core cpus? And when are they launching? I'll try to read some articles.

 

[iamgenius]

I would agree, however, Zen 5 may also be a groundbreaker. When is Meteor lake desktop supposed to launch ? as in reviews/for sale. ETA for Zen 5 I hear is early April.

WOW, also Zen 5. I cant wait until April. But, in the end it will just be like somebody said: If you always wait for new technology, you will never buy anything. So buy while you are still young and can operate a PC before old age hits you.

 

[Thibsie]

I just hope it is not an issue. I mean if you allow overclocking and advertise that your product is overclockable and be proud about it, but then make customers lose their piece of mind if they use it. I think that's a bit unethical.

This is usual, every brands do that.
Besides, I don't see how (till now at least) any maker would be able to detect that the CPU you just RMA'ed id dead because of overclocking so this is mostly a non-issue.

Of course if there are traces of LN2....

 

[iamgenius]

This is usual, every brands do that.
Besides, I don't see how (till now at least) any maker would be able to detect that the CPU you just RMA'ed id dead because of overclocking so this is mostly a non-issue.

Of course if there are traces of LN2....

There is no way for them to know if a cpu went bad because of overclocking but the way it was written indicates that they will know it was overclocked and since it was sent back to them because it is bad, they will assume it is because of overclocking since it was overclocked evident by the marker.

 

[aigomorla]

Why do you say the 7950X is faster in everything?

Core per core 7950X is faster then Threadripper.
The advantage in threadripper is that you have way more then 16 cores, as Threadripper starts at 16.
But the 7950X is faster core to core.

Do you need more then 16X cores, unless you do Threadripper makes no sense if you do not need the pci-e lanes.

This is usual, every brands do that.
Besides, I don't see how (till now at least) any maker would be able to detect that the CPU you just RMA'ed id dead because of overclocking so this is mostly a non-issue.

Of course if there are traces of LN2....

Did you not read the article i linked?
Overclocking Threadripper causes it to Fuse a Register, meaning they will know if you overclocked it or not, as that register will be fused.
Its sort of like how when you jail break a samsung phone, it does something to hardware to notify samsung the phone was jail broken, and they refuse to honor any warrenty on the phone even if its not your fault.

 

[StefanR5R]

But the 7950X is faster core to core.

Until you raise Threadripper's power limit, I guess.

Edit,

Also I hear Meteor Lake will drastically change how we think of modern processors,

Who is we? From what I know about Meteor Lake (and unlike AMD, Intel has already been very talkative about planned CPUs), you can definitely count me out from that 'we', for one.

[Edit 2, rationale: If I was interested in mobile computers, which I am not for the time being, I wouldn't loose sleep over how the CPU is built in particular. The biggest power consumer most relevant battery life consumer in mobile computers, at least those with which you interact via a display, is the display. Apropos, laptop chips are a bit off topic in a HEDT & WS thread. Sorry that I went off track here.]

 

[Markfw]

Until you raise Threadripper's power limit, I guess.

I sounds to me like the OP needs more than 16 cores, but maybe 64 is overkill for him, but with the VM's, more memory and bandwidth, along with maybe 32 cores and more 1T core speed, and the ability for a mild OC, TR might be the answer.

 

[iamgenius]

Core per core 7950X is faster then Threadripper.
The advantage in threadripper is that you have way more then 16 cores, as Threadripper starts at 16.
But the 7950X is faster core to core.

Do you need more then 16X cores, unless you do Threadripper makes no sense if you do not need the pci-e lanes.



Did you not read the article i linked?
Overclocking Threadripper causes it to Fuse a Register, meaning they will know if you overclocked it or not, as that register will be fused.
Its sort of like how when you jail break a samsung phone, it does something to hardware to notify samsung the phone was jail broken, and they refuse to honor any warrenty on the phone even if its not your fault.

Yeah...I definitely need more than 16 cores. Will desktop mainstream have 32 cores in the next gen? I don't know, and I want all performance cores. I dislike intel idea of mixing p cores and e cores.

 

[iamgenius]

Thankfully I don't need many lanes. I remember you aigomorla use so many SSD's for gaming!

I just hope it is not an issue. I mean if you allow overclocking and advertise that your product is overclockable and be proud about it, but then make customers lose their piece of mind if they use it. I think that's a bit unethical.

It looks like it is not an issue and overclocking will not void the warranty:

AMD Ryzen Threadripper 7970X (Pre-order & Price & Specs & Release Date) - CPU-Rumors

AMD’s new Ryzen Threadripper 7970X represents the pinnacle of desktop CPU performance. With 32 high-powered “Zen 4” cores, speeds up to 5.3GHz, and support for fast DDR5 memory, this flagship Threadripper chip delivers elite multi-core and single-core speed for enthusiasts. Unlocked for...

www.cpu-rumor.com

I don't know how reliable this source though.

 

[iamgenius]

More about the fuse:

Overclocking Your Ryzen Threadripper 7000 Blows a Hidden Fuse, Says AMD

The blown fuse won't hurt your CPU, but it proves you've enabled overclocking.

www.howtogeek.com

I think it is safe to say that overclocking will not void the warranty.

I have been reading a little about overclocking threadrippers. Manual overclocking can get very complicated and you lose some single-core performance because you didn't use PBO. I don't like this. Manual overclocking should be the ultimate way to overclock with the most options to get optimum performance. I don't understand why should we lose single-core performance if we discard PBO and opt for manual. I mean in manual we should be able to tweak things for single-core performance to match its performance in PBO. Is this really a thing??? I want the best of the two worlds

 

[aigomorla]

I think it is safe to say that overclocking will not void the warranty.

Overclocking will void the warrenty.
Its written in the TOS, and also they now have a way to prove you overclocked.
Bascially you will get caught with your hand in the cookie jar.

Its just a matter of how AMD decides to enforce it or not, or how hard they intend to enforce it.
Its like how Intel also said they would void warrenty if overclocked, but they didn't enforce it strictly.
However intel also had no real way to prove you overclocked.

Threadripper now has a way to prove you overclocked, and have grounds to void warrenty if they feel like it, as its written in the Terms of Service booklet you get with the CPU.

 

[StefanR5R]

its written in the Terms of Service booklet you get with the CPU.

What is the exact wording there?
Besides, law still applies.

 

[aigomorla]

What is the exact wording there?
Besides, law still applies.

This limited warranty does not cover damages due to external causes, including improper use, problems with electrical power, accident, neglect, alteration, repair, improper installation, or improper testing.

I dont see why they would put in an extra register that fuses when overclocked unless they want to prove something.

I would like an offical announcement from AMD, in regards to this, but i highly doubt we will see any in lue of blacklash.
The only real thing to do is either wait for someone to attempt, or see if any cpu's are dying from overclocking.

If we hear nothing in regards to both, then i guess its all thumbs up.

 

[StefanR5R]

Thanks. That is, overclocking doesn't void the warranty, but may make customers' claims more easily contested by AMD. But keep in mind that AMD advertises overclocking support by Threadripper.

We don't know for sure how the fuse works. IMO it would be appropriate if it triggered when the user increases Voltages over specification, with a hysteresis.

 

[iamgenius]

Myself I don't really understand how you can support and advertise something that will void the warranty.