|
bull3964 posted:The first wave of the 12G servers are live now. You can order the R620, R720, and R720xd with them. Holy poo poo these are amazing. Even the 620 is loaded with 24 DIMM slots. I'm in love. E: And I can price them out in my Dell Premier ghetto, even better.
|
# ? Mar 6, 2012 22:53 |
|
|
# ? Mar 28, 2024 09:29 |
|
I hadn't seen this posted in the thread yet - the pricing list and model equivalents of Ivy Bridge to Sandy Bridge processors. http://www.techspot.com/news/46727-intel-ivy-bridge-desktop-chip-prices-leak.html So it looks like the hallowed i2500k will now be the Core i5-3570K, and cost just about the same. However, I still can't find anything that gives a clear date for retailers carrying the processors. Everyone keeps hinting at early April, is this still the estimated date?
|
# ? Mar 6, 2012 23:10 |
|
Abeya Minora posted:However, I still can't find anything that gives a clear date for retailers carrying the processors. Everyone keeps hinting at early April, is this still the estimated date?
|
# ? Mar 6, 2012 23:20 |
|
FISHMANPET posted:Holy poo poo these are amazing. Even the 620 is loaded with 24 DIMM slots. I'm in love. Yeah, the R620, R720 and R720xd are all basically the same machine now, just different chassis configurations (1U, 2U, 2U with assload of drives.) Specing these machines out seem so much less byzantine. I also love how you can do 1 socket now and still load the server up with 192gb of ram if you need to. It can be a real help with SQL Server Enterprise licensing costs if you are memory bottlenecked but have a low CPU load.
|
# ? Mar 6, 2012 23:52 |
|
RealWorldTech has posted their Sandy Bridge-EP article. It's not a review, but it goes into a lot of low-level detail about the technology.
|
# ? Mar 7, 2012 00:26 |
|
Alereon posted:that Intel China exec. Sean Maloney?
|
# ? Mar 7, 2012 01:18 |
|
Anand has posted a non Intel sanctioned preview of the Ivy Bridge i7 3770K. It's about what was predicted. ~5%-15% better CPU performance and ~20%-40% better GPU performance. It also has low power consumption, about on par with the 2400 under load. It's AF performance is also better than with SB, though I can't seem to find a picture for comparison. This is the Ivy Bridge AF test from Anand's review.
|
# ? Mar 7, 2012 02:57 |
|
Pretty much exactly what we expected, which is impressive. AMD is really going to need to hit it out of the park with Trinity to remain competitive for integrated graphics, and I really hope this is the final nail in the coffin of poo poo low-end graphics cards.
|
# ? Mar 7, 2012 03:48 |
|
Alereon posted:Pretty much exactly what we expected, which is impressive. AMD is really going to need to hit it out of the park with Trinity to remain competitive for integrated graphics, and I really hope this is the final nail in the coffin of poo poo low-end graphics cards. It seems like for the IGP part Llano is still generally 'faster' than even this IVB cpu he's benching. I would assume Trinity will be even a step beyond this. It seems to me like AMD actually has a solid niche for 'gaming' laptops. IVB is obviously the better CPU, but my guess is Trinity will be the better 'mix' of GPU/CPU for 1366x768 resolution, which seems to be the typical resolution for your average 15.6" laptop. Especially since you can get the hybrid crossfire thing going. I would think you could have a pretty nice gaming experience at that resolution, most graphics features turned on and nice framerates.
|
# ? Mar 7, 2012 17:18 |
|
Alereon posted:Pretty much exactly what we expected, which is impressive. AMD is really going to need to hit it out of the park with Trinity to remain competitive for integrated graphics, and I really hope this is the final nail in the coffin of poo poo low-end graphics cards.
|
# ? Mar 7, 2012 17:24 |
|
Add to that the lower power Trinity APUs that can be put into "ultrathin" notebooks are supposed to be as powerful as Llano while drawing half as much power. So even these ultrathins would be more capable at gaming than even a regular laptop with just Ivy Bridge.
|
# ? Mar 7, 2012 19:45 |
|
Sandy Bridge-EP is really impressive. We've seen very impressive numbers across the board and the four socket platform coming out in a month or two is going to be an interesting option.
|
# ? Mar 7, 2012 20:28 |
|
Too bad 32gb RDIMM DDR3 is so expensive and really too bad that 32gb LRDIMM DDR3 is loving insanely expensive. The cost of memory really ramps up when looking at the two socket solutions as capacity goes up. I was looking at various configs for an R620. Yeah, these are dell ram prices but you aren't going to find them a ton cheaper elsewhere. 256gb 16x$400 = $6400 320gb 20x$400 = $8000 384gb 24x$400 = $9600 448gb (4x16gb, 12x32gb) = $30,000 512gb (16x32) = $40,000 768gb 24x32 = $91,200 Once you have to start using 32gb RDIMMs and then 32gb LRDIMMs to get the max density, the price jumps quickly. But it is nice to know I can load a server up with 384gb of ram now, and if the 32gb modules drop an assload in the next two years or so I can double the memory capacity of the server without having to buy a new machine. bull3964 fucked around with this message at 21:16 on Mar 7, 2012 |
# ? Mar 7, 2012 21:11 |
|
I probably missed all this discussion when SNB came out, but, are people not a little miffed that in order to get the unlocked chips you have to also pick up the integrated GPU? 95% of people who intend to overclock are going to have a pcie card anyway. Wouldn't the chip be cheaper, smaller, and potentially cooler without a whole bunch of unused transistors? Can you even run an extra monitor off the IGP if you have a video card?
|
# ? Mar 8, 2012 23:58 |
|
It would cost more to make 2 separate dies, but they do have CPUs with the GPUs cut, the 2550k. It's not like the GPU's wasting a significant amount of power when it's disabled either.
|
# ? Mar 9, 2012 00:10 |
|
craig588 posted:It would cost more to make 2 separate dies, but they do have CPUs with the GPUs cut, the 2550k. It's not like the GPU's wasting a significant amount of power when it's disabled either. Aren't those 2550k's just 2500's with the GPU disabled [probably due to defects?]. Its not as if they got new films for a new ASIC.
|
# ? Mar 9, 2012 00:28 |
|
Fuzzy Mammal posted:I probably missed all this discussion when SNB came out, but, are people not a little miffed that in order to get the unlocked chips you have to also pick up the integrated GPU? 95% of people who intend to overclock are going to have a pcie card anyway. Wouldn't the chip be cheaper, smaller, and potentially cooler without a whole bunch of unused transistors? Can you even run an extra monitor off the IGP if you have a video card? The GPU can be completely power gated, in which case it's completely powered down when not in use. Not sipping a milliwatt. Chip pricing is not very dependent on marginal cost of production, because there are enormous investment costs, both in design and in setting up manufacturing. Intel decided it was more cost-effective to sell the same chip binned different ways than it would be to work up a different chip design and prepare a separate manufacturing chain for the relatively tiny enthusiast market segment. Same reason LGA 1366 and 2011 chips are harvested Xeons and not unique designs. So while the chip might be smaller without the GPU section, it wouldn't be cheaper, and it also wouldn't be cooler. And yes, you can run an extra monitor off the IGP. Two, in fact, or three on Ivy Bridge. Or you can use Virtu and run the dGPU through the IGP's frame buffer, and that will let you use QuickSync and a hefty video card all with only one monitor.
|
# ? Mar 9, 2012 01:36 |
|
Well that makes sense. I guess I got the impression they were all pointless from the SemiAccurate page. In hindsight I think that dude is just butthurt that nobody cares about linux?
|
# ? Mar 9, 2012 02:12 |
|
Since future Intel and AMD processors will have a GPU built in will we be seeing games taking advantage of it for physics?
|
# ? Mar 9, 2012 03:36 |
|
That depends on the willingness of physics middleware providers to write GPGPU support, whether the hardware is available even if a monitor is not plugged into it, and the limits (if any) that nVidia's Aegia/PhysX patent portfolio (if there are patents) place on these attempts. Probably won't be able to see it for a while.
|
# ? Mar 9, 2012 03:50 |
|
I'm excited, I should have two Dell R620s to play with by the end of the month with each having 2x E5-2690. These will be replacing servers with single X5670s in them. So, we'll be going from 6 Westmere-EP cores at 2.93ghz (max turbo 3.3ghz) to 16 Sandy Bridge E cores also at 2.9ghz (max turbo 3.8ghz.) Technology is getting scary cheap.
|
# ? Mar 9, 2012 04:41 |
|
Said the person buying processors costing 4 iPads each.
|
# ? Mar 9, 2012 05:49 |
|
Factory Factory posted:Said the person buying processors costing 4 iPads each. Well, I'M not buying them . Still though, my career is only 11 years old and I am continually amazed at the pace of technology. It doesn't seem that long ago that I was on my first job out of college, upgrading to SQL Server 2000 on a BLAZING fast PIII Xeon 700 who's leased price was probably more than the $15k /machine cost of what was just ordered.
|
# ? Mar 9, 2012 06:42 |
|
Factory Factory posted:Said the person buying processors costing 4 iPads each.
|
# ? Mar 9, 2012 09:35 |
|
japtor posted:They're a bargain when you consider the processing power vs the iPad's Let's see sixteen SB-E cores entertain you whilst you poop, all at 2048x1536
|
# ? Mar 9, 2012 17:56 |
|
Factory Factory posted:Said the person buying processors costing 4 iPads each. Could be worse- they could be top-bin E7s This is pretty cool: Mod note: please re-host that. We frown upon image leeching. edited: leeching is bad in a well actually fucked around with this message at 19:39 on Apr 6, 2012 |
# ? Mar 9, 2012 18:04 |
|
Factory Factory posted:The GPU can be completely power gated, in which case it's completely powered down when not in use. Not sipping a milliwatt. To be pedantic, power gating transistors (like all transistors) still leak a little power. Not a lot, and there are techniques to minimize it, but it could potentially be more than a milliwatt. Fuzzy Mammal posted:In hindsight I think that dude is just butthurt that nobody cares about linux? Oh yes, very much so.
|
# ? Mar 9, 2012 21:49 |
|
Have anyone seen a motherboard with the new Z77 chipset that can do 3 or more sata 6 in raid 0? I got 3x SSD in raid0 on a p55 chipset MB now, and I'd really like to keep running that if I upgrade and use it on sata 6.
|
# ? Mar 14, 2012 10:01 |
|
Supradog posted:Have anyone seen a motherboard with the new Z77 chipset that can do 3 or more sata 6 in raid 0? I got 3x SSD in raid0 on a p55 chipset MB now, and I'd really like to keep running that if I upgrade and use it on sata 6. There will be no such thing unless you do software RAID. Z77 only has 2 SATA III ports, and third-party ports won't do RST FakeRAID.
|
# ? Mar 14, 2012 10:29 |
|
Bah, that's what I feared. Next upgrade gotta include a ServeRAID M1015 or similar, or some new SSDs then.
|
# ? Mar 14, 2012 10:43 |
|
SemiAccurate posted an article explaining the Ivy Bridge delay. There were two factors: Firstly, several large customers shifted their orders to cherry-picked ultra-low TDP processors. Secondly, a "very large customer" (Apple?) shifted their order from CPUs with HD 2500 to HD 4000 graphics. Both of these factors meant that Intel needed several additional production weeks to stockpile the appropriate processors for the launch.
Alereon fucked around with this message at 02:05 on Mar 19, 2012 |
# ? Mar 19, 2012 02:02 |
|
I missed this thread. Intel has been hard at work on Lynx Point (Haswell's PCH), submitting patches ahead of the curve to get VIDs/DIDs into mainline. Some patches for AVX2 are already appearing in GCC as well. Also I discovered today at work that Intel apparently tries to "wear-level" when it comes to making changes to their ethernet adapter NVM, which was kind of interesting!
|
# ? Apr 6, 2012 02:17 |
|
theclaw posted:
|
# ? Apr 6, 2012 08:19 |
|
McGlockenshire posted:That's ... freaking scary. It also doesn't exist on Ark, which would normally lead me to believe that it doesn't actually exist. Except that this EPSD product naming guide on Intel's own site (PDF) mentions it, as well as many of the other server boards that Ark acknowledges. Weird. I guess it shouldn't be a surprise that they were going to make chips for a 4-way system, but I am shocked that they're fitting it all on one level. You could potentially jam that into a 1U, right? 32 cores, 64 threads in a 1U? Now you're starting to get into SPARC territory.
|
# ? Apr 6, 2012 19:41 |
|
FISHMANPET posted:I guess it shouldn't be a surprise that they were going to make chips for a 4-way system, but I am shocked that they're fitting it all on one level. You could potentially jam that into a 1U, right? 32 cores, 64 threads in a 1U? Now you're starting to get into SPARC territory. The only limitation I could see in putting it in a 1U is removing enough heat and having beefy enough power supplies that fit into 1U. This should be the board in the upcoming R820 from dell. I'm just not sure when they are going to launch. From what people are speculating, the next line of 12G servers are going to launch in May with a 3rd line in July. They still have to launch their 1 socket and 4 socket 12G line, so it's a tossup which will come out in May (I'm betting the single socket line.) We got our brandy new R620s last week, but I haven't had the opportunity to even unpack them yet. I really want to start playing with 16 cores of Sandy Bridge goodness.
|
# ? Apr 6, 2012 20:21 |
|
FISHMANPET posted:I guess it shouldn't be a surprise that they were going to make chips for a 4-way system, but I am shocked that they're fitting it all on one level. You could potentially jam that into a 1U, right? 32 cores, 64 threads in a 1U? Now you're starting to get into SPARC territory. There are 4-proc 1Us available right now, like systems based on the Supermicro X8QBE-F motherboard, though the notes I have here say that any processors with a TDP greater than about 95W are "potentially not thermally viable" in 1Us. That board is also only for the previous generation, so 10 cores per processor at best. The last 4-proc machine I saw from Intel was a 4U affair that packed all four processors up front and then used memory risers in the back. Their previous generation 8-proc machine uses dual processor CPU/memory risers.
|
# ? Apr 6, 2012 20:53 |
|
bull3964 posted:The only limitation I could see in putting it in a 1U is removing enough heat and having beefy enough power supplies that fit into 1U. On one hand I'm interested in the E5-1600 series for VMWare, because the chips are so powerful you don't need a lot, and at the smale scale I'd rather build out than up. On the other hand, stuffing twice as many servers full of 10Gb Nics changes the value proposition quite a bit. I guess if Dell can get the their 1 socket E5 down far enough in price it will be worth it. One socket really doesn't need the 24 RAM slots of the dual proc servers.
|
# ? Apr 6, 2012 21:22 |
|
Anandtech has an Intel Z77 Panther Point chipset and motherboard preview up.
|
# ? Apr 8, 2012 05:24 |
|
There's a lot on LucidLogix in there that makes me ache for a GPU thread.
|
# ? Apr 8, 2012 06:02 |
|
|
# ? Mar 28, 2024 09:29 |
|
That auto-BIOS update over WAN is pretty sexy. I guess my company isn't alone anymore. Would be nice to see some PCIe 3.0 SATA/SAS HBAs for cheap NAS platforms too. I had to get a X58 over SNB because it didn't have enough PCIe lanes.
|
# ? Apr 8, 2012 06:13 |