Last year 16 core gen1 THREADRIPPER was selling for 1000$. Now it's 650$. Today, 32 core gen2 THREADRIPPER is selling for 1800$. That means 200$ less for twice the cores (if you were to buy 32 gen1 THREADRIPPER cores one year ago, you'd pay 2000$ for two separate processors, now you can get 32 cores in one single processor for 200$ cheaper) and better power envelope/tech-process.
Your thoughts, Jow Forums? It's obvious that AMD literally wins in every field right now, but do you think 2990WX is a good deal right now, as it is, or is it better to still wait for Zen 2 and inevitable price cut of gen2 THREADRIPPER too? What price do you expect for 2990WX to be at, when Zen 2 THREADRIPPER lands? I personally think they'd cut it down to 1200~1400$. I highly doubt they make it cheaper than 1000$ even when the time to phase out comes upon and price cuts will be abound.
My father's company's hosting has switched entirely to an AMD powered database. An obvious move, considering that due to sheer cost hy performance, AMD is murdering Intel in data centers. You'd be an idiot, out of money or sponsored by Intel to use intel right now.
single core is where it's a.........wait. half a core 1 thread is where it's at!
Justin Harris
Anybody got the MOAR LAKES image?
Dominic Wright
>I will never buy an AMD product.
holy shit. AMD's stock just plunged 645 points since this post.
Ryder Scott
For what? Do you have a better single core speed than 5.3ghz for $350?
Jonathan Wilson
If gen1 THREADRIPPER wasn't considered as "small business-viable" option yet due to it's "streaming enthusiast gaymen" field, gen2 THREADRIPPER with 32 is an absolutely worthwhile legitimate consideration now for anyone who has a small business/server/enterprise platform built and doesn't want to pay much more for EPBYNC. 2990WX is a perfect solution for both streaming/editingenthusiastic geymen AND people who actually WORK. AMD found an ideal field, a perfect spot. EPBYNC is now solely for medium and large sized business, while 32 core 2990WX is for any human who has small business. Lucrative as fuck. If, some time ago in the past, people who had small businesses were mainly choosing to go cheap/old Xeons, now that deranged practice of buying cheap trash is over, as gen2 THREADRIPPER literally runs light year circles around 99.99% of all Xeons garbage out there WHILE consuming less power AND being very decently priced for what it offers (both performance and features-wise).
>2018 >200watt CPU >180watt GPU >fml need special sockets and three-phase current installation The fuck is going on with desktops? It's not an industrial machine.
Blake Brooks
AMD are on track for 5% market share by the end of Q4. 10% next year. Companies test server CPU's long before making qauntity orders. Those tests have come back as no issues found. Those orders are already in.
Cry moar Intel baby.
Joseph Evans
>Inturd's baked """""TDP""""" is same as AMD's TDP
Parker Reyes
>200watts >industrial machine
Are you some sort of fairy 2-3 inches tall?
John Murphy
Just ingore him
Logan Ramirez
>AMD: 32 core for 1800 bucks (no hardware vulnerabilities) >Intel: 18 core for 2000 bucks (17 hardware vulnerabilities AND COUNTING)
get fucked Shintel. Literally kill yourselves
Ethan Adams
DELID
Jose White
>My father's company's hosting has switched entirely to an AMD powered database The next "my dad works at Nintendo trust me" post That's impossible for one of many reasons: >hosting Hosting what? Websites? VPS? VPC? Depending on what he's hosting, he's probably still using an Intel Xeon until all of the clients on the server is ready to be migrated. >considering that due to sheer cost hy performance What the fuck are you even talking about? Xeons have better TCO for very high bandwidth interconnects and latency sensitive clusters, and EPYC has better TCO for memory-intensive and mass scaling. There is no absolute cost-to-performance metric here, if you were legitimately telling the truth you'd be more specific. > AMD is murdering Intel in data centers Still hasn't happened because the downtime needed to take Xeon servers offline and replace them with EPYC is far more than live migrating older Xeons to newer Xeons. And what type of data center, you fucking mong? >You'd be an idiot, out of money or sponsored by Intel to use intel right now See what I just said. The financial impact and operation cost of downtime is one of the biggest reasons why Xeons are still being purchased 9 times out of 10 for existing enterprise solutions. It won't matter how much better EPYC is compared to Xeon, if the cost of downtime and restoring services is ten to a hundred times the cost of those machines over their 5 year lifecycle.
There's no Epyc on Zen+ so they have more of supply of dies so the price is lower. Zen2 will be split between Ryzen 3xxx, TR3, and Epyc2 so the price will be higher.
Jack Cook
zen2 will also be smaller dies so better yields again.
Cameron Brown
Let's not shoot ourselves on the dick, EPYC is still equally vulnerable to the newer versions of Spectre as Xeons.
Jeremiah Jackson
>t. Supantha Mukherjee
Jeremiah Russell
i REALLY wanted to wait for 7nm cpu's from AMD, their only releasing a 32 core server cpu this year, and by the time release comes in feburary, price settle down in august - i would have waited a year for the next gen cpus
going to buy a Ryzen 2700x literally in an hour for my build, its going to last me until 2022 when the 5nm cpus / nano carbon cpus start being talked about
Wish i bought AMD stock 2 years ago, but who invests in a company that looks like its dying on paper , oh how things have changed :)
Nolan Jenkins
>downtime to replace a node on a cluster You don't even know what you are talking about.
Brandon Hernandez
>AMD: 32 core for 1800 bucks (no hardware vulnerabilities) that doesn't require of you to change socket/motherboard every time and drops/cuts prices massively after 6~8 months after initial launch >Intel: 18 core for 2000 bucks (17 hardware vulnerabilities AND COUNTING), 14nm++++++++++++++++++++++++++++++++++++++++++++++++++++++++++, requires to change socket/motherboard completely every fucking time, an atomic holocaust oven, RF ID, no Hyper Threading, false/fake TDP, anti-competitive dirty-handed kikery practices, OverClocks high only on first 2 cores while downclocking massively on the rest (Zen doesn't get super-inflated useless 5GHz clocks, but it's 4+ GHz OverClocking is stable across ALL cores/threads), NEVER EVER drops prices any
Gavin Evans
He never mentioned whether this "data center" was scaled out or a single rack business. Also, >taking a node offline for 30 minutes to an hour has no impact on the customers being hosted on that particular node
Joseph Wilson
7nm EPBYNC is already coming out at the end of THIS year, so by the time 7nm RyZen, APUs, and THREADRIPPER hit, which will definitely be no earlier than mid-spring, everything will be in perfect balance, as enterprise/business would already have enough time to try out and move to 7nm EPBYNC before consumers get into the line. There will be no reason to increase prices massively due to such factors as great yields (already confirmed), adoption time between 7nm EPBYNC's release and home/mainstream offerings, and full backwards compatibility across RyZen/APUs, THREADRIPPER. It's just a fact of life that absolute majority of people won't be buying new Zen 2 motherboards at first day and would just update their BIOSes for the new stones, thus significantly cutting usage costs. Later, probably yes, some people would undoubtedly upgrade for the mere sake of better/newer features new motherboards might have, but not right away.
Jose Lopez
>EPYC is still equally vulnerable to the newer versions of Spectre as Xeons NetSpectre can't do jack to Zen due to SEV, you dumb FUD-producing fuck.
Evan Butler
>wanted to wait for 7nm CPUs from AMD, they're only releasing a 32 core CPU this year 7nm EPBYNC is coming out at the end of THIS year
Jeremiah Gonzalez
yeah but thats a server CPU?
even if it were for a desktop [ ive read AMD strictly say that is not the intention of this cpu ] its going to cost $1000 vs a $400 ryzen 2700x
jumping from 4 to 8 cores is healthy, jumping 4 to 16 is greedy for me :^)
Tyler Gutierrez
He is talking about servers you fucking idiot.
also fuck off with the 5.3ghz meme.
Wyatt Jones
EPBYNC is whatever you want it to be, baby. It can be enterprise, business, server, MySQL, or GAYMEN. Or all of it at the same time, without any sweat. Same with THREADRIPPER, but it's a lesser variant aimed at home users (while gen2 THREADRIPPER 2990WX is aimed at small businesses, not just home).
>jumping 4 to 16 is greedy How is that any greedy, when AMD officially cuts price of 1950X from 1000$ to 650$ (a fucking 350 DOLLARINO decrease) in less than 12 months? Other models got price cuts too, BTW. This is the very epitome of the phrase "ALTRUISM".
Leo Williams
i see your point, im pretty sure the EPYC cpu is going to have lower clock rates though, which is why its not intented for regular desktop/gaming use
id love to get 16 cores, but spending $420 cad on ryzen 2700x seems to be good enough
For one, 2990WX was already confirmed to hold stable 4GHz across all cores cores & threads, no downclocking. If 32 core THREADRIPPER can do it, 32 core EPBYNC can do it too, naturally.
Samuel Myers
amd got 1% on the server market in literally one q
doesnt understand how huge it is
Easton Edwards
EPYC 2 is going to make the Earth move and Intel jump off a cliff. Threadripper 2 32/64 on 12nm doing 4Ghz+ on all cores? EPYC on 7nm is gonna be the ultimate shit wrecker.
Easton Fisher
>EPYC on 7nm is gonna be the ultimate shit wrecker THREADRIPPER MK III or MK IV will be.
I also work in IT and we are looking at AMD because of the price savings. Intel Xeon are good processors but they're massively overpriced.
Hunter Hernandez
>16-32 cores take up more room on the chip, Threadripper has 4 separate dies, it is basically 4 Ryzens connected through infinity fabric.
Wyatt Brooks
>Intel Xeon are good processors How is it there, in 2012?
Dylan Thompson
What about 64 core Rome?
Hunter Roberts
>That means 200$ less for twice the cores (if you were to buy 32 gen1 THREADRIPPER cores one year ago, you'd pay 2000$ for two separate processors, now you can get 32 cores in one single processor for 200$ cheaper) No, that's not what that means. Learn some fucking English.
Evaluating does not equal buying. My company purchased two Tyan servers with dual eight-core EPYCs to see how much it would cost to transition our locally hosted services that relied heavily on memory bandwidth. The performance was very good and it practically justified replacing three or four dual socket Haswell Xeon E5s that we currently had. BUT There was no way for us to live migrate the virtual servers on those servers and required at least 2 hours of downtime to replace the Xeons on machine at a time. Our cadence was going to be >both EPYCs going online >migrating all the clients of one Xeon to the others so we can shut it down >coping the client VMs of that Xeon server on the SAN for both EPYCs >repeat until all Xeons had been successfully migrated But there was no way for us to cleanly migrate all of the running VMs, or preserve whatever changes were made to the VMs after they were copied within the SAN. The test migration lost about 3 hours of work from a local department (thankfully, none of the changes that were lost were that significant, since the users had been advised of the service disruption beforehand and had access to a temporary local NAS). There were significant bandwidth and connection disruptions as well, since their backbones were being hammered with significantly higher bandwidth usage and caused our NGFs to freak out.
Had we gone full ham and replaced our EOL Xeon servers with EPYC, our company would have lost millions per minute of downtime based on the test migration. There was no way in hell our company could stomach those losses just to get different hardware installed when upgrading to Skylake Xeons would have had near zero downtime (you can live migrate from older Xeons to newer Xeons).
Liam Carter
Honestly we don't know if 32core threadripper is using four dies or two yet. Once it gets delidded we should know.
William Brown
What he meant was $200 less for the same numbee of cores, because he's a fucking ESL.
Angel Rogers
I think the surge in AMDrones is from paid shills or retards with buyer's remorse.
Intel still wins every single benchmark with 2 year old architecture.
It's been this way on Jow Forums for as long as I can remember. AMD is *always* on the cusp of releasing something competitive and *every* time it just forces their competition to actually release what they've been sitting on while AMD catches up.
>Intel still wins every single benchmark Guyse, show him that "Xeon gold turd" collage, please. I don't have it on my main PC as of right now (need to hook up my external HDD for that and I'm kind of lazy right now).
Andrew Bell
The 32 core threadripper to be released is going to be 4 dies as there is no 16 core dies yet. One issue with threadripper 2 is that despite 4 dies the memory configuration is the same with only 4 channels split across 2 dies so the 2 other dies won't have direct memory access.
Noah Martinez
>don't switch to epyc because i lose money >stay on xeon platform with inferior performance and power consumption and a fuck ton of security holes >lose a fuck ton of money for staying on intel for years on end instead of switching to epyc Brilliant.
Ian Butler
Sounds like Intel has you by the balls. You could of course make a copy of the cluster oonto the new servers and when ready flip the switch to AMD. But it's probably not as simple as that. At some point you will need to upgrade and when the time comes you may as well bite the bullet because Intel is not going to be competitive for a while.
>In games >In single threaded Adobe Photoplop >In shitty PS2 emulators OK
Luis Scott
Ironic retardation is still retardation.
Wyatt Jackson
Photoshop been multi-core/threaded since CS 6. PCSX 2 been using at least 2+ cores/threads since 0.9.5 (it's 1.4.0, currently).
Camden Baker
>i lose money >lose a fuck ton of money for staying on intel for years on end instead of switching to epyc Did you not read the part where I said a minute of downtime = millions of dollars? That's no small drop in the bucket. Also, power consumption costs for the newer Skylake Xeons and the EPYC servers were about equal. Neither of them produced any more heat than the Haswell machines, nor did we had any hardware compatibility issues that needed to be addressed. I think EPYC needed some of our VMs to be recreated because of a fault that AMD have addressed by now. >stay on xeon platform with inferior performance and power consumption and a fuck ton of security holes It's not like we had a choice. Intel knows this and upped the pricing for near-equivalent Skylake Xeons. We paid about $5-12k more per server, but that's still more favorable than losing hundreds of millions because our company's day-to-day operation ground to a fucking halt. And yes, we operate in several different time zones, so there was no time period that would have reduced our losses even more. >Sounds like Intel has you by the balls It's VMWare too. I'm pretty sure they could make live migration possible between two x86-64 architectures if we can migrate Core2-era Xeons to Haswells without a problem. >You could of course make a copy of the cluster oonto the new servers and when ready flip the switch to AMD That's pretty much what we planned, but there are still thousands of users/services that still need to access VMs on those servers and any changes that are made between us copying those instances and bringing the new hardware online would be lost. That's why live migration is so important and planning for service outages is not as simple as shutting down one server and flipping the switch on another.
Liam Williams
Now this is EPYC!
Julian Harris
>It's not like we had a choice. Well then you're free to stay on a shitty inferior platform while all the other companies switch to Epyc because "i'm gonna lose 6 gorillion dollars"
>a minute of downtime = millions of dollars Lesser problem/damage than being fucked completely via Meltdown/Spectre or other secuiry hole. Also, you're a fucking retard if you think that up to -57% of performance loss in enterprise/server and MySQL due to "full MeltDown patch" is thing that should be ignored "just because I'm afraid of a little downtime which can cost a couple mils". -57% in performance loss equals massive slowdowns AND downtime for end user, which in itself translates to massive loss of credibility as a service, loss of clients, which in itself leads to monetary losses even bigger than that of a "slight downtime due to change from Xeon to Epyc". You are a fucking moron that doesn't know how to do business and I hope your company goes fucking bankrupt, because there are such idiots like you working in it.
Isaiah Martin
He's probably larping but can't be absolutely sure.
Just wait until the partners switch to a provider that uses AMD.
Dylan Miller
The thing is if they are such a big business they could very well call in AMD engineers to help them transition without too much loss of money. I am pretty sure AMD would be glad to help out if they are getting a big order in the process. But yeah. Stick with Intel and wait for shit to hit the fan.
Jason Cook
A shitton of big and medium-sized businesses alike has already jumped the Inturd shit and switched to Zen, this is one of the main reasons why Intrash had such a massive drop at stock exchange last friday. Sure, there were other additional factors (such as them COMP-FUCKING-LETELY missing the earnings estimates while AMD greatly SURPASSED expectations in a good way), but main one was the mere fact medium-sized and large businesses finally had enough of the retarded garbage BS Inturd been doing for the last 20 months to everyone, switching to Zen by large with no hesitations or remorse whatsoever. Intrash is fucking dead-in-the-water right now, holding by last strands. Literally NO ONE likes them anymore, even most of the dumbest shareholders and apologistic brainwashed cucks of theirs. It's worse than Titanic. Literal fucking TOXIC actives.
John Price
>They could very well call in AMD engineers to help them transition without too much loss of money. I am pretty sure AMD would be glad to help out if they are getting a big order in the process. That's exactly what AMD does currently, with the Project 47.
>Stick with Intel and wait for shit to hit the fan. It already did that, last friday. Big-time.
Wow, this over-reaction is completely unnecessary. The patches for Meltdown and Spectre did not significantly impact any of our VMs because most of the applications they run is memory dependent. 57% is a huge exaggeration and I don't think any of our servers experienced cumulative performance losses higher than 10% since the latest patches. I don't know what the total financial impact of the Spectre and Meltdown bug was, but I can't imagine it being any more than a few thousand dollars from additional power consumption across all of our servers (which, again, wasn't that significant).
And no, BIOS updates did not cause any significant downtime because of, you guessed it, live migration and excess capacity across all of our servers.
>patches for Meltdown and Spectre did not significantly impact any of our >-57% is a huge exaggeration Stopped reading right there and hid the moron. You should do so too, anons of Jow Forums. This is literal imbecility all way through and through.
Isaiah White
>I don't know what the total financial impact of the Spectre and Meltdown bug was, but I can't imagine it being any more than a few thousand dollars Ask Google and Amazon, you dumb fuck. The losses were in tens of millions.
Luke Reed
Calm down guys. He believes his four (4) servers are representative of the entire server industry. You can't argue with someone like that.
Easton Price
>The thing is if they are such a big business they could very well call in AMD engineers to help them transition without too much loss of money We investigated the option when we ordered the two evaluation servers, but we did not want to spend more money invested into a potential dead-end than we already had with those two servers. Before the rabid fanboys jump in on the "dead-end" part, you need to understand that AMD has not been a serious contender in the server market because they pretty much cut us dry with the last of the good Opterons. We had no way of trusting that AMD would be able to provide the level of support and reliability that we're used to with Intel's hardware. AMD could have folded in 2018 and we would have been shit out of luck with dead platforms and no support.
Intel might be making shit, overpriced products, but they solid as far as support goes to OEMs, which then extends over to us, the customers. AMD needs to exhibit this level of support and that Project 47 thing mentioned here is a good start.
Austin Robinson
>AMD could have folded in 2018 and we would have been shit out of luck with dead platforms and no support. So this is someone talking with and intel envelope covering his eyes.
Henry Mitchell
We don't have four servers. I said that a single EPYC would have been enough to replace three to four of the existing Haswell server.
Judging by your complete dismissal that different customers have different uses for the same hardware, I feel that it's safe to ignore your fake news-driven hyperbole as well.
So this is someone talking with an AMD fanboy membership card jammed into both eyesockets. AMD was not in a good place financially in 2016 to 2017, and EPYC was too new to get any reliable figures out of that didn't come from an underfunded marketing department. I never said that EPYC was a bad platform, it just was not suitable, cost efficient, or the lowest risk option at the time. It still isn't until VMWare makes live migration possible between EPYC and Xeon, and when EPYC has enough adoption to have a solid support base we can rely on. Like I said, it's getting there, but it's still a giant risk versus the more ubiquitous and established Xeon options.
Also, I saw how much the newer Intel Xeons were going to cost us. We are hoarding the Skylake Xeons for the next year or so.
Not the same workload. That's like posting the AVX-512 benchmarks and laughing at how far behind EPYC is versus Skylake. Different applications, different requirements, etc.
Michael Allen
If they spent that shit marketing money on ipc performance research they'd have doubled their profit.
Dominic Jackson
Yeah he's a dumbfuck who tries to justify shitty platforms
Gavin Roberts
>Yeah he's a dumbfuck who tries to justify shitty platforms >It's not like we had a choice. Intel knows this and upped the pricing for near-equivalent Skylake Xeons. We paid about $5-12k more per server, >It's VMWare too. I'm pretty sure they could make live migration possible between two x86-64 architectures if we can migrate Core2-era Xeons to Haswells without a problem. >Also, I saw how much the newer Intel Xeons were going to cost us. We are hoarding the Skylake Xeons for the next year or so.
Nobody is justifying anything. Not every purchase decision is done by a single individual, nor is it as straight-forward as "we can replace three dual-socket Xeon Haswells with a single dual socket EPYC, let's buy them immediately". Intel has our company by the balls and are asking us to pay more for the same capability as our existing platform. We don't have a choice if we don't want to disrupt the workflow of several offices across the world and lose more money in a 24 hour period than the TCO cost savings of each EPYC server over its service life.
Talking to rapid AMD fanboys is like talking to a retarded white Trump supporter, no amount of facts or explanations will ever convince him that he's a racist bigot.
Zachary Scott
>I said that a single EPYC would have been enough to replace three to four of the existing Haswell server. Then fucking DO IT ALREADY, you DUMB MONGREL SHITS. Haswell is fucking trash anyway.
Grayson Lewis
>fake news-driven hyperbole You are a fucking lamer. Clear and simple. Absolute dummy.
Gabriel Watson
>AMD was not in a good place financially in 2017 Holy shit you're retarded. Literally couldn't be more wrong.
Brayden Thompson
>I never said that EPYC was a bad platform, it just was not suitable, cost efficient, or the lowest risk option at the time >It still isn't Typical Intel-bribed dumb fuck right here, ladies and gentlemen. Literally every other big business is jumping to Epyc right now, but this stupid fuck decides to put his head into sand while closing his ears and screaming "LALALALALALA, CAn"T HEAR YOU ALL!" in utterly autistic denial of the truth of the surrounding reality that's going on right at this current time. Fucking disgusting. Either just come out of the closet, admit that you've been bribed, and tell us all openly how much Intrash paid you, or fucking go and off yourself hard right this very instant, you piece of shit.
Kevin Campbell
>Not the same workload. It's literally the same in 99.82% of all other business/server/enterprise workloads that worth any damn shit, you piece of cocksucking loser trash. MySQL, for one, was utterly SLAUGHTERED, absolutely ASSSRAPED by this so-called "Meltdown patching".