I was going to buy a 4890 for £125 but the was i wanted ran out of stock should i spend a little bit more and get a 4890, or a little bit less and get a 5770? thanks
I was going to buy a 4890 for £125 but the was i wanted ran out of stock should i spend a little bit more and get a 4890, or a little bit less and get a 5770? thanks
For outright performance in DX9 and DX10 game the HD4890 is a better choice but if you want much lower power consumption and DX11 features the HD5770 is the choice - the HD5770 has performance around that of an HD4870. The HD4870 and HD4890 do also need a more powerful PSU so you will need to have a look at this too.
i dont mind about power consumption, are the DX11 features worth the performance drop?
It depends on whether you want to be able to use tessellation in games or not. TBH,the best thing would be to look at the improvements DX11 brings to the table over DX10 and consider whether these are worth it to you!
This article may prove useful with regards to tessellation:
http://www.hardocp.com/article/2009/..._tessellation/
I would also read this article:
http://www.anandtech.com/video/showdoc.aspx?i=3507
The 5700 is a pooch performance-wise. Your choice here comes down to whether you value power consumption or gaming performance.
I do wish you'd get off your high horse about how "bad" the 5770 is, Rollo. Of all the stuff you spout on here this is the one thing that makes you sound most fanboyish. It's a good performing midrange card with a much better featureset than the equivalent launch cost card from the previous generation. At the minute it is overpriced because it's the newest release from *anyone*, but given it performs between a 4850 and a 4870 it is *not* a "pooch" - now get over yourself.
Last edited by scaryjim; 16-12-2009 at 11:16 AM.
In most games the HD5770 is actually around the same speed as an HD4870 1GB:
http://www.hardocp.com/article/2009/...o_card_review/
Last edited by CAT-THE-FIFTH; 16-12-2009 at 12:08 PM.
Not to mention that in the Hexus review it beats out the GTX260 in Crysis, Far Cry 2, GRID and HAWX...
Personally I'm waiting for Rollo to tell me that the 5770 is crippled by its 128bit memory bus, so I can point him in the direction of the Hexus review where a 17% overclock on memory (i.e. 17% increase in memory bandwidth) only yeilded a 4.5% improvement in framerate. Hardly crippling, is it...
From what i gather the 128bit memory plus GDDR5 gave the 4770 performcance similar to the 4850 with 256 and DDR3. I think the reduced memory bus makes the cards cheaper to produce due to smaller chipsize. I'm no expert though. The larger bus on NV cards is one of the reasons why they are more expensive to produce.
Of course the jump frrm the 8800gtx (384) to the g92(256) chips did not really cripple performance (ok it ruined the expected performance increase) but made the cards more affordable.
Of course please correct me if I'm wrong! Any gittish/smug/annoying marketing splurge will definately get a
You're quite right that a smaller bus makes the card cheaper to produce, but not quite for the right reason: the memory chips themselves interface with the system over a 32bit bus (usually in pairs), so the lower your memory bus width the less memory chips you have to fit on the card. A 128bit bus requires just 4 memory chips, as opposed to 8 for a 256bit bus. DDR5 memory is capable of transferring twice the amount of data per transfer than it's specified bus width (I'm quite sure how though - it's not officially quad-pumped, soit only makes 2 transfers per clock cycle, but those transfers can somehow be twice the size of its memory pathway!? Either way, the effective bandwidth is twice that of equivalent DDR3) so 128bit DDR5 @ 1000MHz has the same bandwidth as 256bit DDR3 @ 1000MHz. Of course, DDR5 clocks a little higher (ATI run it in the 1100 - 1200MHz region) so the bandwidth is slightly greater than a 4850s.
It's worth noting that despite its double-helping of bandwidth (DDR5 on 256bits) the 4870, where it is faster than the 5770, is fairly consistently ~ 10% faster, and Hexus made 4.5% that back with a 17% memory overclock. Clearly, the GPU is held back slightly by its lower memory bandwidth, but not enough to warrant the extra expense of doubling the memory path.
I personally wonder if they could've hit a happy medium with a 192bit interface, but ATI don't seem to like working outside powers of 2 for their memory (unlikely NVidia!), and there's no knowing how much cost that would've added to the card - would it be worth another £15 at retail for a 10% performance boost? Plus they'd've had to choose between 768MB and 1.5GB for the total card memory, whereas sticking to a 128bit interface lets them use 1GB...
Domestic_Ginger (16-12-2009)
I was looking at going for a 4890 but they do get a bit hot.
□ΞVΞ□
The 5770 has the same core clock, the same number of ROPs, the same number of texture units, the same amount of RAM, and runs its RAM faster than a 4890. As the stream processor clusters and texture mapping are next generation, the fact that the 5770 runs games 10-20% slower than the 4890 only leaves three possibilities I can think of:
1. The 128 bit memory access cripples it's performance compared to the 256 bit access on the 4890.
2. Th SPs or texture units of ATi's current gen are inferior to their last gen. (I doubt this, usually things move forward)
3. The shift to DX11 inclusive drivers has hamstrung the the 5770's performance, perhaps by less efficient use of the compiler for their VLIW arch. (as I suggested may be the case in the thread about their DX11 sales)
The last is possible, but I doubt it as the "new" ATi arch isn't very different from the "old" ATi arch.
Of course there is one other possibility I haven't mentioned because it wouldn't be strictly "ethical" on ATi's part:
It could be they've intentionally crippled 5XXX series performance until the NVIDIA products launch, at which point a miracle driver will significantly boost performance. This would not tip NVIDIA's hand to the true performance of the 5XXX series, perhaps make them less aggressive with clock speeds. (and deprive their customers of possible performance in the meantime)
In any case, those are the possible reasons for the 5770's lower performance across the board. I've been going with the memory bandwidth as the most likely in my opinion. If you have an alternative I have not considered, I'd be happy to change my opinion if it seems more reasonable than the above.
BTW- the fact the performance on the the memory overclock went up almost 5% tells me the memory is part of the problem, not that it isn't. 17% isn't a high enough OC to make up the bandwidth deficit, a true test of the theory would be clocking the 128 bit RAM high enough to yield the same bandwidth as the 256 bit RAM. I think both cards would perform the same if you did.
You mean, they might have copied nVidias normal non-ethical antics of crippling drivers until the right moment?
I doubt it though. I think as others have been trying to tell you over and over...5770 is a MID RANGE CARD. It is not a replacement for the 4870/4890.
But seriously, do not let the truth get in the way of your constantly mud spreading. It's not like this is the 3rd or 4th time someone here has pointed out the reason above to you.......
Main PC: Asus Rampage IV Extreme / 3960X@4.5GHz / Antec H1200 Pro / 32GB DDR3-1866 Quad Channel / Sapphire Fury X / Areca 1680 / 850W EVGA SuperNOVA Gold 2 / Corsair 600T / 2x Dell 3007 / 4 x 250GB SSD + 2 x 80GB SSD / 4 x 1TB HDD (RAID 10) / Windows 10 Pro, Yosemite & Ubuntu
HTPC: AsRock Z77 Pro 4 / 3770K@4.2GHz / 24GB / GTX 1080 / SST-LC20 / Antec TP-550 / Hisense 65k5510 4K TV / HTC Vive / 2 x 240GB SSD + 12TB HDD Space / Race Seat / Logitech G29 / Win 10 Pro
HTPC2: Asus AM1I-A / 5150 / 4GB / Corsair Force 3 240GB / Silverstone SST-ML05B + ST30SF / Samsung UE60H6200 TV / Windows 10 Pro
Spare/Loaner: Gigabyte EX58-UD5 / i950 / 12GB / HD7870 / Corsair 300R / Silverpower 700W modular
NAS 1: HP N40L / 12GB ECC RAM / 2 x 3TB Arrays || NAS 2: Dell PowerEdge T110 II / 24GB ECC RAM / 2 x 3TB Hybrid arrays || Network:Buffalo WZR-1166DHP w/DD-WRT + HP ProCurve 1800-24G
Laptop: Dell Precision 5510 Printer: HP CP1515n || Phone: Huawei P30 || Other: Samsung Galaxy Tab 4 Pro 10.1 CM14 / Playstation 4 + G29 + 2TB Hybrid drive
Whether it's a mid range card has nothing to do with it, the things I listed are the only indicators of performance I know of. Please discuss the issue and technology, not what the card is marketed as.
The question is whether a 4890 that performs 10-20% better for about the same price is a better buy than a 5770 that performs 10-20% less for about the same price.
No one buying a card cares one bit about what the card is marketed as, they care about performance, price and usable feature set.
Price isn't indicative of range or competition slot. Market forces dictate price (demand, supply, cost of production, etc). Right now demand far far exceeds supply, and the cost of production isn't exactly cheap either thanks to TSMC's 40nm fab fail. Apples to Oranges, mate. You can argue that the 4890 is currently better bang per buck, but that doesn't mean people who buy 5770's are stupid, either.
Part of the feature set is low power. This could be implemented at the cost of performance. The power budget of the 5xxx series is without doubt, the best I've seen for years.
The fact that the 5770 can perform this well within this power budget is significant, and more important to many people who are looking for a mid range card than features such as [CENSORED] or [CENSORED].
I'd therefore argue that the way in which a card is marketted is therefore VERY important.
[CENSORED by intent. I don't want to get started on yet another fruitless ping pong discussion on 3d Vision and PhysX.... Aw crup now I've done it]
Corsair Air 540, Asus Prime X570-Pro, Win 10 Pro, AMD R9 3900X, Corsair HX 750, EVGA 1080 Ti, 2x Corsair 2TB MP600, 2x 2TB WD20EZRX, 4x8GB Corsair Dominator, custom watercooled (single loop, 2 rads)
Corsair 550D, Asus X470-Prime Pro, Win 10 Pro, AMD R7 2700, Corsair RM750i, Asus GTX780 Poseidon, 2x Sammy 500GB 970 EVO, 2x 2TB Seagate Barracuda, 2x8GB Corsair Vengeance, custom watercooled (single loop, 2 rads)
Synology DS918+ w/ 2xWDC Green 3TB + 2x Seagate Barracuda 6TB, N2200 w/ 2xSammy 1.5TB
backup:
Corsair 500R, Gigabyte GA-Z97MX Gaming 5, Win 10 Pro, i5 4690, Corsair HX750i, Sapphire Fury X, 256GB Sammy SM951 M.2 (System), WD SE16 640GB, 2x8GB Corsair Vengeance, Corsair H100i
There are currently 1 users browsing this thread. (0 members and 1 guests)