Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
repiv
Aug 13, 2009

Intel is shook: https://www.techpowerup.com/256700/intel-to-cut-prices-of-its-desktop-processors-by-15-in-response-to-ryzen-3000

Adbot
ADBOT LOVES YOU

Deuce
Jun 18, 2004
Mile High Club

ItBurns posted:

I feel the same way but I'm convincing myself that they'll also be out of stock for months as a consolation.

I'm about to put a down payment on a house so the temptation will be gone along with four digits off my bank account balance. Hooray?

iospace
Jan 19, 2038



Yeah. A non-goon friend pretty much pointed out that Dr. Su was hoping that the 3000 series would compete with Intel.

Turns out she was wrong, but in a good way :v:

GRINDCORE MEGGIDO
Feb 28, 1985


Oo, a whole 15%.

ufarn
May 30, 2009
And that's on top of eating the tariffs I assume.

wargames
Mar 16, 2008

official yospos cat censor

ufarn posted:

And that's on top of eating the tariffs I assume.

Intel is mostly made outside of china so tariff don't hit hard.

Malcolm XML
Aug 8, 2009

I always knew it would end like this.
Lmao. At least Bob Swan knows imte is in trouble in this segment

iospace
Jan 19, 2038


wargames posted:

Intel is mostly made outside of china so tariff don't hit hard.

Yeah, my 3570k was made in Costa Rica so.

Harik
Sep 9, 2001

From the hard streets of Moscow
First dog to touch the stars


Plaster Town Cop

iospace posted:

Epyc pricing leaked;

Here's the intel lineup for comparison:

Maybe AMD needs to undercut by this much to break through the institutional inertia and illegal OEM kickbacks (HP still won't offer AMD servers, only allow you to buy them from "partners") but it sure seems like they're leaving a lot of profit on the table. Depending on how IPC and TDP work out, the comparable epyc to the top-tier xeon platinum is the 32-core, a quarter the price, and that's only in the middle of the epyc range.

The only absolute lead intel has is quad/octo SMP. There's just no way to put 256 or 512 zen cores in a single machine, yet.

E: possibly cache latency as well. Intel has a decent lead in cache-resident computing, so it depends on how well zen2 closes that gap.

Harik fucked around with this message at 17:01 on Jun 21, 2019

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!

iospace posted:

Yeah. A non-goon friend pretty much pointed out that Dr. Su was hoping that the 3000 series would compete with Intel.

Turns out she was wrong, but in a good way :v:

Yeah they expected Ryzen 3000 to be released competing against Ice Lake desktop.

Number19
May 14, 2003

HOCKEY OWNS
FUCK YEAH


AMD has a manufacturing advantage over Intel where they (reportedly) get good yields from their silicon due to their chiplet design. Intel’s yields have been by all accounts pretty bad and they are having a tough time meeting demand

If AMD has a product that has superior specs/performance, is cheaper, AND has abundant supply then they are going to sell so many of them that they’ll make up that margin difference with no issue and gain a major foothold in the server market

It’s a pretty amazing turnaround really

iospace
Jan 19, 2038


Harik posted:

Here's the intel lineup for comparison:

Maybe AMD needs to undercut by this much to break through the institutional inertia and illegal OEM kickbacks (HP still won't offer AMD servers, only allow you to buy them from "partners") but it sure seems like they're leaving a lot of profit on the table. Depending on how IPC and TDP work out, the comparable epyc to the top-tier xeon platinum is the 32-core, a quarter the price, and that's only in the middle of the epyc range.

The only absolute lead intel has is quad/octo SMP. There's just no way to put 256 or 512 zen cores in a single machine, yet.

E: possibly cache latency as well. Intel has a decent lead in cache-resident computing, so it depends on how well zen2 closes that gap.

Slight caveat here, that doesn't include the 2nd gen Scalable Xeons (read: 9000 series). They are available to OEMs at the moment, so pricing info isn't well known, but they do have this: https://ark.intel.com/content/www/us/en/ark/products/194146/intel-xeon-platinum-9282-processor-77m-cache-2-60-ghz.html which is 56/112.

But with 400W TDP. So yeah.

iospace fucked around with this message at 17:19 on Jun 21, 2019

iospace
Jan 19, 2038


Number19 posted:

AMD has a manufacturing advantage over Intel where they (reportedly) get good yields from their silicon due to their chiplet design. Intel’s yields have been by all accounts pretty bad and they are having a tough time meeting demand

If AMD has a product that has superior specs/performance, is cheaper, AND has abundant supply then they are going to sell so many of them that they’ll make up that margin difference with no issue and gain a major foothold in the server market

It’s a pretty amazing turnaround really

Yeah I think AMD's strategy is to try to gain a foothold in the "build your own" consumer market (which judging by steam metrics is a slow, but steady, march), stick their foot in the door for the OEM one, and push HARD on Server side stuff where their strengths come out the most.

Gwaihir
Dec 8, 2009
Hair Elf

iospace posted:

Slight caveat here, that doesn't include the 2nd gen Scalable Xeons (read: 9000 series). They are available to OEMs at the moment, so pricing info isn't well known, but they do have this: https://ark.intel.com/content/www/us/en/ark/products/194146/intel-xeon-platinum-9282-processor-77m-cache-2-60-ghz.html which is 56/112.

But with 400W TDP. So yeah.

Those chips are so stupid they almost don't bear talking about, heh.

iospace
Jan 19, 2038


Gwaihir posted:

Those chips are so stupid they almost don't bear talking about, heh.

How so?

I mean when you're comparing AMD's best offering, you might as well include Intel's best. Yes the companies who will need a 64/128 or 56/112 in a server application* is probably countable on your hands, but it's still interesting to compare.

*Google, Salesforce, Microsoft, Amazon, Oracle, SAP... and that's really all I can think of of off the top of my head.

iospace fucked around with this message at 17:25 on Jun 21, 2019

Broose
Oct 28, 2007

MaxxBot posted:

Yeah they expected Ryzen 3000 to be released competing against Ice Lake desktop.

Question for anyone who can hazard a guess: How would Zen 2 compare to to a theoretical ice lake competitor? I get the suggestion that if Intel didn't flub 10nm so bad that, even with zen, the gap would have stayed the same as it always had at least on desktop.

iospace
Jan 19, 2038


Broose posted:

Question for anyone who can hazard a guess: How would Zen 2 compare to to a theoretical ice lake competitor? I get the suggestion that if Intel didn't flub 10nm so bad that, even with zen, the gap would have stayed the same as it always had at least on desktop.

I think AMD was shooting for roughly equal single core performance* with superior multi-thread performance. More and more programs are taking advantage of threads, so might as well go for it. I think worst case scenario for AMD (and best for Intel) at this point is the 3rd party benches for Zen 2 come up exactly as that.

Given that the build-your-own market tends to be people who are PC gamers, multi-tasking is getting more and more relevant, and with higher core and thread counts, it allows AMD an advantage there as well. That's not getting into engines that can take advantage of MT as well.

*+/- 2% on benchmarks for SC tasks, in my opinion. Or, close enough that "eh, if you're buying on single core performance, buy the cheapest option"

iospace fucked around with this message at 17:41 on Jun 21, 2019

K8.0
Feb 26, 2004

Her Majesty's 56th Regiment of Foot
Yes and no. Architectural improvements also get harder over time, so it will always be easier to kinda catch up than to grow a lead. Intel was dominant because they had the best architecture and the best process. Now they're presumably still ahead architecture wise, but process wise they are getting killed since their 7nm is nowhere to be seen and other people are now producing high performance 7nm in volume.

If Intel maintained their process lead, odds are AMD would still be getting smoked, but the gap would trend smaller and smaller because of the difficulty of innovating increasing over time.

Gwaihir
Dec 8, 2009
Hair Elf

iospace posted:

How so?

I mean when you're comparing AMD's best offering, you might as well include Intel's best. Yes the companies who will need a 64/128 or 56/112 in a server application* is probably countable on your hands, but it's still interesting to compare.

*Google, Salesforce, Microsoft, Amazon, Oracle, SAP... and that's really all I can think of of off the top of my head.

They just don't make remote sense from a power/performance combo really. Also likely just straight overboard for cooling in a lot of places if you actually tried to fill a rack with them. It's more like "Hey these exist so we can say they exist, how many will we sell? Ehhhhhhh... don't think too hard about that part."

iospace
Jan 19, 2038


K8.0 posted:

Yes and no. Architectural improvements also get harder over time, so it will always be easier to kinda catch up than to grow a lead. Intel was dominant because they had the best architecture and the best process. Now they're presumably still ahead architecture wise, but process wise they are getting killed since their 7nm is nowhere to be seen and other people are now producing high performance 7nm in volume.

If Intel maintained their process lead, odds are AMD would still be getting smoked.

Eh, I'm going to slightly disagree on the arch bit. They have a superior single-core arch, but when multi-threading comes into play, it's not as good as AMD's. So is it a good arch in one way? Yeah, it is. Is it a bad arch in another? Also true.

That's also not involving the side-channel exploits either.

iospace
Jan 19, 2038


Gwaihir posted:

They just don't make remote sense from a power/performance combo really. Also likely just straight overboard for cooling in a lot of places if you actually tried to fill a rack with them. It's more like "Hey these exist so we can say they exist, how many will we sell? Ehhhhhhh... don't think too hard about that part."

True, but there is an (Extremely small) market for them :v: (though thinking about it more it may be more for supercomputer poo poo than server poo poo)

I am curious for the low/mid-end Xeon 9000 and Epyc Rome comparisons, though.

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!

iospace posted:

How so?

I mean when you're comparing AMD's best offering, you might as well include Intel's best. Yes the companies who will need a 64/128 or 56/112 in a server application* is probably countable on your hands, but it's still interesting to compare.

*Google, Salesforce, Microsoft, Amazon, Oracle, SAP... and that's really all I can think of of off the top of my head.

This is what ServeTheHome has said about Cascade Lake-AP.

https://www.servethehome.com/intel-xeon-platinum-9200-series-lacks-mainstream-support/

quote:

To date, no major vendor such as Dell EMC, HPE, Inspur, Lenovo, QCT, Huawei, or Supermicro has announced mainstream support for the Intel Xeon Platinum 9200 series. Other vendors like Wiwynn, Gigabyte, and Tyan also have told us they are not planning to support the Platinum 9200 series. In April 2019, I asked eight of those ten vendors why they do not support the Intel Xeon Platinum 9200 series and have gotten shockingly similar responses. HPE will support the Platinum 9200 when the Cray acquisition closes, but Cray systems are not mainstream ProLiant. Server customers and Intel rely upon a broad partner ecosystem and the partner ecosystem has largely rejected the Intel Xeon Platinum 9200 series since its launch. In this article, we are going to talk about why.
https://www.servethehome.com/amd-epyc-rome-namd-intel-xeon-computex-2019/

quote:

The first benchmark Intel offered was of the Intel Xeon Platinum 9200 series. Let us be clear, a 64 core AMD EPYC Rome part we expect to be in the 200-300W range, not at 400W. AMD will have broad ecosystem support. We saw several dozen Rome platforms at Computex 2019 alone. Conversely, we explained why the Intel Xeon Platinum 9200 series lacks mainstream support. For our readers, we suggest disregarding Intel Xeon Platinum 9200 benchmarks wholesale for competitive analysis. Virtually the only time they are better than the Intel Xeon Platinum 8200 is if you need extreme density where you can have 4x 400W CPUs per 1U.

iospace posted:

True, but there is an (Extremely small) market for them :v: (though thinking about it more it may be more for supercomputer poo poo than server poo poo)

I am curious for the low/mid-end Xeon 9000 and Epyc Rome comparisons, though.

Yeah that's true that HPC is the one area where they can make sense.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

Broose posted:

Question for anyone who can hazard a guess: How would Zen 2 compare to to a theoretical ice lake competitor? I get the suggestion that if Intel didn't flub 10nm so bad that, even with zen, the gap would have stayed the same as it always had at least on desktop.

Ice Lake should at a minimum bring Intel on par with Zen2 IPC and if you take their claims at face value they'll probably be on par with Zen3 IPC. The fly in the ointment is that 10nm clocks won't be as high as 14++, that's the whole reason Intel dropped the idea of a desktop Ice Lake, the clockrate drop will offset the IPC gains substantially (or perhaps even regress enthusiast performance).

It all depends on just how much of a trainwreck 10+ and 7nm turn out to be, and when they can actually get them going. If they'll do 4.6 (in an enthusiast desktop setting) and they're going up against Zen2, they win by maybe 5-10%. If they are going up against Zen3 (say +5-10% IPC) at 5 GHz peak clocks with a 4.6 GHz part, they're losing by 5-10%. If it's a complete trainwreck and they max out at 4.2 or something stupid like that they're obviously hosed.

So you also have to speculate about when Intel would have launched these hypothetical Ice Lake desktop parts. If yields wouldn't support it until Q3 next year anyway, or whatever, then they're obviously going up against Zen3 and a much more mature 7nm/7+.

Eventually 7nm will get to 14++ level clocks, the process will mature and AMD will tweak their layout to open up timing bottlenecks to increase clockrates and so on. 5+ GHz isn't that far away. It's just not going to happen in Q1 2019 on a first-gen design like a certain Scotsman predicted :v:

Paul MaudDib fucked around with this message at 18:20 on Jun 21, 2019

Cygni
Nov 12, 2005

raring to post

We should all probably pump the brakes a little here. The only numbers we have are ES leaks and AMD's own internal numbers, both of which they have manipulated multiple times before. We don't really know the IPC gains, we don't really know the clock behavior, we don't really know anything for a few more weeks when 3rd parties get their hands on em.

I'm definitely cautiously optimistic, but all the grand strategy discussions about how Intel will counter and all that stuff has no foundation until we actually know if Zen 2 will deliver not only in performance vs Intel, but in price vs Intel. 16 days, yall.

K8.0
Feb 26, 2004

Her Majesty's 56th Regiment of Foot
Yeah people are being absurdly optimistic about the position AMD will be in. I expect Zen2 to do pretty well, and I expect the enormous amounts of cache they're able to throw in by being MCM and 7nm to really help fill a lot of the performance gap vs Intel and even give them some solid wins, but they are not going to match or surpass Intel's single threaded performance in general. People also don't need all the threads AMD is selling, the average person is going to be fine with 6-8 cores for at least the next 5 years. AMD is probably going to be the most sensible option for almost everyone for at least the next year, but let's not be setting expectations so high that benchmarks disappoint compared to the fantasy. The realist take is that Zen2 is going to make it possible once again to buy a $200-300 CPU and probably not need to upgrade for 5 years, which is great.

K8.0 fucked around with this message at 19:23 on Jun 21, 2019

ilkhan
Oct 7, 2004

I LOVE Musk and his pro-first-amendment ways. X is the future.
Also remember that either one (Intel or amd) will do fine for 99.999% of people and situations. Even if Zen 2 does win the difference is not going to be night and day. Just having viable competition is the name of the game.

pixaal
Jan 8, 2004

All ice cream is now for all beings, no matter how many legs.


Cygni posted:

We should all probably pump the brakes a little here. The only numbers we have are ES leaks and AMD's own internal numbers, both of which they have manipulated multiple times before. We don't really know the IPC gains, we don't really know the clock behavior, we don't really know anything for a few more weeks when 3rd parties get their hands on em.

I'm definitely cautiously optimistic, but all the grand strategy discussions about how Intel will counter and all that stuff has no foundation until we actually know if Zen 2 will deliver not only in performance vs Intel, but in price vs Intel. 16 days, yall.

Intel slashing prices is probably a good sign that the benchmarks are at least somewhat accurate. There's no way Intel hasn't managed to get their hands on a sample.

okay not slashing, but Intel hasn't discounted much of anything recently so 15% is pretty big for them.

Gwaihir
Dec 8, 2009
Hair Elf

Cygni posted:

We should all probably pump the brakes a little here. The only numbers we have are ES leaks and AMD's own internal numbers, both of which they have manipulated multiple times before. We don't really know the IPC gains, we don't really know the clock behavior, we don't really know anything for a few more weeks when 3rd parties get their hands on em.

I'm definitely cautiously optimistic, but all the grand strategy discussions about how Intel will counter and all that stuff has no foundation until we actually know if Zen 2 will deliver not only in performance vs Intel, but in price vs Intel. 16 days, yall.

Intel never in a million years actually lowers their prices unless they know they have no other options at all to compete, which should probably tell us about all we need to know.

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!

K8.0 posted:

but they are not going to match or surpass Intel's single threaded performance in general.

I don't see how you can be any more certain of this than saying that they will match or surpass them, all of the numbers we have seen so far show them being very close.

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE
It's hard to know how the latency vs cache thing will play out. I still think AMD is most likely to slide in 0-5% behind in gaming/single-threaded perf but that's close enough in my book to not really matter. There will probably still be curveball titles that just really suck on Ryzen but we should now be talking about a 15% loss or something instead of like 40%.

The real pump-the-brakes thing is the lower-end parts. It's only the 3900X and 3950X that have the nice 4.6+ GHz boosts. If you can open up the TDP limits and XFR will boost to make up the difference on the lower-end parts then OK but that's definitely a "wait-for-reviews" thing.

Also I suppose in general we don't know how XFR will work at all on these, hopefully they can hold close to those specified boosts under all-core load. Again, assuming you open up the power limits, because lol at 16C boosting to 4.6 GHz on 105W, that's not going to happen. Pretty much figuring more like 200W there.

And finally there are thermals. These are real tiny chiplets, like 1/3 the size of Zeppelin, if you pump a lot of power through them (say 100W) they will get toasty, even though they are more spread out. How toasty? Wait for reviews. There's too many variables here.

But I mean, if they can hit a 10% clock increase and a 15% IPC increase then should pretty much close the gap, and barring any surprises that seems to be where things are pointing so far.

Paul MaudDib fucked around with this message at 19:58 on Jun 21, 2019

K8.0
Feb 26, 2004

Her Majesty's 56th Regiment of Foot

MaxxBot posted:

I don't see how you can be any more certain of this than saying that they will match or surpass them, all of the numbers we have seen so far show them being very close.

I'm quite certain of it because AMD seems quite certain of it. If they had Intel beat in ST performance, do you think they would have settled for yet again claiming parity like they have with Zen and Zen+ (which of course turned out to not even be close to reality)?

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!

K8.0 posted:

I'm quite certain of it because AMD seems quite certain of it. If they had Intel beat in ST performance, do you think they would have settled for yet again claiming parity like they have with Zen and Zen+ (which of course turned out to not even be close to reality)?

They showed parity in 1080p high fps gaming but they showed it ahead in non-gaming stuff, I expect them to be a bit short in gaming but general single-threaded performance would also include non-gaming tasks. Did they ever claim parity in gaming with Zen/Zen+ other than some dumb "see at 4k they're exactly the same" demo?

ConanTheLibrarian
Aug 13, 2004


dis buch is late
Fallen Rib
We know Zen2 CPUs will cheaper for equivalent performance than Intel, but I got the impression that the motherboards will be pricier. Will new builds still end up cheaper by going AMD once RAM and motherboards are accounted for? I'm assuming that there will still be some motivation to buy better RAM for Ryzens.

Drakhoran
Oct 21, 2012

Boards with the X570 chipset are only required if you need PCIe 4. If you don't, there's nothing stopping you from using a $115 B450 board.

Harik
Sep 9, 2001

From the hard streets of Moscow
First dog to touch the stars


Plaster Town Cop

iospace posted:

How so?

I mean when you're comparing AMD's best offering, you might as well include Intel's best. Yes the companies who will need a 64/128 or 56/112 in a server application* is probably countable on your hands, but it's still interesting to compare.

*Google, Salesforce, Microsoft, Amazon, Oracle, SAP... and that's really all I can think of of off the top of my head.
I had a segment on the ridiculous BGA-only parts but I trimmed it for brevity. It lives on in the high-core count per server because there's no way to approach 512 cores with 8x24.

"Costs as much as a luxury car, each, and you need four or eight of them and they have to be built bespoke for you" vs 8k off-the-shelf part isn't really a realistic comparison for anything but the absolutely most niche markets - which I already said are going to be Intel only for now.

PC LOAD LETTER
May 23, 2005
WTF?!

ConanTheLibrarian posted:

We know Zen2 CPUs will cheaper for equivalent performance than Intel, but I got the impression that the motherboards will be pricier.

Mobos are about the same or somewhat cheaper for AMD if you go X470/B450.

Its the higher end X570 mobos that will be expensive and that is already true for Intel as well.

That being said even the "garbage tier" X570 mobos in the ~$150-200 are looking to be better than the best X470 mobos going by what Buildzoid has mentioned so far. You're gonna want to wait for reviews because the BIOS is gonna matter lots if you want to push memory clocks and timings though.

K8.0 posted:

I'm quite certain of it because AMD seems quite certain of it. If they had Intel beat in ST performance, do you think they would have settled for yet again claiming parity like they have with Zen and Zen+ (which of course turned out to not even be close to reality)?

I think the biggest indicator so far that Zen2 will meet or beat Coffeelake's ST performance is that Intel is doing some preemptive (mild-ish) price cuts which they pretty much never do ever.

PC LOAD LETTER fucked around with this message at 23:14 on Jun 21, 2019

Combat Pretzel
Jun 23, 2004

No, seriously... what kurds?!
Hmmm, judging the EPYC prices, I'd figure the Threadrippers past 16C will be pretty pricey. Conversely, I hope that the 16C ones will come in a four chiplet comedy configuration, i.e. 4x2/2, so that there's twice the L3 per core (plus quad channel) over the 3950X.

Theris
Oct 9, 2007

PC LOAD LETTER posted:

That being said even the "garbage tier" X570 mobos in the ~$150-200 are looking to be better than the best X470 mobos going by what Buildzoid has mentioned so far.

Better in terms of build quality and power delivery, not features/connectivity. So if you want, say, tons of SATA ports or 2.5G Ethernet or the like (or lots of RGB) you might still be better off with a mid/high-end X470 board vs a similarly priced low-end x570 board.

For someone like me who doesn't care about that stuff, though, the Asus X570 Prime P is looking really good if it actually comes in at the rumored $150.

Theris fucked around with this message at 03:22 on Jun 22, 2019

ItBurns
Jul 24, 2007

Theris posted:

For someone like me who doesn't care about that stuff, though, the Asus X570 Prime P is looking really good if it actually comes in at the rumored $150.

Pair it with the $300 5Ghz 16 core 3700x, according to rumors.

Adbot
ADBOT LOVES YOU

taqueso
Mar 8, 2004


:911:
:wookie: :thermidor: :wookie:
:dehumanize:

:pirate::hf::tinfoil:

ItBurns posted:

Pair it with the $300 5Ghz 16 core 3700x, according to rumors.

If that is real they will sell so many of those.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply