|
Factory Factory posted:Is that the top 4-core or comparing 8-core to the i7-2600K's 4-core? AMD seems to be going all-in on multithreaded performance if that's an 8 vs. 4 comparison. Then again, that would mirror their graphics strategies. If that slide is real, the bench-marked used doesn't efficiently scale past 4 threads I don't think. So if its not a fake, then what you're seeing is effectively the performance of a 2 module 4 core BD chip being equivalent to a 2600K. Personally I think the slide is a fake.
|
#
¿
May 4, 2011 23:26
|
|
|
|
| # ¿ Apr 26, 2024 15:39 |
|
|
PC LOAD LETTER posted:Its supposed to be the 8 "core" 4 module chip that is being bench marked there. Clock speed is unknown but its probably at or near peak base speed at the very least. I'm aware of that, but if the benchmark only effectively uses 4 cores, then the performance you are seeing is closer to that, or nearly identical to, a 4 core BD chip. it also makes the 2600K similar or almost identical in performance to a 2500K. If thats the case, AMD has a decent chip, because being able to compete with the 2500K with a quad core BD would be amazing. Current AMD offerings are probably 30% slower than a 2500K.
|
|
#
¿
May 5, 2011 16:21
|
|
|
Alereon posted:Here's the Anandtech article about Graphics Core Next (GCN) for those who haven't seen it. I'm pretty sure that GCN is still pretty far out, more likely a target for the Radeon HD 8000 or even 9000 series. I think it's confirmed that Southern Islands, the Radeon HD 7000-series due for release in a couple months, will be VLIW4-based. Bulldozer APUs are definitely VLIW4-based. I think AMD is likely to maintain their power usage lead for awhile, since it's more of a philosophy of aiming for a lower target, and AMD has a significant technology lead in memory controllers (one of the reasons the R6870 has such good efficiency). I do not believe this is accurate. Right now, based on what we can see in the drivers [check beyond3d forums] there will likely be a low end VLIW-4 card, to be used in the hybird CF for Trinity [BD APU], with the other designs all being GCN based.
|
|
#
¿
Jul 11, 2011 05:31
|
|
|
Alereon posted:I'm really pretty excited about the Radeon 7800-series myself. While the 7900-series looks cool, it's a brand new architecture so I'd rather wait until the kinks get ironed out. The leaked specs for the 7800-series seem to be exactly what I was hoping for, a die-shrunk 6900-series with much lower power consumption (and hopefully great overclocking!). The 77xx, 78xx, and 79xx are all based on the new GCN architecture. Although rumors are that the 78xx will have exactly the same amount of CU's as 69xx series, which will allow us to make some interesting comparisons.
|
|
#
¿
Dec 19, 2011 19:49
|
|
|
Beelzebubba9 posted:So basically AMD gets a full node process shrink and over a year to make a card that's <25% faster than the GTX 580, which itself was just a re-spin of the GF100/GTX 480. And the 7970 is priced higher. Maybe I'm just spoiled by the previous generation of GPUs from AMD (or Intel's ability to get 25-300% performance increases between CPU generations on the same process), but it's hard not to feel a little disappointed. The 580 is about 15-25% faster than a 6970, and plenty of people purchased the 580 instead of the 6970, at a very large premium in price. Those same KIND of people who are the market for a high end video card right now would certainly be willing to pay $50 more for a 7970 for 15-30% more performance than a 580 today. I'm not one of those people, but I never expect the 'top end' videocard to be one that I will be interested in buying. I also expect that over the next few months there will be improvements in the driver to increase that margin slightly. I'm pretty sure when the 560ti launched compared to the 6950 1GB, the 6950 was the same or marginally slower, and now its the same or marginally faster. GCN being a new graphics arch, I think we'll see similar modest gains just from the driver. Will be interesting to see what nvidia is able to do, a few months from now. I wouldn't be surprised if nvidia's midrange is almost as fast as the 79xx series at launch. We'll have to see if nvidia has figured out how to do a new process without falling on its face. If we compare the 5870 to the 7970, the 7970 looks like a very nice healthy jump, and considering that the 6970 was only born out of a TSMC node cancelation, the 7970 is probably what we could have expected if the node hadn't be canceled. In that respect the 7970 is pretty respectable. the 7990 will probably be able to be almost as fast as 7970CF, without going over the PCI-E spec. Unless nvidia has more performance per watt, I think AMD will still have the better SINGLE UBER CARD. When you are power constrained, it pays to be the best perf/watt.
|
|
#
¿
Dec 22, 2011 19:42
|
|
|
Not sure what to make of this. http://semiaccurate.com/2012/01/19/nvidia-kepler-vs-amd-gcn-has-a-clear-winner/
|
|
#
¿
Jan 19, 2012 22:44
|
|
|
Factory Factory posted:I may be working on a small sample size, but SemiAccurate seems like a site full of hysterical horseshit to me. This is true mostly. I take everything he says with a huge container of salt. He does know people at foundry and engineering levels though. There is often information that comes out pretty early from him that ends up being correct later. Not always, but often. Mostly he's the most anti Nvidia person in the world, so if this is a serious post, then it's quite something to say 'they win handily', considering usually he finds ways to knock nvida products even when they are pretty good.
|
|
#
¿
Jan 19, 2012 23:00
|
|
JESUS gently caress THIS IS AN ABOMINATION! THIS IS A CRIMINAL ACT OF DECEPTION AND FRAUD BEYOND ALL KEN, JUST LIKE THAT TIME NVIDIA USED WOOD SCREWS AMIRITE?"
|
The midrange is very much stagnated. I got my 5850 for $260 over 2 years ago. The part to upgrade to right now would be the 6950 for about the same $$, and although it does have more performance, it's not appreciably enough so for me to drop ANOTHER $260. At least not right now. Because of AMD moving the 79xx up in price to 550/450, this means the 78xx will be slightly faster than the 69xx, while costing as much, or slightly more. Leaving me nothing to upgrade to at $260 that feels compelling. I'm hoping that the GK104 is as competetive with Tahiti as has been rumored, forcing AMD to lower prices to $415/$330 for the 79xx, making the 78xx probably $300/$240 or so. Maybe finally I'll be able to afford an upgrade. It's time to build a new computer this year to replace my E8400 box. So IVB + the best price/perfromance video card + 240/256GB SSD is what I want for this spring.
|
|
#
¿
Feb 9, 2012 19:06
|
|
|
Agreed posted:That tech enthusiasts are speculatively interested in the 28nm process shrink competition between the two modern major graphics providers comes as a surprise? The only difference is the last time this happened, AMD launched a far more affordable price. The HD5870 was faster than the GTX285 [new process vs old process] and then was subsequently slightly slower again than the 480 once it finally launched. Nvidia refreshed that part faster than AMD refreshed their part, and the 580 pleasently surprised people, whereas the 6970 was somewhat underwhelming [perception is reality]. This time AMD has the best overall part again, new process vs. old process. But instead of great 'bang for buck' they just priced it slightly higher than their competetion. That is where people are disastisfied. The pricing. Those people are dumb though. Why wouldn't you want AMD to get as much profit out of these chips while they can. In the long run that is better for all consumers.
|
|
#
¿
Feb 9, 2012 19:49
|
|
|
Alereon posted:On the other hand, the Radeon HD 7570 1GB does seem like a pretty good option, given that it's only slightly more expensive than the Radeon HD 6670 1GB GDDR5 and will thoroughly kick its rear end with lower power consumption. It's a decent upgrade option for OEM systems with weak power supplies too. In addition it's looking like the switch to GCN is a smart one over VLIW-5. http://ht4u.net/reviews/2012/amd_radeon_hd_7700_test/index4.php With 7770 at 5770 clocks, it has a 25% less theoretical GFLOP's than the 5770 does. Even with that, it generally outperforms or is the same performance than the 5770. GCN looks really good, but the pricing for these products is off. Eventually I suspect the 68xx's will all be sold out and this card can settle down in price. Probably 115-120 for the 7770.
|
|
#
¿
Feb 15, 2012 21:03
|
|
|
Star War Sex Parrot posted:Cross-posting because this is a drat joke. Agreed. They beat nvidia to market by 4 months [at least] and yet it looks like it was too much for the driver team. GCN is the second 'new' arch from them in a year. I know VLIW-4 wasn't as big a change from VLIW-5 as GCN is, but still, they've been tasked with a big load. I think they will figure it out and get it right. I feel like improving their drivers is going to increase performance a good amount.
|
|
#
¿
Mar 1, 2012 18:32
|
|
|
Aleksei Vasiliev posted:FXAA also blurs everything on the screen slightly, though. If the developer implements FXAA I believe it is possible that the text can remain untouched by the process.
|
|
#
¿
Mar 8, 2012 20:44
|
|
|
Peechka posted:I just replaced my 8800GS that I had in my crappy Dell system, and this card pretty much kept up until SWTOR, but that game has a lovely engine so its partly the games fault. And were talking about a 5 yr old $130 card. I replaced that with a $135 ATI 6850, got about a 100% performance increase and fully expect this to last at least 3-4 yrs. What resolution is your monitor? I have a 5850 with an i5-750 OC'd to 3.6ghz and SW:TOR ran pretty good [45-60 fps] with medium+ settings on 2048x1152. If you're rocking a CRT monitor with a low resolution your statement makes sense. if you have very very low standards for what kind of graphic fidelity you require your statement makes sense.
|
|
#
¿
Mar 8, 2012 22:39
|
|
|
It's the 19th. Availabiilty of the 78xx's is basically 0 right now. Seems like if it was anything other than a logistic's problem [perhaps shipments to retailers were delayed?] I would have expected some sort of announcement postponing the launch. I suppose it could be TSMC problems that Charlie claims exist. Really suspicious that there are none available though.
|
|
#
¿
Mar 19, 2012 20:15
|
|
|
This is a pitiful launch if 1 card from 1 vendor is all the 'stock' that can be managed.
|
|
#
¿
Mar 20, 2012 00:03
|
|
|
Images from Tom's Hardware review of the 680 leaked and it appears as if nvidia has finally out engineered AMD. The die size is still in question, but all the rumors indicate that its about 40 - 60mm2 smaller than Tahiti [79xx], meanwhile it tends to be 15% or so faster than Tahiti and consume less power while doing so. This is great news to me, because if nvidia prices this GPU at around $500, then AMD will have to lower the 79xx considerably. That *should* make prices cascade and make the midrange parts cheaper. I can at least hope.
|
|
#
¿
Mar 20, 2012 17:04
|
|
|
Crackbone posted:Link? http://imgur.com/a/RCDqK Obviously Tom's is probably the least reputable benchmarking site, but it seems that the 680 is going to be a very good performer.
|
|
#
¿
Mar 20, 2012 17:08
|
|
|
Crackbone posted:Assuming those are real, looks good. Of course, it all depends on retail pricing, and for most people what Nvidia does for the mid-range market. Yep. Apparently they were up on the real website for a while, someone was smart and downloaded them all before they were pulled down. I think they are real. In recent generations AMD has had a lot more performance per mm2 and generally more performance per watt. So the fear was that nvidia had to postpone the GK100 [big chip] and wouldn't be able to compete with AMD without having their large monolithic die. Rumors of a 300 - 320mm2 GK104 led me to believe that they would be competing with Pitcairn, so Nvidia really has made a big turn around here. This reminds me of the HD4870, which surprised everyone, including nvidia. I know nvidia won't be pricing this chip at those kinds of low prices, but at the very least it should force AMD to sell GPU's for less than they are now. I suppose it's possible that nvidia could sell the GK104 for like, $600, and AMD wouldn't have to move prices, but I think it's more likely to launch at $500, and then force AMD to drop the price of the 7970 down quite a bit. In the midrange I think the 7870 is a great looking chip, but it's about $50 too much, otherwise I would have ordered it today. Maybe the cascade of price cuts that GK104 *should* cause will trickle down to the 7870. 
|
|
#
¿
Mar 20, 2012 17:22
|
|
|
Gwaihir posted:Wasn't the 680 supposedly the equivalent of the situation we had with the 480/580, in that the 480 was a sorta rushed to market chip that didn't meet the original shader core count/whatever due to production issues? Which were then fixed and revised in to the 580 which was the same architecture but with the outrageous power usage fixed and the original core count intact. Yes, the 680 is the GK104 part. This is 300 - 320mm2 by the best estimates from the 'die shots' we've seen. The larger GK100/GK110/GK112 [whatever they call it] won't be out until August at the earliest. technically this part is the upgrade to the GF114 part, but it's performance is at least equal Tahiti, perhaps even 10 - 20% better. Of course I think AMD will release either a super clocked version, a 'performance driver' or perhaps both to compete with the 680, and the reality is both will still have a higher priced card. The other factor here is availability. No matter how fast these things are if they are always out of stock it won't impact current pricing much, if at all.
|
|
#
¿
Mar 20, 2012 18:02
|
|
|
http://forums.overclockers.co.uk/showpost.php?p=21514515&postcount=61 He's not giving details and seems to be basing his opinion purely on synthetic benchmarks... however, this is looking very promising. If it's selling for $500, then that's going to force competition.
|
|
#
¿
Mar 20, 2012 18:54
|
|
|
the GTX 680 is not a good GPGPU peformer. Unless it's driver problems, it's getting beaten by the 7870 from what I have read.
|
|
#
¿
Mar 21, 2012 15:55
|
|
|
I ordered one. I've never bought a top tier card. This is a pretty compelling one, with really nice features like Adaptive V-Sync and TXAA. It's generally faster than AMD's fastest while being slightly slower. It's massive overkill for me @ 2048x1152, but it will be the first time I can play games with the graphics settings set to max all the time. I'm pretty stoked. First Nvidia GPU I've bought since Riva 128. In a dell I had a 8800 GT for a while also I think.
|
|
#
¿
Mar 22, 2012 16:33
|
|
|
Gwaihir posted:With how these benches turned out, performance on the GK110 should be really exceptional. Not to mention the inevitable dual chip 690 or whatever they end up calling it. With the much lower power envelope compared to the 580, they should be able to do an actual decent job of it, instead of the 590 which had to sacrifice so much clockspeed to keep the thermals in check. The GK110 or whatever it will be called is likely going to be a very GPGPU focused chip. No doubt it will have better games performance, but I suspect a lot of the die size increase will go to increasing DP performance.
|
|
#
¿
Mar 22, 2012 16:51
|
|
|
Factory Factory posted:According to AnandTech, there's an even-higher-end single chip Kepler forthcoming. The 680 GTX is the current single-chip king, but it won't be the single-chip king of this generation. I think by the time big Kepler launches, GCN 2.0 will be out.
|
|
#
¿
Mar 22, 2012 19:26
|
|
|
I bought it too. First $500 video card I've ever had. I'm building an IVB box and this will complete it. I'm probably going to overspend and get the i7-3770k as well. The computer will be with me 4 years. The $230 difference between what I SHOULD get and what I WANT to get [should get the i5-3570k and HD7870] is like $4.70 a month over that period of time. I can't say no to what I want for such a small difference in money over the length of time i'll own these cards. Also, given that the development cycle for GPU's seems to have slowed down, I'm pretty sure that the 680 will really be very solid over that length of time. We will get at least 1 more generation on 28nm for sure, and I wouldn't be surprised if we are on 28nm for even a bit longer than we were on 40nm. Anyway, here comes maxed out adaptive V-synced TXAA2 graphics to my humble 2048x1152 monitor [in about a month]. I'm pretty excited. 
|
|
#
¿
Mar 23, 2012 00:29
|
|
|
HalloKitty posted:With that, AMD are back in the game What's interesting is that the 7970 is more expensive [or rather it should be] than the GTX 680. 3GB of RAM, larger PCB, more expensive VRM components [I'm pretty sure on that one], larger die size. If their yields are good then i'm sure they are still profitable at $480 each, but I'm really surprised they haven't come out with a 1.5GB version of the card yet.
|
|
#
¿
Apr 16, 2012 17:55
|
|
|
I would love a Trinity laptop ultra book style. Where are they? 14" monitor, 1600x900 resolution, Trinity APU, 128GB SSD, about 3.5 lbs. Why isn't this a thing? Trinity's CPU is plenty fast for browsing the internet and office software type products. Trinity's GPU is faster than an HD4000. It's really frustrating because it's actually a perfect product for this market IMO. Where are these products?
|
|
#
¿
Jun 23, 2012 00:14
|
|
|
This is probably the wrong thread, but I'm not sure that there is a right thread for this question, and the conversation recently has become highly technical, so I'm hoping some eyeballs here can answer this question. How much do we know about the 64b Denver SoC? From a high level, does it sound a little bit like what intel does with x86 on their CPU's, [x86 being CISC, but internally the instructions are broken down into smaller bits and micro-ops]. ARM's ISA is already RISC so I'm not sure that the two ideas are comparable, but I'm just looking for someone to explain in a way I can understand what exactly nvidia thinks it's gaining by doing this transmetaish implementation of the ARM v8 ISA. My best guess from reading about the Denver 64b SoC is that they are moving to an in order execution pipeline and that they are saving more power that way (over OoO) than they are spending on the extra translation step done internally, and that this results in overall better perf/watt. But that just seems impossible to me, like cold fusion.
|
|
#
¿
Sep 3, 2014 22:35
|
|
|
|
| # ¿ Apr 26, 2024 15:39 |
|
|
Dante80 posted:No. There is no reason at this point in time to choose a low frequency 8c/16t processor for gaming. I agree with this. But if someone didn't game and wanted to have hobby time editing/exporting dozens of photos in Lightroom and edit/exporting videos in Premier, then would I be correct in thinking that 1700x or 1800x are pretty good options?
|
|
#
¿
Mar 2, 2017 19:15
|
|